International Journal of Fatigue 53 (2013) 33–39

Contents lists available at SciVerse ScienceDirect

International Journal of Fatigue

journal homepage: www.elsevier .com/locate / i j fa t igue

Adhesion strength of ceramic top coat in thermal barrier coatings subjectedto thermal cycles: Effects of thermal cycle testing method and environment

Masakazu Okazaki a,⇑, Satoshi Yamagishi a, Yasuhiro Yamazaki b, Kazuhiro Ogawa c,Hiroyuki Waki d, Masayuki Arai e

a Department of Mechanical Engineering, Nagaoka University of Technology, 1603-1 Kamitomiokamachi, Nagaoka-shi, Niigata 940-2188, Japanb Department of Mechanical Engineering, Niigata Institute of Technology, Japanc Fracture and Reliability Research Institute, Tohoku University, Japand Department of Mechanical Engineering, Iwate University, Japane Materials Science Research Laboratory, Central Research Institute of Electric Power Industry, Japan

a r t i c l e i n f o a b s t r a c t

Article history:Received 7 May 2011Received in revised form 22 February 2012Accepted 24 February 2012Available online 14 March 2012

Keywords:Thermal Barrier Coatings (TBCs)Thermal cycleIsothermal exposureElastic modulusAdhesion strength

0142-1123/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.ijfatigue.2012.02.014

⇑ Corresponding author.E-mail address: okazaki@mech.nagaokaut.ac.jp (M

This paper deals with the adhesion strength of ceramic top coat in thermal barrier coatings (TBCs) sub-jected to thermal cycles under several different test conditions. Here the TBC specimens consisting of 8%yttria stabilized zirconia, CoNiCrAlY alloy bond coat and Ni-base superalloy were prepared by plasmaspraying. The isothermal exposure and the thermal cycles were applied to the TBC specimens by severalconditions at high temperatures. A series of the test results clearly demonstrated that the adhesionstrength of the top coat was significantly changed by the application of thermal cycles and by the isother-mal exposure. It was also found that the thermal fatigue damage might be evolved depending on of thetesting method by which the thermal cycles are applied. Some background of these findings were dis-cussed, based on the measurements of elastic modulus, tensile strength, and thermal conductivity ofthe ceramic top coat, as well as both the thermally grown oxide at the top coat/bond coat interfaceand the residual stress in the TBC specimens.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Thermally insulating ceramic coatings, known as thermal bar-rier coatings (TBCs) are essential to improve the performance andefficiency of advanced gas turbines in service at extremely hightemperatures [1–6]. The key role of TBCs is, of course, to protectthe metal substrate from high temperature oxidation and environ-mental attack [1,2]. In general the TBC system consists of at leastthree layers; ceramic top coat, bond coat and metal substrate.

The most critical issue limiting durability of TBCs is spallation ofthe ceramic top coat. Once this type of damage is realized, hot sec-tion components made of superalloy substrate may be overheated,resulting in complete failure. Adhesion strength is a majorparameter characterizing the resistance of the ceramic top coatagainst spallation [7,8], where the strength has been oftenevaluated according to the ASTM standard [7]. In general the adhe-sion strength of top coat is significantly changed during service. (i)Thermal stress, which is promoted by the mismatches in thermalexpansion coefficient and thermal conductivities between the metalsubstrate, bond coat and ceramic top coat, and (ii) the influence of

ll rights reserved.

. Okazaki).

environment, i.e. formation of thermally grown oxides (TGOs) atthe bond coat/top coat interface, are essential factors [9,10,12–14]. The interaction between both factors is also important in somecases [12–14]. In the actual TBCs in service the thermal cycle failureis often a critical issue to be concerned, to which all of the above fac-tors commit. Nevertheless, the basic understanding of failure mech-anisms and their interaction still is still on the way of research.

It is an objective of this work to get basic understanding oneffects of thermal cycle testing method on the thermal fatiguedamage of TBCs, through the measurements of both the mechani-cal and physical properties of top coat and the remaining adhesionstrength.

