Thin Solid Films 539 (2013) 345–349

Contents lists available at SciVerse ScienceDirect

Thin Solid Films

j ourna l homepage: www.e lsev ie r .com/ locate / ts f

Thin film platinum–palladium thermocouples for gas turbine engine applications

Ian M. Tougas, Otto J. Gregory ⁎Department of Chemical Engineering, University of Rhode Island, Kingston, RI 02881, USA

⁎ Corresponding author at: University of Rhode Island, 20Road, Kingston, RI 02881, USA. Tel.: +1 401 874 2085; fax:

E-mail address: gregory@egr.uri.edu (O.J. Gregory).

0040-6090/$ – see front matter © 2013 Elsevier B.V. Allhttp://dx.doi.org/10.1016/j.tsf.2013.05.076

a b s t r a c t

a r t i c l e i n f o

Article history:Received 21 November 2012Received in revised form 29 April 2013Accepted 1 May 2013Available online 23 May 2013

Keywords:Radio-frequency sputteringPlatinumPalladiumThermocoupleGas turbine engine

Thin film platinum:palladium thermocouples were fabricated on alumina andmullite surfaces using radio fre-quency sputtering and characterized after high temperature exposure to oxidizing environments. The ther-moelectric output, hysteresis, and drift of these sensors were measured at temperatures up to 1100 °C.Auger electron spectroscopy was used to follow the extent of oxidation in each thermocouple leg and interdif-fusion at the metallurgical junction. Minimal oxidation of the platinum and palladium thermoelements wasobserved after high temperature exposure, but considerable dewetting and faceting of the films were ob-served in scanning electronmicroscopy. An Arrhenius temperature dependence on the drift rate was observedand later attributed tomicrostructural changes during thermal cycling. The thin film thermocouples, however,did exhibit excellent stability at 1000 °C with drift rates comparable to commercial type-K wire thermocou-ples. Based on these results, platinum:palladium thin film thermocouples have considerable potential foruse in the hot sections of gas turbine engines.

© 2013 Elsevier B.V. All rights reserved.

1. Introduction

Gas turbine engine components based on advanced ceramics suchas ceramic matrix composites have superior thermomechanical prop-erties at temperatures above 1000 °C when compared to convention-al superalloys [1]. Additionally, ceramic thermal barrier coatings haveenabled gas turbine engine blades comprised of superalloys to oper-ate at temperatures as high as 1500 °C [2], which translates intoboth improved fuel efficiency and reduced NOx emissions. Due tothe severe temperature gradients that can exist in these advancedmaterials, there is a need to integrate sensors into these componentsto monitor their thermomechanical behavior. Thin film sensors can bedirectly deposited onto both stationary and rotating engine compo-nents for health monitoring as well as to verify existing models [1–3].

Thin film thermocouples offer several advantages over conven-tional wire thermocouples or thermal spray instrumentation used ingas turbine engine applications. Thin film sensors do not adversely af-fect gas flow patterns through the engine since their thicknesses arewell below the boundary layer thickness formed on rotating surfaces[3,4]. Thin film sensors also have negligible mass and therefore have aminimal effect on the vibrational modes of rotating components [5].They are more accurate and have faster response times than conven-tional wire sensors (b1 μs) due to their lower thermal mass. Addi-tionally, thin films are not affected by the g forces acting on theblades which are rotating at high velocities [3,6]. Furthermore, thefilms can be directly deposited onto the surface of engine components

5 Crawford Hall, 16 Greenhouse+1 401 874 4689.

rights reserved.

without the need for adhesives, permitting more accurate surfacetemperature measurements than wire based sensors [3].

Wire thermocouples have been employed in engine applicationsfor more than 30 years, however, the presence of oxygen can causeinstability and drift at temperature in metallic thin film thermocou-ples due to the significantly reduced diffusional distances in thesedevices. For example, conventional type-S thin film (platinum:90%platinum–10% rhodium) thermocouples can suffer from selectivede-alloying of rhodium in the film due to the formation of rhodiumoxides at temperatures between 600 and 800 °C [7]. This results in aprogressive, time dependent degradation or drift in the thermoelec-tric output by as much as 5 °C/h as the platinum–rhodium thermo-couple legs undergo microstructural and compositional changes dueto this selective oxidation [3]. This effect can be mitigated by applyingalumina overcoats to protect the thermocouple elements from oxida-tion [3,7]. Therefore, there is a need for more stable thermocouplematerials that do not experience detrimental microstructural changesand selective oxidation issues at high temperature without resortingto protective overcoats.

