Journal Archive

Johnson Matthey Technol. Rev., 2023, 67, (1), 65
doi: 10.1595/205651323X16692809325480

Reliable and Traceable Temperature Measurements Using Thermocouples

Key to ensuring process efficiency and product consistency

  • Frank Edler
  • Physikalisch-Technische Bundesanstalt (PTB), Abbestraße 2–12, D-10587, Berlin-Charlottenburg, Germany
  • Email:

Received 1st June 2022; Revised 16th August 2022; Accepted 30th August 2022; Online 12th January 2023

Article Synopsis

Temperature is the most frequently measured process variable in almost all industrial sectors from the chemical industry to glass and ceramics, refrigeration and power generation. During many manufacturing processes, continuous temperature control is an important part of product quality assurance and a matter of avoiding malfunctions or detecting them at an early stage. Measuring points can be located at different places such as in containers, pipe systems, machines, ovens or reactors, whereby different gaseous, liquid or solid media, for instance, steam, water, oil or special chemical substances may be involved. In view of these extremely complex tasks, flexibility is one of the most important requirements for measurement technology and signal processing. And this is where thermocouples, which can be adapted to almost all measuring tasks due to their simple design, become relevant. The basic design and operating principle of thermocouples are described in this paper; issues relating to calibration, traceability and measurement uncertainty are addressed. Recent developments to improve temperature measurement with thermocouples are presented. New, drift-optimised thermocouples, novel designs and alternative calibration methods are described, and their advantages over conventional thermocouples or calibration methods are specified.

1. Introduction

The accurate control of temperature is the key to ensuring process efficiency and product consistency in industrial production processes. One basic requirement in this respect is reliable temperature measurements which are traceable to the International Temperature Scale of 1990 (ITS-90) (1) and have low measurement uncertainty and high reproducibility.

Thermocouples are the most widely used temperature sensors. Their basically simple design, their practicability, as well as their wide range of application (–270°C–2700°C) are reasons for their use in industrial temperature measurement and control technology. Furthermore, existing documentary standards (24) defining the temperature-electromotive force (emf) characteristics of different standard thermocouple types ensure that, for most applications, thermocouples of known quality and characteristics are readily available. The principles of thermocouple thermometry were outlined by William Thomson (Lord Kelvin) in the mid-19th century. He established the relationship between the principle thermoelectric effects discovered by Seebeck (1821) and Peltier (1834) and verified the effect now known as the Thomson effect. A general overview of thermocouple properties and applications can be found in (58).

The main problem in accurate temperature measurement with thermocouples are structural changes in the thermocouple wires due to external mechanical influences and even by thermal stress alone. This leads to local changes of the Seebeck coefficient (one refers here to thermoelectric inhomogeneities), which result in drift effects and consequently in erroneous temperature measurements.

This article briefly describes the basic design and the operating principle of thermocouples in Section 2. Section 3 addresses issues relating to traceability, calibration and measurement uncertainty, while Section 4 presents recent developments to improve temperature measurement with thermocouples by reducing drift effects.

2. Design and Operating Principle

Due to the variety of possible measuring tasks and different applications, thermocouples have various designs. Depending on the type of insulating material and sheathing, three major categories of thermocouples can be distinguished: thermocouples with inflexible hard-fired multi-bore ceramic insulation (commonly used as reference thermocouples), mineral-insulated metal-sheathed (MIMS) thermocouples (most versatile assemblies as they are flexible and provide protection from oxidation and contamination), and thermocouples with flexible plastic or woven fibre insulation materials (8).

Nevertheless, the general and essential components of a thermocouple can be identified. A thermocouple consists of two thermoelectrically different conductors (the thermoelements), which are connected to each other at one end (the measuring junction). The thermoelements are electrically insulated against each other by using suitable materials. At the other end, known as the reference junction, the thermocouple wires are connected to compensating leads or equal copper lines. A protection tube completes the thermocouple. Ceramic or metallic materials are often used here to protect the wires from influences from the measuring environment and to ensure minimum mechanical stability. If the junctions in this closed circuit are at different temperatures T1 and T2, a direct current (DC) voltage VT = f(T1, T2) is produced, which can be measured at any point in this circuit. The DC voltage VT, referred to as the emf, is a measure of the temperature difference between the connection points (Figure 1). The emf is only generated in the sections of the thermocouple wires that are exposed to the temperature gradient of a furnace or bath and not at the isothermal junctions. Provided that the thermoelements are thermoelectrically homogeneous, the following applies (Equation (i)):


with S (T) as the Seebeck coefficient (temperature-dependent) and T1, T2 as the temperatures of the measuring and reference junction, respectively.

Fig. 1.

Thermocouple measuring circuit, with A and B being the thermoelements and C representing the copper lines

The thermoelectric voltage of a material is often given against a reference material, usually against platinum. If such thermoelectric voltages are arranged in order, the result is a thermoelectric voltage series. From a multitude of possible material combinations, only a limited number have prevailed in practice: those that best meet the general requirements for the use of thermocouples. The corresponding emf-temperature reference functions of letter-designated thermocouple types are summarised in documentary standards (24).