2. Experimental procedures

2.1. Preparation of specimens

The TBC specimens consisting of three layers; Ni-base superal-loy, bond coat and top coat, were prepared in the present work.Here, an 8 wt.% yttria partially stabilized zirconia (YSZ), METCO204NS, and a CoNiCrAlY alloy, AMDRY9951, were selected as thetop and bond coat materials, respectively. The chemical composi-tions of the powders used are listed up in Table 1. The substrate

Table 1Chemical compositions of the powders used (wt.%).

(a) Top coating powder

ZrO2 Y2O3 HfO2 MgO SiO2 TiO2 CaO

90.5 7.5 1.6 <0.01 0.04 0.11 0.02

(b) Bond coating powder

Co Ni Cr Al Y

38 32 21 8 0.5

Fig. 1. Geometry of specimens used. (a) Adhesion strength, (b) residual stress, and(c) thermal conductivity.

34 M. Okazaki et al. / International Journal of Fatigue 53 (2013) 33–39

material employed was a polycrystalline Ni-base superalloy, MarM247.

The geometries of TBC specimens used are illustrated in Fig. 1.These TBC specimens were fabricated as follows: after sprayingthe bond coat alloy by 100 lm in thickness on the substrate, theYSZ top coat was overlay coated by 500 lm in thickness. Theseprocesses were managed and performed by atmospheric plasmaspraying at Plasma Giken Co. Ltd., Japan. The details are presentedin Refs. [11,18].

Moreover, the free-standing top coat specimens were also pre-pared to measure their some basic properties: elastic stiffness, ten-sile strength and thermal conductivity. Here, for the former twoproperties, the specimens were extracted by removing the sub-strate material from the TBC specimen by chemical solution tech-

Table 2Test program of thermal cycle and isothermal exposure tests.

Type of test Environment T

High temperature (isothermal) exposure In air 8METHOD-I thermal cycle (uniform heating and cooling) In air (in vac.) 1METHOD-II thermal cycle (non-uniform heating and cooling) In air 1

nique. The outline of the measurement method will be given ineach subsection in next chapter.

2.2. Thermal cycle and isothermal exposure tests

Either thermal cycles or isothermal exposure was applied to theTBC specimen in air, according to the test conditions summarizedin Table 2. Thermal cycles were applied, following the two differenttype of test methods; hereinafter, denoted by METHOD-I and -II inthis work, respectively. In the ‘‘METHOD-I’’ one single furnace wasused to cyclically heat-up and cool-down the TBC specimen undersuch a cycle frequency low enough that the temperature in the TBCspecimens might be changed in steady state without significanttemperature gradient [17] (Fig. 2a). The heating and cooling ratesemployed in this work were approximately 0.133 �C/s and0.067 �C/s, respectively. The control of test temperature wasconducted via R type thermocouples which represented the tem-perature within the electric furnace. At the same time the TBCspecimen temperature was monitored by the thermocoupleswelded on the substrate, to ensure insignificant temperature dif-ference between the furnace and specimen. In the ‘‘METHOD-II’’thermal cycle test, on the other hand, two furnaces isolated witheach other; higher and lower furnaces, were used. The temperatureof the two furnaces was kept constant so that they corresponded tohigher and lower temperatures in the thermal cycle test. At thesame time, the specimen temperature was continuously monitoredby the thermocouples, as well. During the test the TBC specimentraveled periodically between the two furnaces, through a mechan-ical driving system. Here, the traveling time was so short (i.e.,within 30 s) that the temperature gradient in the specimens wassignificant during the traveling period, as will be given in Section3.2. The heating and cooling rates in the MEHOD-II were approxi-mately 1.5 �C/s and 0.67 �C/s in average, respectively.

The thermal cycle test conditions are summarized in Table 2.The isothermal exposure test was also carried out in air, by meansof an electric furnace.