Platinum–rhodium alloys are currently being employed as comple-mentary thermoelements for wire-based thermocouples and thinfilm-based thermocouples. However, noblemetals such as gold, palladi-um, and platinum do not suffer from selective oxidation processes at el-evated temperatures since they are not alloys. Unfortunately, the use ofgold for high temperature thermocouple applications above 1000 °C islimited due to its low melting point (1064 °C) and dewetting of themetal on oxide substrates at temperatures of 900 °C [8]. Wire thermo-couples based on platinum and palladium have been investigated tosome extent [9–13], however, investigations of thin film platinum:pal-ladium thermocouples have been somewhat limited. Kreider et al. has

346 I.M. Tougas, O.J. Gregory / Thin Solid Films 539 (2013) 345–349

examined the use of thin film platinum:palladium thermocouples onoxidized silicon wafers for radiometric temperature measurements upto 1050 °C [14,15]. However, these thermocouples could only measuretemperatures up to 850 °C and required the use of a titanium bondcoat between the silicon wafer and thermocouple elements due todewetting issues [14,15]. The use of thin film platinum:palladiumthermocouples for higher temperature applications has not beenpreviously investigated.

In this study, platinum:palladium thin film thermocouples weredeposited onto ceramic substrates, including alumina and mullite,using r.f. sputtering. The thermoelectric output and drift were mea-sured at temperatures up to 1100 °C and these thin film thermocou-ples exhibited remarkable stability at high temperature in anoxidizing environment. Auger electron spectroscopy (AES) was usedto characterize the chemical composition of the metallurgical junctionsas a function of depth and tomonitor the extent of oxidation of the junc-tion. The microstructure of the thermocouple legs and metallurgicaljunctions was examined using scanning electron microscopy (SEM).

2. Experimental details

Both alumina and mullite (CoorsTek, Inc.) were employed as thesubstrates for all thin film platinum:palladium thermocouples. Thesubstrates (190 mm × 25 mm × 2 mm) were cleaned with acetone,methanol, and deionized water. A dry film negative photoresist(DuPont™ MX 5050™) was applied to the surface of the substratesand was soft-baked prior to patterning thin film thermocouples.After application of the photoresist, the thin film thermocouple pat-terns were transferred to the resist using an Optical Associates, Inc.aligner (350 nm wavelength ultraviolet light) to expose the resist.The photoresist coated substrates were developed to reveal the de-sired pattern and placed inside a Materials Research Corporationmodel 8667 sputtering system to deposit the metal thermocouplelegs. Platinum and palladium films were deposited in pure argon. Abackground pressure of 2.7 × 10−4 Pa was achieved in the sputteringchamber prior to sputtering and the stage (work piece)wasmaintainedat temperatures below 100 °C using water-cooling. Table 1 lists thesputtering parameters employed to deposit the platinumand palladiumfilms. Deposition resulted in 1.5 μmof platinum and 2 μmof palladium.

All thin films were annealed in nitrogen for 5 h at 500 °C to removepoint defects, including trapped argon, and densify the films. The ther-moelectric output was measured as a function of time and temperatureby placing the metallurgical junction (hot junction) in the hot zone of atube furnace and attaching the cold junction to an aluminum chill blockoutside the furnace. In this way, a temperature difference was appliedalong the length of the beam and the cold junction was maintained ator below100 °Cusing chilledwater. The thermocoupleswere thermallycycled in air to 900 °C at 4 °C/min for several cycles to determine thethermoelectric voltage and the drift rate was determined at 800 °C,900 °C, 1000 °C, and 1100 °C over a 10 h period at each temperature.The temperature difference was continuously monitored using type-K(cold junction) and type-S (hot junction) wire thermocouples and cop-per extension wires were attached to the bond pads of the platinum:palladium thermocouples using a silver paste to acquire the voltage sig-nals. All copper wires and wire thermocouples were connected to anIOTech Personal Daq 54 USB data acquisition system with PDaq ViewPlus© software recording the temperature and voltage signals. AES

Table 1Sputtering parameters used for the deposition of platinum and palladium.