3. Calibration and Measurement Uncertainty of Thermocouples

In order to obtain reliable as well as spatially and temporally comparable results of temperature measurements, it is necessary to link all individual measurement results via calibrations to a common, stable reference standard. The associated aspects of traceability, calibration and measurement uncertainty are considered below.

3.1 Traceability

Reliable and accurate temperature measurements are essential for the economy, society and the technical and scientific world. Measurement data are used to prove that manufacturing standards are met, and also to verify the quality, properties and functions of products, processes and services. This results in the necessity of calibrating the thermometers used in a traceable way to the kelvin, the International System of Units (SI) unit of thermodynamic temperature, to obtain consistent, comparable and valid measured values. Metrological traceability is the property of a measurement result, whereby the result is related to a reference through a documented unbroken chain of calibrations, each contributing to the measurement uncertainty, see Figure 2. The ‘reference’ in temperature measurement is the definition of the SI unit, the kelvin, through its practical realisation and its embodiment in national and international standards (9). International and national standards may be primary thermometers directly measuring thermodynamic temperatures or thermometric fixed points with well-defined phase transition temperatures. The standards of accredited laboratories are in some cases also thermometric fixed points or standard thermometers (for example, platinum resistance thermometers (PRTs)). Industrial standards can be resistance thermometers, simple fixed points or noble metal thermocouples.

Fig. 2.

Metrological traceability from the SI unit kelvin to the measurement instrument

3.2 Calibration

A calibration comprises an overall set of operations under defined conditions to establish a relationship between the values indicated by the measuring instrument and the corresponding values realised by standards.

The emf-temperature reference functions for thermocouples are defined as mathematical equations, and manufacturers supply thermocouples that largely conform to the reference functions within the specified tolerance ranges. As a rule, the emf varies within the tolerance band with deviations between different wire batches, different wire compositions and different manufacturers. The aim of calibrating a thermocouple is to measure the emf deviations from the reference function at a few temperatures so that corrections can be made and measurement uncertainties reduced. The calibration of thermocouples must provide:

  • The functional relationship between the emf and temperature over the range of interest

  • The uncertainties in temperatures inferred from the relationship

  • The conditions under which the relationship and uncertainties apply.

The usefulness of a calibration should always be considered. Using a thermocouple, even at temperatures that are not too high, can cause changes in chemical compositions and crystal structures. Such alterations are accompanied by changes in the Seebeck coefficient and thus by changes in the emf-temperature relationship, making any calibration meaningless. This is especially true of base metal thermocouples. Noble metal thermocouples are thermoelectrically more stable and homogeneous and allow, to a certain extent, the repeated setting of a defined thermoelectric state as the starting point of a calibration.

The calibration of a thermocouple therefore involves the determination of the thermoelectric homogeneity of the thermoelements. This is in addition to measuring thermoelectric voltages at a sufficiently large number of known temperatures to characterise the emf-temperature dependence of the thermocouple over its range of use. Special calibration methods taking into account the spatial indeterminacy of the temperature-sensitive region of the thermoelements for generating the emf (see Section 3.3) are in situ calibrations and the implementation of self-validating principles. For in situ calibrations, the temperature profile during calibration is identical to the temperature profile in the application.

Noble metal thermocouples are usually calibrated at fixed points of the ITS-90 or by the comparison method against high-precision thermometers, such as PRTs, in baths and furnaces. At temperatures above 1100°C, calibration is carried out by the comparison method against radiation thermometers, at eutectic metal-carbon (M-C) fixed points (10, 11) (see Section 4.3), or by means of the wire-melting or miniature crucible method (1214) at the melting points of palladium (1553.4°C in air) and platinum (1769°C). If necessary, base metal thermocouples are calibrated against more accurate thermometers (such as noble metal thermocouples) by using comparison methods in suitable baths or furnaces. In all cases, the deviations from the corresponding reference values are calculated from the measured emfs at the known temperatures and approximated by low-order functions (first to fourth order) as shown in Figure 3. These deviation functions, together with the respective emf-temperature reference function, describe the individual thermoelectric properties of the calibrated thermocouple.

Fig. 3.

Deviations from a reference function Δemf, approximated by a low-order correction equation

3.3 Measurement Uncertainties

Compared to other thermometers, the sensitive part of a thermocouple is more spatially extended and not fixed at a particular position. As thermoelectric voltage is generated in the sections of the thermocouple wires which are exposed to temperature gradients of the furnace or bath, special care is required in estimating the influence of the sources of uncertainty on the measured emf.

Because of the special importance of thermoelectric inhomogeneities which often contribute most to the overall measurement uncertainty of a thermocouple calibration, their determination will be described in a more detailed manner. Thermoelectric inhomogeneities are abnormal location-dependent changes of the Seebeck coefficient along the thermoelements. Measurement errors occur when thermoelectrically inhomogeneous sections of a thermoelement are exposed to temperature gradients. Inhomogeneities can be caused by impurities, mechanical stresses during usage or manufacturing, unavoidable structural changes when thermoelements are used at high temperatures, or oxidation or changes in the vacancy concentration in thermoelements. A few of the inhomogeneities are reversible and can be at least partially eliminated by annealing. New type S and R thermocouples have relative thermoelectric inhomogeneities of the order of ±0.02%, type B thermocouples of the order of 0.05% and base metal thermocouples have relative inhomogeneities of about ±0.1%. Noble metal thermocouples made of pure metals (gold-platinum and platinum-palladium) are generally more homogeneous and have relative inhomogeneities of ±0.01% (15).