After applying either thermal cycles or isothermal exposure, theresidual adhesion strength of the ceramic top coat was evaluated,according to the ASTM standard, C633 [7], as a representative mea-sure to assess the damages. The residual stress in the top coat wasalso measured by the stress relief method. Here the stress wasevaluated from the measurement of relief strain between before/after the removal of metallic substrate and bond coat.

3. Results and discussion

3.1. Isothermal exposure test

Elastic stiffness of the ceramic top coat was measured by apply-ing an external tensile load to the free-standing 0.5 � 10 � 70 mmrectangular plate specimen, where the tab plates were adhered onthe both sides of the specimen parts for clamping (Fig. 3a).Stress–strain relation was monitored via a load cell and a straingauge directly adhered at the center of specimen gauge section(see Fig. 3a), from which slope the elastic modulus was determined.Fig. 3b reveals some typical stress–strain curves. The elastic modu-lus evaluated is summarized in Fig. 4 as a function of isothermal

emperature/temperature range Time/number of cycle

00 �C, 900 �C, 1000 �C, 1100 �C 0, 100, 300, 1000 h000–400 �C, 1000–500 �C, 900–400 �C, 900–650 �C 0, 10, 100, 1000 (cycles)000–400 �C 0, 10. 100,1000 (cycles)

Fig. 2. Test equipments used for the thermal cycle tests. (a) METHOD-I. (b) METHOD-II.

Fig. 3. Tensile test of the self-standing top coat specimen. (a) Grip tab and straingauges adhered on the specimen surface. An appearance after the tensile test, (b)typical stress–strain curves measured.

Fig. 4. Change in elastic stiffness with high temperature exposure.

M. Okazaki et al. / International Journal of Fatigue 53 (2013) 33–39 35

exposure time and temperature. It is found that the modulus signif-icantly increased when the specimens were exposed at high tem-perature for long time. For example, the elastic modulus after theexposure at 1000 �C increased from an initial value of about20 GPa to about 35 GPa after 1000 h. When the exposure timewas long enough, the modulus seemed to be almost saturated.These increases might be resulted from progress of sintering ofthe top coat during isothermal exposures. As a matter of fact, a de-crease in porosity fraction with isothermal exposure has been con-

firmed in the Phase II collaboration activities in the Society ofMaterials Science, Japan [11].

Fig. 5 expresses the change in tensile strength of the top coat asa function of isothermal exposure time, where the measurementswere carried out following the same method as that for the Young’smodulus measurement. It is found from Fig. 5 that the tensilestrength also dramatically increased and then converged into a sat-urated value. It is worthy to note that the increasing behavior intensile strength is very similar to that in elastic modulus, suggest-ing that these two changes were caused by the same mecha-nism(s). One of mechanisms may be a progress of sintering.

The thermal conductivity of the top coat, ktc, was also measuredapplying the laser heat flux rig test method employed elsewhere[12,13,18]. In this work the value of ktc was approximately

Fig. 5. Change in tensile strength with high temperature exposure.

Fig. 6. Change in thermal conductivity with high temperature exposure.

Fig. 7. Change in residual stress built-up in the ceramic top coat with hightemperature exposure.

Fig. 8. Change of the remaining adhesion strength of ceramic top coat with hightemperature exposure.

36 M. Okazaki et al. / International Journal of Fatigue 53 (2013) 33–39

calculated from the steady state temperature gradient in the TBCsystem by

ktc ¼ kbmðT i � TbÞ=tbm

ðTs � T iÞ=ttcð1Þ

where kbm is thermal conductivities of the base metal, and ttc andtbm are thickness of the top coat and that of base metal includingbond coat, respectively. Ti, Tb and Ts are temperatures at thebond/top coat interface, at the bottom of substrate and at the sur-face of top coat, respectively: those were measured by thermocou-ples mounted inside or outside of the specimen (see Fig. 1c).

The changes in thermal conductivity are summarized in Fig. 6.As shown here, whereas the thermal conductivity rapidly in-creased at the beginning of exposure, the increasing rates almostconverged after a prolonged exposure, depending on the exposingtemperature. These behaviors must be also corresponding to pro-gress of sintering.