Sputteringparameters

Targetdiameter(cm)

Targetpower(W)

Powerdensity(W/cm2)

Sputteringgas pressure(Pa)

Depositionrate(μm/h)

Filmthickness(μm)

Platinum 10.16 200 3.88 1.20 Ar 0.6 1.5Palladium 12.70 200 2.49 1.20 Ar 1.0 2.0

depth profiles were used to determine the chemical composition ofthe metallurgical junction and identify oxidation products formed onthe platinum and palladium thermocouple legs. AES was performedusing a Perkin Elmer 5500 Multi-Technique Surface Analyzer. A back-ground pressure of 6.0 × 10−7 Pa was established and a 1 × 1 mmarea on the surface of each thermocouple sample was sputter cleanedfor 10 s prior to acquiring each depth profile. SEM was performed onthe same films to analyze the microstructure after thermal cyclingusing a JEOL JSM-5900LV SEM and a 20 kV accelerating voltage.

3. Results and discussion

3.1. Thermoelectric measurements

The thermoelectric response of several platinum:palladium thinfilm thermocouples is shown in Fig. 1a and b. A peak temperature of900 °C was employed at the hot junction during thermal cycling anda temperature difference as large as 750 °C was established alongthe length of the ceramic substrates. The maximum thermoelectricvoltage of thin film thermocouples subjected to a temperature differ-ence of 750 °C was 9.00 mV on alumina substrates and 9.52 mV onmullite substrates. This thermoelectric response was comparable to aconventional type-S wire thermocouple, which typically has a ther-moelectric response of 8.45 mV, when a similar temperature differ-ence is applied [16]. Remarkable stability was observed for the thinfilm platinum:palladium thermocouples when tested over severalthermal cycles. However, the peak thermoelectric voltage at 900 °Cfor the thin film thermocouples diminished slightly after each temper-ature ramp, which most likely results from microstructural changessuch as grain growth, pore growth, and dewetting of the metallicfilms. Hysteresis upon heating and cooling to 900 °C is shown inFig. 2. Here, the thin film thermocouples on mullite exhibited greaterhysteresis than those formed on alumina. However, the hysteresiswas comparable to that of previously studied ceramic thin film ther-mocouples deposited on alumina [2]. Furthermore, ceramic filmshave inherently greater stability in air because they have all the oxy-gen that they can take in their structures.

The Seebeck coefficient and drift rates of the thin film thermocou-ples were determined as a function of hot junction temperature andtime, respectively. The Seebeck coefficient defined in this study isgiven by Eq. (1)

S ¼ −ΔVΔT

¼ − ΔVa−ΔVb

Th−Tc

� �ð1Þ

where S is the Seebeck coefficient given in μV/°C, ΔV is the voltage po-tential difference between the two thermocouple materials (ΔVa −ΔVb), and ΔT is the temperature difference between the hot junction(Th) and the cold junction (Tc). Eq. (2) gives the temperature depen-dence of the Seebeck coefficient for metal thermocouples

S NDð Þ ¼ − π3ND

� �2=3 8k2m � T3eh2

!Aþ 3

2

� �ð2Þ

where ND is the carrier concentration, k is the Boltzmann constant,m * is the effective electron mass, T is the absolute temperature, e isthe electron charge, and A is a transport constant [2]. There is no appre-ciable change in the carrier concentration of these highly conductiveplatinum and palladium films over the temperature range of interest,and it was observed that the Seebeck coefficient increased in a linearfashion over most of the temperature range investigated, as shown inFig. 3. Therefore, the Seebeck coefficient, which was expected to be alinear function of temperature, was confirmed in this study.

Fig. 1. Thermoelectric output of two different platinum:palladium thermocouples on alumina (a) and mullite (b). A peak hot junction temperature of 900 °C was used in each case.