The thermoelectric inhomogeneity of a thermocouple can be determined by measuring immersion profiles. At constant temperatures of the measuring and reference junctions, the position of the thermocouple is changed in a temperature gradient, and the change in the thermoelectric voltage is used as a measure of the inhomogeneity. The temperature used when investigating inhomogeneity effects on the measured emf should not be chosen to be too high so that changes in the thermoelectric properties of the thermocouple during the investigation can be avoided. Uncertainties, uInh, due to inhomogeneity at temperatures (for example, 1000°C) other than the examination temperature (for example, 200°C) are calculated according to Equation (ii):


with ΔemfInh as the maximum thermoelectric voltage difference measured during the test and emf1000 along with emf200 as thermoelectric voltages measured at 1000°C and 200°C, respectively (15).

Further uncertainty contributions in the calibration of thermocouples must be considered and may result from:

  • Parasitic heat fluxes to or from the measuring junction

  • Reduced insulation resistances when measuring high temperatures (factor 10 when increasing the temperature by 200 K)

  • Incorrect reference junction temperatures

  • Slightly different Seebeck coefficients when using compensating or extension lines

  • Errors of the electrical voltage measurement

  • Interpolation errors in evaluation of the calibration results.

4. New Developments and Trends in Thermocouple Thermometry

This section presents information and research results on alternative thermocouple types, alternative thermocouple designs, as well as on calibration procedures to improve the reliability of temperature measurement with thermocouples.

4.1 New Drift-Optimised Thermocouples

Many different thermoelement materials and a variety of combinations of them have been investigated for their suitability for temperature measurement under various conditions, but for most applications only some materials and material combinations proved to be suitable. These are the letter-designated base metal thermocouple types K, N, E, J and T, as well as the noble metal thermocouples S, R and B. These thermocouples have been defined in relevant standards (2, 4), and their origin dates back to approximately 100 years ago. The selection of these thermocouple combinations was often based on features such as their ready availability, the ease with which they were manufactured and the reproducible adjustability of their compositions with resulting thermoelectric voltages that were comparatively high. A good overview of the historical development of the letter-designated thermocouples can be found in Webster (16).

It is noticeable that all letter-designated thermocouple types consist of at least one alloyed thermoelement. The main reasons for the thermoelectric instability and inhomogeneity of these thermocouples are temperature-dependent changes of the crystal structures and different evaporation rates of alloy constituents, which result in changes of the Seebeck coefficient. An improvement of this situation has been achieved in the past by the development of thermocouples made of pure metals (platinum-palladium and gold-platinum thermocouples) and more recently by the investigation of optimised alloy compositions, especially for platinum-rhodium alloyed noble metal thermocouples.

4.1.1 Pure Noble Metal Thermocouples

Pure elements (metals) are thermoelectrically more homogeneous and stable than alloys because they are free from effects caused by having more than one type of atom in the crystal lattice. Such effects are, for example, the selective evaporation or oxidation of one component and the partially more unstable lattice order of alloys. Noble metal thermocouples made of pure metals were developed in the late 1980s (17). The gold-platinum thermocouple is presently the most accurate thermocouple (18, 19). It consists of the two pure noble metals gold and platinum, which are both thermoelectrically homogeneous and stable. It therefore suffers only little from typical temperature-induced changes in thermoelectric homogeneity. With gold-platinum thermocouples, measurement uncertainties can be achieved that approach those of a high-temperature standard PRT (HSPRT), usually about ±0.01°C. In contrast, there are disadvantages that seldom allow its use in industrial practice. Gold-platinum thermocouples are comparatively expensive (about US$10,000), the thermoelements used must be very pure (usually >99.999%) and the maximum temperature is limited to 1000°C due to the melting point of gold at 1064°C. Furthermore, gold and platinum, as well as the ceramic insulation used (pure alumina (Al2O3, 99.7%)), have different thermal expansion coefficients. With changing temperatures, mechanically induced stresses may occur, which change the Seebeck coefficients. Therefore, a stress-compensating coil of thin platinum wire (wire diameter (0.1–0.2 mm) is often welded between the two thermoelements as the measuring junction. On the one hand, this allows the thermoelements to move easily relative to each other in the ceramic insulation tube, but on the other hand, it makes this setup mechanically more unstable.