The residual stress, rr, built-up in the ceramic top coat was esti-mated from the relief strain, erelief, by

rr ¼ Etc erelief ð2Þ

where Etc and erelief are elastic modulus of top coat (measured inFig. 4), and a released strain on the substrate being removed from

the TBC specimen, that is measured via strain gauge. In Eq. (2) thestiffness of substrate is approximated to be high enough, comparedwith that of top coat. The measurement result is summarized inFig. 7. The residual stress in the as-sprayed top coat was almost zeroor slightly tensile. After being exposed at high temperature, itvaried into compression, which was corresponding with the typicalresults reported elsewhere [14].

The change of the adhesion strength of top coat with isothermalexposure time and temperatures is given in Fig. 8, where the mea-surements were on the basis of the ASTM standards [7]. It is foundfrom this figure that the adhesion strength was slightly increasedwith the increase of exposure time. This trend is roughly similarto the results by other research works [14–16]. The final rupture oc-curred almost within the top coat near the interface in all the spec-imens, as shown in Fig. 9. The formation of thermally grown oxide(TGO) was pronounced at the bond/top coat interface, as indicatedby Fig. 10. However, it was not directly conjunctive with the finalrupture area in the adhesion test. Comparing Fig. 9 with Fig. 5, itis reasonably interpreted that the increase in adhesion strengthwould be attributed to the increase of top coat strength itself.

Fig. 9. Cross sections of the specimens exposed to isothermal aging after the adhesion test.

Alumina

Mixed oxide

(a) 300 hours at 900 (b) 1000 hours at 1000

Bond coat

Top coat Top coat

Bond coat

Alumina

Fig. 10. Thermally grown oxide (TGO) formed at the bon/top coat interface, after the high temperature exposure test.

Fig. 11. Temperature history at representative local areas in the TBC specimen during the METHOD-II thermal cycle test.

M. Okazaki et al. / International Journal of Fatigue 53 (2013) 33–39 37

3.2. Thermal cycle test

It is important to know previously how the temperaturegradient is inside of the TBC specimen and whether there are anysignificant differences between the METHOD-I and -II testingmethods. Fig. 11 depicts the temperature history at the two repre-sentative areas inside of the TBC specimen subjected to the ther-mal cycles by the METHOD-II in air. Here the temperature at thetop/bond coat interface and that at the metal substrate were mea-sured by the thermo-couples mounted at inside the specimen andon the specimen surface, respectively. The measurement for theformer part was realized as follows: after the thermocouples were

welded at the interface through a small artificially drilled holeacross the top coat, the hole was filled again by a zirconia pasteto minimize possible influences of the hole; see an illustration inFig. 11. For the latter measurement location the thermocoupleswere welded directly near the bond coat/substrate interface onthe specimen surface. It is seen from Fig. 11 that the temperaturewas not uniform inside of the specimen, especially during the trav-eling period of the specimen between the two furnaces. This mustbe due to a difference in thermal conductivity between the top coatand substrate. Thus, thermal stress is generated not only from thethermal expansion coefficient mismatch between the YSZ top coatand metal substrate but also from the temperature gradient in

Fig. 13. Cross sections after adhesion test of the specimen subjected to thermal cycle

Fig. 12. Remaining adhesion strength of the ceramic top. (a) After thermal cycles byMETHOD-I (associating with almost uniformly heating and cooling), (b) afterthermal cycles by METHOD-II (associating with non-uniformly heating and cooling).

38 M. Okazaki et al. / International Journal of Fatigue 53 (2013) 33–39

specimen. In the METHOD-I, on the other hand, the temperaturegradient inside of the TBC specimen was negligible, because theheating and cooling rates were pretty slow (Table 2).