347I.M. Tougas, O.J. Gregory / Thin Solid Films 539 (2013) 345–349

Thermocouple drift was defined according to Eq. (3) below

DR Tð Þ ¼ ΔV Tð ÞV Tð Þref

⋅ TΔt

ð3Þ

whereDR(T) is the drift rate given in °C/h,ΔV(T) is the change in voltageat constant temperature, V(T)ref is the initial voltage, T is the tempera-ture, and Δt is the elapsed time at temperature. Table 2 shows thedrift rate of thin film platinum:palladium thermocouples on aluminaand mullite surfaces at different temperatures. The drift rates of thesethin film thermocouples were less than 1 °C/h in magnitude at temper-atures up to 1000 °C. Microstructural changes in the films during ther-mal cycling may have been responsible for drift since the drift rate hadan apparent Arrhenius temperature dependence. Fig. 4 shows the Ar-rhenius temperature dependence on drift rate for thin film platinum:palladium thermocouples deposited on alumina. This Arrhenius tem-perature dependence was modeled according to Eq. (4) below:

DR Tð Þ ¼ A· exp − EaRT

� �ð4Þ

where A is a constant, Ea is the activation energy, R is the gas constant(8.314 J/mol·K), and T is the absolute temperature. The activation ener-gy associated with the drift rate of the platinum:palladium thin filmthermocouples was 132.61 kJ/mol, as given by the slope of the log ofdrift rate versus 1/T. This activation energy falls between the activationenergies for surface diffusion and volumetric diffusion of platinum andpalladium, which suggests that a combination of surface driven forcesand bulk microstructural changes was responsible for the observeddrift [17]. Processes such as the constrained grain growth and dewetting,which reduce the surface free energy of the films, are the most likely

Fig. 2. Hysteresis upon heating/cooling platinum:palladium thermocouples on aluminaand mullite. Corresponds to second cycle of thermoelectric data in Fig. 1a and b.

causes of drift. Nonetheless, the drift rates of these platinum:palladiumthermocouples were comparable to wire type-K thermocouplesoperating at temperatures up to 1100 °C, confirming that theyhave remarkable stability relative to current commercially avail-able instrumentation.

3.2. AES depth profiling and SEM imaging

The chemical composition of the thin film thermocouples, at themetallurgical junctions, was determined as a function of depth usingAES. Oxide formation on platinum was limited to thicknesses lessthan 1 nm, regardless of the substrate used. Oxygen was present atno more than 5 at.% throughout the cross-section and was likelydetected from the alumina or mullite substrates where dewetting ofthemetallic films had occurred. Oxygen solubility in platinum and pal-ladium is very low (on the order of 0.035 at.% in palladium at4.0 × 104 Pa oxygen partial pressure and 850–900 °C) [18]. Platinumdoes not readily form an oxide at temperatures above 500 °C but in-stead experiences material loss through the evaporation of the metalor one of its volatile oxides, such as PtO2 [19,20]. The loss of platinumdue to evaporation is proportional to the oxygen partial pressure inthe ambient, and is appreciably higher at atmospheric pressure be-tween 1000 °C and 1100 °C, where PtO2 has a partial vapor pressureof 0.05 Pa [19]. Decomposition of the volatile oxides at high tempera-ture can result in the redeposition of the metal onto the original sur-face, so the overall material loss is minimal compared to other noblemetals [19]. Therefore, any oxides that were present on the surfaceof the platinum legs likely formed during cooling from elevatedtemperatures.

Oxide formation on palladium when deposited on alumina waslimited to a thickness of less than 1 nm, which was comparable to

Fig. 3. Seebeck coefficient of platinum:palladium thermocouples on alumina and mulliteas a function of temperature.

Table 2Drift rates of platinum:palladium thermocouples on alumina and mullite at varioustemperatures.

Drift rate (°C/h) 800 °C 900 °C 1000 °C 1100 °C

Alumina 0.18 −0.06 −0.76 −2.06Mullite −1.46 −0.83 – –

348 I.M. Tougas, O.J. Gregory / Thin Solid Films 539 (2013) 345–349

that of platinum films formed on the same surface. Again, oxygen wasstill detected at greater film depths but was likely due to the exposedsubstrate in areas where the metal film had dewetted. Oxygen wasdetected at approximately 30 at.% to a depth of more than 25 nmfor palladium deposited on mullite. However, the formation of con-densed oxide phases on the palladium is not likely; while materialloss from evaporation of the metal or of its volatile oxide, PdO, ismuch more likely [19]. At temperatures greater than 850 °C, surfaceoxides on palladium typically decompose. The vapor pressure of pal-ladium metal at temperatures above 1000 °C exceeds that of PdO, es-pecially at atmospheric pressure, so material loss due to palladiummetal evaporation likely dominates the material loss mechanism[19,20]. Therefore, it is likely that the high atomic percentage of oxy-gen detected in the films is derived from the exposed mullite surfacein the vicinity of the dewetted palladium films. This dewetting phe-nomenon was much more prevalent in the palladium films than plat-inum, as given by the atomic percent oxygen detected for both thinfilms.