Second only to the gold-platinum thermocouple in terms of accuracy is the platinum-palladium thermocouple (20), which is also made of pure noble metals. However, the highest purity of palladium commonly available is only about 99.99%, so various residual impurities may affect the emf-temperature relationship both randomly and unpredictably. In addition, the oxidation of palladium occurs between 500°C and 700°C, which affects the Seebeck coefficient via structural lattice changes (21). However, this oxidation process is reversible, and the palladium wire can be restored to its initial state by high-temperature annealing at temperatures above about 1000°C. An advantage of the platinum-palladium thermocouple compared to the gold-platinum thermocouple is the higher application temperature of about 1500°C, which is limited by the melting point of palladium at about 1554°C. The lowest uncertainties achievable with a platinum-palladium thermocouple are of the order of about ±0.1°C, due to the problems of palladium oxidation mentioned above and the effects of impurities. The cost of a platinum-palladium thermocouple ranges from US$5000 to US$7000, depending on the wire grade. Considering the high costs, susceptibility to impurities, and problems similar to those with gold-platinum thermocouples caused by the differential expansion of thermoelements, platinum-palladium thermocouples are also not best suited for industrial applications. Nevertheless, they can be used as reference thermometers to calibrate other thermocouples with high accuracies. As an example of the superiority of the thermoelectric homogeneity of pure metal thermocouples compared to conventional platinum-rhodium alloyed thermocouples, Figure 4 shows the immersion profiles of a type S and a platinum-palladium thermocouple. Both thermocouples were moved through the temperature gradient of a fixed-point furnace when realising the freezing plateau of aluminium at a constant temperature of 660.323°C. The maximum emf change of the platinum-palladium thermocouple is lower than about 0.2 μV, which corresponds to a temperature equivalent of less than 20 mK. For the type S thermocouple, a maximum emf change of about 1 μV was obtained, which corresponds to a temperature equivalent of about 0.1 K.

Fig. 4.

Immersion profiles of a type S and a platinum-palladium thermocouple measured at the freezing point of aluminium (660.323°C)

4.1.2 Optimised Platinum-Rhodium Alloyed Thermocouples for High Temperatures

Within the framework of the EMPRESS (22) and EMPRESS-2 (23) European Metrology Programme for Innovation and Research (EMPIR) projects, investigations were carried out on the thermoelectric stability of platinum-rhodium alloyed thermocouples at temperatures above 1100°C. Previous tests had shown that the high-temperature thermoelectric stability of wires made of platinum-rhodium alloys improves with the mass fraction of rhodium (24). Further investigations showed that the most important factor for the thermoelectric stability and homogeneity of platinum-rhodium wires above 1100°C is the oxide vapour transport, provided that no significant impurities are present (2528). Using a new technique to determine an optimal thermoelectrically stable combination of two platinum-rhodium alloys, the drift behaviour of different combinations was investigated comparatively on the basis of multi-wire thermocouples. The multi-wire thermocouples consisted of up to seven thermoelements of different platinum-rhodium alloys and were continuously exposed to high temperatures for up to 6000 h, as well as being repeatedly calibrated in situ or under repetitive conditions between this high-temperature annealing (29). As an example, Figure 5 shows the emf-drift behaviour of various platinum-rhodium alloys with rhodium contents of 10% or higher during annealing at 1300°C, 1350°C and 1400°C. In addition, these alloys were measured repeatedly at the melting point of Co–C (approximately 1324°C). During the first approximately 200 h, an increase of the emfs occurred for all combinations, which corresponded to temperature equivalents of 1.5 K to 3.5 K. With the exception of the combinations containing the platinum-17% rhodium alloy (red symbols), the repeatedly measured emfs of all platinum-rhodium combinations remained constant within a temperature equivalent of ±0.2 K, regardless of the different high-annealing temperatures up to 1400°C.

Fig. 5.

Drift behaviour at cobalt-carbon (1324°C) of combinations of different platinum-rhodium alloys during about 3800 h of annealing at 1315°C, 1350°C and 1400°C

Taking into account all drift data obtained in the scope of the EMPRESS project and bearing in mind the practically available platinum-rhodium compositions, the combination of platinum-40% rhodium vs. platinum-6% rhodium was the most promising combination for low-drift behaviour of a platinum-rhodium alloyed thermocouple at high temperatures. This combination was therefore used to determine an emf-temperature reference function within the EMPRESS-2 project.

To determine the emf-temperature reference function for platinum-40% rhodium vs. platinum-6% rhodium thermocouples, a total of 10 thermocouples were calibrated. This was undertaken by using comparison methods and at fixed points of the ITS-90 according to uniform and agreed procedures. Based on the results of these measurements, the two platinum-40% rhodium vs. platinum-6% rhodium thermocouples with the most emf-temperature pairs and the highest degree of equivalence of the results were selected to determine the reference function in the temperature range between 0°C and 1769°C (30). The expanded uncertainty of this reference function between 250°C and 800°C is about 0.5 K. It increases to about 1 K at 1350°C, 1.6 K at 1550°C and 3 K at 1769°C. The special feature of the platinum-40% rhodium vs. platinum-6% rhodium thermocouples is that, for the first time, the composition and combination of the thermocouple wires were optimised specifically to ensure only a low drift at high temperatures (T > 1100°C). Therefore, the main application temperature range of this new thermocouple type is at temperatures above 1100°C, where the high-thermoelectric stability becomes particularly apparent. It should be noted that the Seebeck coefficient at temperatures below about 250°C is lower than 3 μV·K–1. Due to this low thermoelectric sensitivity, the use of platinum-40% rhodium vs. platinum-6% rhodium thermocouples is not recommended at temperatures below 250°C as better alternatives are available.