The adhesion strengths of the ceramic top coat were summa-rized in Fig. 12. It is found from Fig. 12a that the adhesionstrength by the METHOD-I increased at least up to 300 cycles.This trend is similar to the test results shown in Fig. 8. Here,the final decohesion of the ceramic top coat occurred withinthe top coat somewhat apart from the interface, as shown inFig. 13a. It is also interesting in Fig. 12a that there were no signif-icant effects of the thermal cycle test environment; compare theresults between in air and vacuum. Note again that a major factorto produce thermal stress is thermal expansion mismatch be-tween the top coat and substrate in the METHOD-I, since theTBC specimen was heated and cooled with negligible tempera-ture gradient.

In the case of the METHOD-II thermal cycle test, the remainingadhesion strength was monotonically decreased with the numberof thermal cycles (see Fig. 12b). This trend was in contrast withthat in Fig. 12a. As indicated by Fig. 11, the temperature gradientinside the specimen was significant in the METHOD-II, especiallyduring the traveling period. In the other words, in the METHOD-IIthe thermal stress is generated, not only by the thermal expansioncoefficient mismatch between YSZ top coat and metal substrate,but also by the temperature gradient in specimen.

It is worthy to note from the comparison between Fig. 12a and bthat not only the adhesion strength itself but also their changingbehavior were not always similar. In addition, the change of resid-ual stress with thermal cycles was also different between the M-ETHOD-I and -II, as shown in Fig. 14. Thus, it is natural topostulate that the temperature gradient induced thermal stressmight change the internal stress state, resulting in a change of

s. (a) After thermal cycles by METHOD-I, (b) after thermal cycles by METHOD-II.

Fig. 14. Change of residual stress with thermal cycles depending on the testingmethods.

M. Okazaki et al. / International Journal of Fatigue 53 (2013) 33–39 39

thermal fatigue damage. Here a possible additional phenomenonmust be an acceleration of microcracking inside the top coat.

Between the METHOD-I and -II there was little difference in thefinal rupture site in the adhesion test; that was inside of top coatapart enough from the prior bond/top coat interface where the for-mation of TGO layer was significant (Fig. 13). This means that theTGO was not responsible directly for the difference in remainingadhesion strength between the METHOD-I and -II. Thus, it is rea-sonable to consider that a mechanical role to produce thermalstress played more essentially in the differences seen in Figs. 12–14. It should be noted that the lower temperature in the presentthermal cycle test was 400 �C, which is relatively close to the duc-tile–brittle transition temperature (DBTT) of MCrAlY bond coat al-loys; from 300 to 500 �C in many cases [19,20]. This kind ofmechanical properties of bond coat can also commit to the differ-ences shown in Figs. 12–14, since thermal cycle damage seems tobe evolved during cooling at temperatures below the DBTT [19].When this is the case, the difference between METHOD-I and -IImay get more pronounced with the decrease of minimum temper-ature in thermal cycle test. No matter which factors are moreresponsible for, it is important to keep in mind that the thermal cy-cle failure life should be dependent on the testing method. Now anadvanced project has begun to more quantitatively explore thebackgrounds of dependence of testing method, via numerical stressanalysis.

4. Conclusion

This paper dealt with the adhesion strength of ceramic top coatin thermal barrier coatings (TBCs) subjected to thermal cycles un-der several different test conditions and testing systems, compar-ing with that exposed to isothermal aging at high temperature.Here the experimental variables were the maximum/minimumtemperatures in thermal cycles, and the testing systems used forthe tests. A series of tests clearly demonstrated that the adhesionstrength was significantly changed by the application of thermalcycles and the isothermal exposure. Of particular important results

was found in the remaining adhesion strength as well as the resid-ual stress, those were strongly dependent on the testing method togive the thermal cycles. It was suggested that the mechanical fac-tor(s) to induce thermal stress might play more major role in theabove differences and little contribution from environmental influ-ence in this work.