Considerable interdiffusion of platinum and palladium at the met-allurgical junctions of the thin film thermocouples was verified byAES depth profiling. Since a complete solid solution of platinum andpalladium will be formed at the junction, solid solutions containingapproximately 90/10 at.% platinum/palladium on alumina and 78/22 at.% platinum/palladium on mullite were not surprising. Theplatinum–palladium phase diagram provided further evidence of solidsolution formation in both metallurgical junctions [21]. Oxygen wasdetected at depths up to 15 nm at both metallurgical junctions; how-ever, the detection of oxygen was most likely from exposed substratesurfaces since it is not likely a metastable oxide film would form atthese depths in either platinum or palladium [19]. A larger atomicpercentage of oxygen in the metallurgical junction was detected onmullite because it experienced a greater amount of dewetting afterthermal cycling than the same junction formed on alumina.

Dewetting in conjunction with faceting of both the platinum andpalladium films was observed in SEM, and these microstructuralchanges were apparent in the thin film thermocouples deposited onboth alumina and mullite. In all cases, dewetting resulted in recrystal-lization and pore growth where the pore edges were faceted,

Fig. 4. Drift rates of platinum:palladium thermocouples as a function of temperature.Note the Arrhenius temperature dependance of the drift rates.

particularly at grain boundaries. Fig. 5 shows the microstructure ofplatinum and palladium films on alumina and mullite. When deposit-ed on alumina, the palladium films contained larger pore sizes alongwith larger grains and more faceting, than observed in the platinumfilms due to differences in the melting temperature of the two metals.Furthermore, the surface recession rate of palladium is four orders ofmagnitude higher than platinum due to surface energy consider-ations [19].

When deposited on mullite, the platinum resembled a contiguousnetwork of high aspect ratio crystals and faceted islands with discretepores. Palladium films deposited on mullite appear to have similardewetting characteristics relative to the palladium film formed onalumina. Due to differences in surface roughness, the topography ofthe palladium film appears different on mullite because it generallycontains larger pores and randomly dispersed smaller islands, whichwere not observed in the films formed on alumina. The differencesin the dewetting of these films result from dissimilar surface energyand surface roughness of the substrates, where the rougher surfacestend to induce more residual stress on films, increasing their surfacefree energy. Therefore, there is a greater driving force for dewettingof platinum and palladium on mullite due to its relatively roughersurface. This is primarily responsible for the different microstructuresobserved on alumina. All microstructures here appeared to be similarto those formed via a vapor phase transport mechanism, in whichcontinuous and fully faceted networks of metal were formedthroughout the film due to the dewetting process [3]. Higher magni-fication revealed faceted striations in the thermocouple legs, whichare also suggestive of the vapor phase transport. Various images ofstriations in the films are shown in Fig. 6a and b.

The microstructure of the metallurgical junctions showed no evi-dence of phase separation since a complete solid solution was formedbetween platinum and palladium after thermal cycling. The formationof faceted pores was evident in the dewetted films; however, the ex-tent to which this occurred in relation to platinum and palladiumwasminimal. This was due to the increased thickness of the overlappingplatinum and palladium films at the junction relative to the individuallegs. If more material is initially present, as in the case of the thermo-couple junctions, the dewetting process has the same microstructuraleffect. Similar to the platinum and palladium films, faceted striationswere observed at the metallurgical junction on both surfaces. Delam-ination of the film was also observed at the metallurgical junction/palladium interface where the junction had receded from the palladi-um film and was likely driven by dewetting of the palladium filmduring thermal cycling. Delamination and dewetting of these filmswould ultimately lead to device failure where, in the limit, the plati-num and palladium films would debond completely, most likely atthe junction/thermocouple leg interface.