4.1.3 Optimised Platinum-Rhodium Alloyed Thermocouples for Low Temperatures

The well-known platinum-rhodium alloyed noble metal thermocouples (types S and R) are frequently used as reference thermocouples (working standards). They are, however, also found in industrial applications with high-accuracy requirements in the temperature range between 0°C and 1100°C. The measurement uncertainties occurring in routine use are of the order of 0.5°C to 1°C. They are caused by reversible but unavoidable changes in the crystal structure at temperatures below approximately 600°C and also by selective rhodium oxidation above approximately 550°C (31, 32). A reduction of the measurement uncertainty to 0.1–0.2°C is possible in principle but requires regular annealing and proof of thermoelectric homogeneity. This is, however, not possible or too complex for many users. For this reason, there are efforts to develop an intrinsically stable thermocouple to reduce the requirements for regular thermal treatments and to maintain the achievable minimum measurement uncertainty in the temperature range between 0°C and 1100°C.

In a recent study (32), the temperature-induced drift properties of several platinum-rhodium alloys (5%, 13%, 17%, 20%, 30%, 40%) were investigated at temperatures between 150°C and 950°C. This study showed that the drift due to ordering processes below about 550°C and rhodium oxidation above 550°C, as occurs with type S and type R thermocouples and as found for the other platinum-rhodium alloys investigated in this study, was almost non-existent with the platinum-20% rhodium alloy. The various platinum-rhodium alloys were annealed at 1100°C for 2 h and removed (quenched) directly from an annealing furnace. They were then subjected to gradient annealing for several hours in a special tube furnace with a linear temperature gradient (approximately 150–950°C) and then underwent a homogeneity scan at about 100°C by using a water heat-pipe (33). For comparison purposes and therefore only qualitatively, Figure 6 shows the measured location-dependent, and therefore temperature-dependent, changes in the emf of the respective platinum-rhodium alloys against pure platinum after 100 h of annealing in the gradient furnace. The temperature on the y-axis corresponds to the annealing temperature in the gradient furnace. The curve marked by the full triangles is the platinum-20% rhodium alloy.

Fig. 6.

Homogeneity scans of different platinum-rhodium alloys measured against pure platinum

The results of another publication investigating the thermoelectric stability and homogeneity of the thermocouple combination of platinum-20% rhodium vs. platinum using platinum-20% rhodium wires from four different manufacturers (34) confirmed the results of the previous comparative study (32). It also put them on a broader basis. Furthermore, a preliminary emf-temperature reference function was determined in the temperature range between 0°C and 962°C, based on 12 data points (emf-temperature) per thermocouple. With a fifth order polynomial emf = f(T), the residuals were randomly distributed within ±4 μV of the reference function. Accuracies of about ±0.5°C to the silver point can be achieved only by using this reference function. With an additional calibration, the accuracy can be improved to better than ±0.1°C (34).

4.2 New Design of Mineral-Insulated Metal-Sheathed Thermocouples

MIMS thermocouples made of base metals are the most widely used temperature sensors in industrial applications. However, they are not comparable to noble metal thermocouples in terms of the measurement uncertainties that are achievable. The causes of drift effects are similar to those of noble metal thermocouples, but they are more pronounced in base metal thermocouples. The sources of drift effects are temperature-dependent close-ordering processes, changes in alloy compositions (only partially reversible) and impurities from the environment and from the protection sheaths made of metal. Nevertheless, with a few base metal thermocouple types, measurement uncertainties of the order of a few tenths of a kelvin can be achieved in limited temperature ranges or under unchanging measuring conditions. Measurement uncertainties of a few degrees are, however, normally to be expected.

Studies at the University of Cambridge, UK, have shown that the main contamination elements in the thermocouple wires have been manganese and chromium, regardless of the thermocouple type (35). It was therefore assumed that these impurities mainly came from the sheath material (various Inconel alloys: 600, 310, 316, 304) which is used for all MIMS thermocouple types. Depending on operating temperatures and durations, the impurities from the sheath material diffuse through the densely packed MgO insulation into the thermoelements and cause irreversible changes in the composition and thus drift-causing changes in the Seebeck coefficient.

To reduce drift effects, it is therefore necessary to prevent or minimise the transport of impurities into the thermoelements as far as possible. This goal can be achieved by a novel design developed and studied by CCPI Europe Ltd, UK, and the University of Cambridge (35). The key point here is the use of an additional second inner sheath made of materials that are compatible with those of the thermoelements used. A schematic cross-sectional view of the dual-wall thermocouple is shown in Figure 7. In a series of tests with conventional and dual-wall type K and N thermocouples, it was shown that drift reductions of between 75% and 89% could be achieved with the dual-wall thermocouples when used under cycling conditions between 450°C and 1250°C (35). The reduced amount of drift found with the dual-wall thermocouples indicated that the thermocouples could stay within a given tolerance for a longer time, therefore reducing the measurement uncertainty and allowing extended recalibration intervals.

Fig. 7.