Acknowledgements

Some of the present tests have been carried out as a part of col-laborative research organized by Sub-committee on ‘‘Superalloysand Coatings’’, The Society of Material Science, Japan. The authorsare expressing their gratitude to the Sub-committee member fortheir fruitful discussions. One of authors, M. Okazaki, is expressinghis gratitude to a financial support to a part of this project by theGrain-in-Aid for Scientific Research by JSPS (No. 21246022).

References

[1] Miller RA. Oxidation based model for thermal barrier coating life. J Am CeramSoc 1984;67:517–21.

[2] National Research Council, Coatings for high temperature structuralmaterials—trends and opportunities. Washington DC: National AcademyPress; 1996.

[3] Seth BB. Superalloys. In: Pollock T et al. The utility gas turbine perspective,Superalloys 2000. Warrendale (PA): TMS; 2000. p. 3–12.

[4] Schneider K, Gruling HW. Influence of coatings and hot corrosion on thefatigue behavior of Ni-based superalloys. Thin Solid Films 1981;84:29–36.

[5] Sudrashan TS, editor. Surface modification technologies III. Warrendale(PA): TMS; 1990.

[6] Okazaki M. The potential for the improvement of high performance thermalbarrier coatings. Mate Res Soc Int 2003;55:3–10.

[7] ASTM standard test method for adhesion or cohesion strength of thermal spraycoatings, C633; 2008.

[8] Hutchinson JW. Metal–ceramic interfaces. In: Ruhle M, Evans AG, Ashby MF,Hirth JP, editors. London: Pergamon Press; 1990. p. 295–306.

[9] Rabiei A, Evans AG. Failure mechanisms associated with thermally grownoxide in plasma sprayed thermal barrier coatings. Acta Mater2000;48:3936–50.

[10] Okazaki M, Take K, Kakehi K, Yamazaki Y, Sakane M, Arai M, Sakurai S, KanekoH, Harada Y, Itoh A, Okuda T, Nonaka I, Fujiyama K, Nanba K. Collaborativeresearch on thermo-mechanical and isothermal low-cycle fatigue strength ofNi-base superalloys and protective coatings at elevated temperatures in thesociety of materials science, Japan (JSMS). ASTM STP 1428; 2003. p. 180–94.

[11] Subcommittee on superalloys and coatings, Phase II Activities. Japan: TheSociety of Materials Science; 2005 [chapters 2–5].

[12] Zhu D, Miller RA. Thermal conductivity and elastic modulus evolution ofthermal barrier coatings under high heat flux conditions. NASA/TM-1999-209069; 1999.

[13] Kaneko H, Torigoe T. Reliability improvement technology of thermal barriercoating for industrial gas turbines. In: Proceeding for the 31st annualconference of the GTSJ; 2003. p. 289–92.

[14] Subcommittee on superalloys and coatings, Phase III Activities. Japan: TheSociety of Materials Science; 2008 [chapters 2–5].

[15] Yamazaki Y, Schmidt A, Scholz A. The determination of the delaminationresistance in thermal barrier coating system by four-point bending tests. SurfCoat Technol 2006;201:744–54.

[16] Arai M, Sakuma T, Iwata U, Saitou M. Proposal of delamination life predictionmethod of thermal barrier coating layer under thermal cycle condition. J SocMater Sci Jpn 2001;50(6):651–6.

[17] JIS H 8451. Testing methods for thermal cycle and thermal shock resistance ofthermal barrier coatings; 2007.

[18] Okazaki M, Yamazaki Y, Namba K, Ogawa K, Ohki M, Fujiyama K, et al.Collaborative research on adhesion strength of thermal barrier coatings inJSMS. J Solid Mech Mater Eng 2010;4:252–62.

[19] Trunova O, Beck T, Herzog R, Steinbrech RW, Singheiser L. Damagemechanisms and lifetime behavior of plasma sprayed thermal barriercoating systems for gas turbines—Part I: Experiments. Surf Coat Technol2008;202:5027–32.

[20] Okazaki M. High temperature strength of Ni-base superalloy coatings. SciTechnol Adv Mater 2001;2:357–66.