4. Conclusion

Platinum:palladium thermocouples were developed to replaceconventional thin film thermocouples based on platinum–rhodiumalloys. These alloys suffer from high drift rates due to instability anddegradation as a result of compositional changes and selective oxida-tion processes during thermal cycling. The platinum:palladium thinfilm thermocouples did not show the detrimental effects of selectiveoxidation and its adverse effect on the chemical stability of the metal-lurgical junction. Therefore, the drift rates observed for these thermo-couples were comparable to conventional type-K wire thermocouplesduring thermal cycling. Thicker (thin film) thermocouples would pro-long the stability of these sensors by inhibiting dewetting and unfa-vorable microstructural changes in the films during thermal cycling.The thermoelectric performance of these thermocouples at high tem-perature makes them promising candidates for implementation inthe harsh environments of gas turbine engines, and given their

Fig. 5. SEMmicrographs using backscatter electron imaging (BSEI) of various thin film thermocouple legs after high temperature cycling: platinumon alumina (a); palladium on alumina(b); platinum on mullite (c); and palladium on mullite (d). Each film exhibited dewetting with distinctly different microstructures due to long term high temperature exposure.

Fig. 6. SEM micrographs (BSEI) of the faceted striations in the thermocouple legs after high temperature cycling: palladium on alumina (a); platinum on mullite (b).

349I.M. Tougas, O.J. Gregory / Thin Solid Films 539 (2013) 345–349

stability, long term sensing at temperatures in excess of 1100 °C willbe possible.

Acknowledgments

The authors wish to thank the NASA Glenn Research Center,Cleveland, Ohio for the support of this work under NASA GrantNNX07AB83A (Aircraft Ageing and Durability Project) and Mr. RobertBalikov in the Chemical Engineering Department at the University ofRhode Island for the fabrication and testing of some of the thermocou-ples used in this study.

References

[1] J.D. Wrbanek, G.C. Fralick, D. Zhu, Thin Solid Films 520 (2012) 5801.[2] O.J. Gregory, M. Amani, I.M. Tougas, A.J. Drehman, J. Am. Ceram. Soc. 95 (2012) 705.[3] G.E. Aniolek, O.J. Gregory, Surf. Coat. Technol. 68–69 (1994) 70.

[4] J.F. Lei, H.A. Will, Sens. Actuators, A 65 (1998) 187.[5] L.C. Martin, R. Holanda, NASA TM-106714, 1994.[6] H. Choi, X. Li, Sens. Actuators, A 136 (2007) 118.[7] K.G. Kreider, J. Vac. Sci. Technol. A 11 (1993) 1401.[8] D. Wang, P. Schaaf, J. Mater. Sci. Mater. Electron. 22 (2011) 1067.[9] K.D. Hill, Metrologia 39 (2002) 51.

[10] M.G. Ahmed, K. Ali, J. Metrol. Soc. India 23 (2008) 225.[11] G.W. Burns, D.C. Ripple, M. Battuello, Metrologia 35 (1998) 761.[12] Y.A. Abdelaziz, F.M. Megahed, M.M. Halawa, Measurement 35 (2004) 413.[13] N.P. Moiseeva, Meas. Tech. 47 (2004) 915.[14] K.G. Kreider, G. Gillen, Thin Solid Films 376 (2000) 32.[15] K.G. Kreider, F. DiMeo, Sens. Actuators, A 69 (1998) 46.[16] Revised Thermocouple Reference Tables: Type S, Omega Engineering, Stamford,

CT. 0[http://www.omega.com/temperature/Z/pdf/z208-209.pdf][17] I. Beszeda, E.G. Gontier-Moya, D.L. Beke, Surf. Sci. 547 (2003) 229.[18] J. Gegner, G. Hörz, R. Kirchheim, J. Mater. Sci. 44 (2009) 2198.[19] H. Jehn, J. Less-Common Met. 100 (1984) 321.[20] J.C. Chaston, Platin. Met. Rev. 19 (1975) 135.[21] S.R. Bharadwaj, A.S. Kerkar, S.N. Tripathi, S.R. Dharwadkar, J. Less-Common Met.

169 (1991) 167.