Cross-sectional view of a dual-wall thermocouple

Recent comparative studies of the drift behaviour of conventional and dual-wall type N and type K MIMS thermocouples were carried out within the EMPRESS-2 EMPIR project. For the drift measurements, 48 thermocouples were tested at six different national metrology institutes (NMIs). At each NMI, four conventional and four dual-wall thermocouples were subjected to a homogeneity scan, calibrated at the iron-carbon eutectic fixed point, and then either exposed to isothermal annealing at 1200°C for 500 h or to temperature cycling process between 300°C and 1150°C (36). For the type N thermocouples, a three times lower drift was found for the dual-wall thermocouples compared to the conventional thermocouples during the isothermal treatment at 1200°C. Within the cycling treatment, a five to six times lower drift was observed for the dual-wall thermocouples, which agrees well with the results in (35). For the type K thermocouples, no significant difference in the drift behaviour was observed during isothermal aging at 1200°C. When cycling the type K thermocouples, the conventional thermocouples even showed a lower drift than the dual-wall thermocouples, which is in contrast to previously obtained results (35). The reasons for this discrepancy are not clear but could be due to the use of different batches of thermoelements.

A slightly different approach to minimise drift effects in MIMS thermocouples is described in (37). A new and robust multipoint thermocouple was developed to ensure higher process reliability. It combines a metallic outer sheath as a protective tube and several thinner MIMS thermocouples placed in the inner MgO insulation. The outer sheath forms the first barrier. If there is any leakage, the internal MgO powder might become contaminated. This does not, however, change the accuracy of the temperature measurement, as each of the inner thermocouples has its own additional sheath. This is a second protective barrier for the thermocouple measuring system. The measuring junctions of the internal thermocouples can be arranged at different positions within the measuring system, so that temperature distributions can also be recorded reliably in a simple way.

4.3 Alternative Calibration Methods

4.3.1 Metal-Carbon Eutectic Fixed Points

For high-level calibrations of noble metal thermocouples up to temperatures of 1100°C, fixed points of pure metals are available from the ITS-90. With the development of eutectic M-C fixed points at the beginning of the 2000s (10, 38), the temperature gap was closed between the temperatures of the freezing point of copper (1084.62°C) and the melting point of palladium (1553.4°C in air). Since that time, the M-C melting points iron-carbon (1553.7°C), cobalt-carbon (1324°C), nickel-carbon (1329°C) and palladium-carbon (1491°C) have been used to calibrate noble metal thermocouples in this temperature range. The measurement uncertainty of the melting temperatures is of the order of a few tenths of a kelvin (11), so that the calibration uncertainty in this temperature range has been reduced by approximately a factor of two compared to an interpolation. The further advantage of M-C fixed points is that the carbon of the graphite crucibles is a component of the fixed-point material. Because graphite is available with high purities of 99.999%, the risk of contamination of the fixed-point material by impurities from the crucible material is minimised.

A schematic diagram of a fixed-point cell for M-C eutectics of the Physikalisch-Technische Bundesanstalt (PTB), Germany, used for the calibration of thermocouples is shown in Figure 8. The cell consists of a double-wall graphite cylinder with an axially symmetrically located thermometer well. The double-wall design minimises the risk of failure of the M-C eutectic fixed-point cell during use. Often the fixed-point material bonds strongly to the inner walls, and due to the different thermal expansion coefficients of the fixed-point materials and the graphite, the cells can break and fail. With the double-wall design, cracks in the inner crucible due to the above reasons do not necessarily result in the failure of the fixed-point cell. If further cracks appear in the outer crucible, it can often still be exchanged, and the fixed-point cell can continue to be used.

Fig. 8.

M-C eutectic fixed-point cell of PTB usable for the calibration of thermocouples (dimensions in mm)

4.3.2 In Situ Calibrations

The calibration of thermometers in the installed state, i.e. at the position of the required measurement and under the typical measuring conditions of the application, is referred to as in situ calibration. As in the case of thermometer calibration under laboratory conditions, a known reference temperature TRef with sufficient accuracy for the desired uncertainty of the calibration must also be provided for in situ calibrations in general. To this end, the two different calibration principles, the fixed-point method and the comparison method (see Section 3.2), can be used. A detailed description of applying the fixed-point method for in situ calibrations can be found in another article in this themed issue of Johnson Matthey Technology Review (39). When using the comparison method, a more stable reference thermometer is installed next to the thermocouple under process conditions, and the measured temperature values are compared. If necessary, the measured values of the thermocouple under calibration are corrected.

The general advantages of the in situ calibration of thermocouples (and other thermometers) under process conditions are:

  • The entire measurement chain from the sensor to the transmitter can be checked under process conditions

  • The thermometers to be calibrated do not have to be removed, thus saving time and costs

  • The continuous control of the temperature is possible

  • Especially for thermocouples, in situ calibration is the only calibration method capable of significantly reducing the effects of inhomogeneities.

5. Summary

This article describes recent research activities to improve temperature measurement with thermocouples in industrial applications. In addition to reducing measurement uncertainties, the focus is on increasing the reliability of temperature measurements and extending the service period and service life of thermocouples in use. New, drift-optimised thermocouples, novel designs and alternative calibration methods are described, and their advantages over conventional thermocouples or calibration methods are specified. With platinum-20% rhodium vs. platinum thermocouples, reduced temperature-induced drift effects have been verified at temperatures up to approximately 950°C compared to conventional type S and R thermocouples, which allow temperature measurements with a measurement uncertainty reduced by a factor of two to three. The combination of platinum-40% rhodium vs. platinum-6% rhodium thermoelements was the most promising combination for low-drift behaviour of a platinum-rhodium alloyed thermocouples at high temperatures above 1100°C. An emf-temperature function was established with an expanded uncertainty of 0.5 K (T <800°C), which increases to about 1 K at 1350°C, 1.6 K at 1550°C and 3 K at 1769°C. Tests with dual-wall MIMS thermocouples of type N and K revealed in some cases considerably lower drift rates (up to 89%) in direct comparison with equivalent conventional MIMS thermocouples, so that correspondingly longer service lives can be achieved with dual-wall MIMS thermocouples. The introduction of M-C eutectic fixed points for the calibration of thermocouples bridged the fixed-point gap between the freezing point of copper and the melting point of palladium, so that the calibration uncertainty could be reduced by a factor of two compared to interpolation when using suitable M-C fixed points in this temperature range. Moreover, this paper introduces specific points on traceability, calibration and measurement uncertainty in the use of thermocouples. These result from a particular feature of thermocouples, which is that their temperature-sensitive part is more spatially extended than in other thermometers.


  1. 1.
    H. Preston-Thomas, Metrologia, 1990, 27, (2), 107 LINK
  2. 2.
    ‘Thermocouples – Part 1: EMF Specifications and Tolerances’, IEC 60584-1:2013, International Electrotechnical Commission, Geneva, Switzerland, 2013, 136 pp LINK
  3. 3.
    ‘Temperature – Electromotive Force (EMF) Tables for Pure-Element Thermocouple Combinations’, IEC 62460:2008, International Electrotechnical Commission, Geneva, Switzerland, 2008, 57 pp LINK
  4. 4.
    ‘Standard Specification for Temperature-Electromotive Force (emf) Tables for Standardized Thermocouples’, ASTM E230/E230M-17, ASTM International, West Conshohocken, Pennsylvania, USA, 2017 LINK
  5. 5.
    T. J. Quinn, ‘Thermocouples’, in “Temperature”, 2nd Edn., Ch. 6, Academic Press, Cambridge, Massachusetts, USA, 1990, pp. 286–331 LINK
  6. 6.
    R. E. Bentley, “Handbook of Temperature Measurement: The Theory and Practice of Thermoelectric Thermometry”, Vol. 3, Springer-Verlag Singapore Pte Ltd, Singapore, 1998, 245 pp
  7. 7.
    J. V. Nicholas and R. D. White, ‘Thermocouple Thermometry’, in “Traceable Temperatures: An Introduction to Temperature Measurement and Calibration”, 2nd Edn., Ch. 8, John Wiley & Sons Ltd, Chichester, UK, 2001, pp. 295–342 LINK
  8. 8.
    F. Edler, Y. G. Kim, H. Liedberg, E. Webster and D. R. White, ‘Guide to Secondary Thermometry: Thermocouple Thermometry: 1. General Usage’, Consultative Committee for Thermometry, Bureau International des Poids et Mesures (BIPM), Sèvres, France, 2021, 62 pp LINK
  9. 9.
    Joint Committee for Guides in Metrology, ‘International Vocabulary of Metrology –Basic and General Concepts and Associated Terms (VIM)’, JCGM 200:2012(E/F), 3rd Edn., Bureau International des Poids et Mesures (BIPM), Sèvres, France, 2012, 108 pp LINK
  10. 10.
    Y. Yamada, H. Sakate, F. Sakuma and A. Ono, Metrologia, 1999, 36, (3), 207 LINK
  11. 11.
    F. Edler, Y. G. Kim, G. Machin, J. Pearce and D. R. White, ‘Guide to Secondary Thermometry: Specialized Fixed Point above 0°C’, Consultative Committee for Thermometry, Bureau International des Poids et Mesures (BIPM), Sèvres, France, 2022, 37 pp LINK
  12. 12.
    M. Tischler, and M. J. Koremblit, ‘Miniature Thermometric Fixed Points for Thermocouple Calibrations’, in “Temperature: Its Measurement and Control in Science and Industry”, ed. J. F. Schooley, Vol. 5, American Institute of Physics, New York, USA, 1982, p. 383
  13. 13.
    F. Edler and K. Huang, Int. J. Thermophys., 2016, 37, (12), 126 LINK
  14. 14.
    F. Edler and K. Huang, Int. J. Thermophys., 2016, 37, (12), 119 LINK
  15. 15.
    F. Jahan, and M. Ballico, ‘A Study of the Temperature Dependence of Inhomogeneity in Platinum-Based Thermocouples’, in “Temperature: Its Measurement and Control in Science and Industry”, ed. D. C. Ripple, Vol. 7, American Institute of Physics, New York, USA, 2003, 469
  16. 16.
    E. Webster, Metrologia, 2021, 58, (2), 025004 LINK
  17. 17.
    E. H. McLaren and E. G. Murdock, “The Pt/Au Thermocouple: I. Essential Performance; II. Preparatory Heat Treatment, Wire Comparisons and Provisional Scale”, Publication No. NRCC-27703, National Research Council of Canada, Ottawa, Canada, 22nd April, 2009, 111 pp
  18. 18.
    G. W. Burns, G. F. Strouse, B. M. Liu, and B. W. Mangum, ‘Gold versus Platinum Thermocouples: Performance Data and an ITS-90 Based Reference Function’, in “Temperature: Its Measurement and Control in Science and Industry”, ed. J. F. Schooley, Vol. 6, American Institute of Physics, New York, USA, 1992, p. 531
  19. 19.
    D. C. Ripple and G. W. Burns, “Standard Reference Material 1749: Au/Pt Thermocouple Thermometer”, Special Publication (NIST SP) – 260-134, National Institute of Standards and Technology, Gaithersburg, USA, 1st March, 2002
  20. 20.
    G. W. Burns, D. C. Ripple and M. Battuello, Metrologia, 1998, 35, (5), 761 LINK
  21. 21.
    W. S. Ohm and K. D. Hill, Int. J. Thermophys., 2010, 31, (8–9), 1402 LINK
  22. 22.
    J. V. Pearce, F. Edler, C. J. Elliott, L. Rosso, G. Sutton, R. Zante and G. Machin, ‘A European Project to Enhance Process Efficiency Through Improved Temperature Measurement: EMPRESS’, 17th International Congress of Metrology, Paris, France, 21st–24th September, 2015, EDP Sciences, Les Ulis, France, 2015, 6 pp LINK
  23. 23.
    J. V. Pearce, A. Heyes, A. Fateev, G. Sutton, A. Andreu and G. Machin, “Enhancing Process Efficiency Through Improved Temperature Measurement: The EMPRESS Projects”, National Physical Laboratory, Teddington, UK, 2019, 22 pp LINK
  24. 24.
    F. Edler and P. Ederer, AIP Conf. Proc., 2013, 1552, (1), 532 LINK
  25. 25.
    J. V. Pearce, Johnson Matthey Technol. Rev., 2016, 60, (4) 238 LINK
  26. 26.
    J. C. Chaston, Platinum Metals Rev., 1975, 19, (4), 135 LINK
  27. 27.
    C. B. Alcock and G. W. Hooper, Proc. R. Soc. Lond. A, 1960, 254, (1279), 551 LINK
  28. 28.
    E. S. Webster, A. Greenen and J. Pearce, Int. J. Thermophys., 2016, 37, (7), 70 LINK
  29. 29.
    J. V. Pearce, F. Edler, C. J. Elliott, A. Greenen, P. M. Harris, C. Garcia Izquierdo, Y.-G. Kim, M. J. Martin, I. M. Smith, D. Tucker and R. I. Veltcheva, Metrologia, 2018, 55, (4), 558 LINK
  30. 30.
    F. Edler, J. Bojkovski, C. G. Izquerdo, M. J. Martin, D. Tucker, N. Arifovic, S. L. Andersen, L. Sindelarova and V. Zuzek, Int. J. Thermophys., 2021, 42, (11), 150 LINK
  31. 31.
    E. S. Webster, Int. J. Thermophys., 2015, 36, (8), 1909 LINK
  32. 32.
    E. S. Webster and F. Edler, Int. J. Thermophys., 2017, 38, (2), 29 LINK
  33. 33.
    E. S. Webster, D. R. White and H. Edgar, Int. J. Thermophys., 2015, 36, (2–3), 444 LINK
  34. 34.
    E. Webster, Metrologia, 2020, 57, (1), 014005 LINK
  35. 35.
    T. D. Ford and M. Scervini, “Low Drift Type K and N Mineral Insulated Thermocouple Cable for Aerospace Applications”, TE Wire & Cable LLC, New Jersey, USA, White Paper, 2020, 11 pp
  36. 36.
    D. J. L. Tucker, F. Edler, V. Žužek, J. Bojkovski, C. Garcia-Izquierdo, M. Parrondo, L. Šindelářová and N. Arifovic, Meas. Sci. Technol., 2022, 33, (7), 075003 LINK
  37. 37.
    S. Mella, A. Löffler and M. Schalles, ‘C4.2 Reliable Multipoint Temperature Profiling in Hydroprocessing Units’, SMSI 2021 Conference, Online, 3rd–6th May, 2021, Sensor and Measurement Science International, Wunstorf, Germany, 2021, pp. 199–200 LINK
  38. 38.
    Y. Yamada, F. Sakuma and A. Ono, Metrologia, 2000, 37, (1), 71 LINK
  39. 39.
    J. V. Pearce, D. L. Tucker, R. I. Veltcheva and G. Machin, Johnson Matthey Technol. Rev., 2023, 67, (1), 4 LINK

The Author

Frank Edler graduated from the Department of Physics, Humboldt University of Berlin, Germany, with a degree in crystallography. From 1986 onwards, he worked in Berlin as a scientific assistant at the Office for Standardization, Metrology and Commodity Testing in the field of temperature measurement technology. Since 1990, he has been employed in the same field of work at the Physikalisch-Technische Bundesanstalt (PTB), Germany, with a focus on thermocouple thermometry and noise thermometry. He received his doctorate from the Faculty of Mathematics and Physics at the University of Hannover, Germany, in 1999, where he focused on “Noise Thermometric Determination of the Thermodynamic Temperature of the Palladium Melting Point Using Miniaturized Fixed-Point Cells”. Since 2007, he has headed the Thermoelectrics Working Group in the PTB’s Temperature Department.

Related articles

Industrial Low Pressure Hydroformylation: Forty-Five Years of Progress for the LP OxoSM Process

The Kelvin Redefinition and Practical Primary Thermometry

Find an article