Cyclic behavior of EB-PVD thermal barrier coating systems with modified bond coats

Cyclic behavior of EB-PVD thermal barrier coating systems with modified bond coats

Surface & Coatings Technology 203 (2008) 449–455 Contents lists available at ScienceDirect Surface & Coatings Technology j o u r n a l h o m e p a g...

1MB Sizes 0 Downloads 40 Views

Surface & Coatings Technology 203 (2008) 449–455

Contents lists available at ScienceDirect

Surface & Coatings Technology j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / s u r f c o a t

Cyclic behavior of EB-PVD thermal barrier coating systems with modified bond coats Uwe Schulz ⁎, Klaus Fritscher, Andrea Ebach-Stahl DLR-German Aerospace Center, Institute of Materials Research, 51170 Cologne, Germany

a r t i c l e

i n f o

Available online 4 September 2008 Keywords: Thermal barrier coating Thermally grown oxide EB-PVD Bond coat

a b s t r a c t The lifetime of thermal barrier coating (TBC) systems depends on substrate, bond coat, thermally grown oxide (TGO), and ceramic top coat. In the present paper NiPtAl bond coats as well as NiCoCrAlY(X) deposited by LPPS and EB-PVD (electron-beam physical vapour deposition) underneath conventional EB-PVD yttria stabilized zirconia top coats were investigated on three different substrate alloys. Several bond coat treatments such as polishing, annealing in vacuum, and grit blasting were employed in order to study effects on TBC life, and particularly the underlying mechanisms of TGO formation. Samples were thermally cycled at 1100 °C and partly at 1150 °C. Spallation of the TBCs is mainly correlated with TGO formation that is influenced by bond coat type and pre-treatment. The longest lifetimes were achieved on a novel Hf-doped EB-PVD NiCoCrAlY-X bond coat owing to a differing TGO formation and failure mechanism. Activation energies derived from lifetimes and test temperatures were calculated to identify key failure mechanisms within these complex coating systems. © 2008 Elsevier B.V. All rights reserved.

1. Introduction Thermal barrier coatings (TBCs) offer the potential to significantly improve efficiency of aero engines and stationary gas turbines for power generation. All their constituents that are bond coat, thermally grown oxide, ceramic top and even the substrate they protect are crucial for lifetime of the coatings [1,2]. State-of-the-art TBCs consist of a PtAl diffusion or an MCrAlY overlay bond coat and a ceramic top coat deposited by electron-beam physical vapor deposition (EB-PVD) or plasma spraying. The EB-PVD process offers the advantage of a superior strain and thermal shock tolerance for the ceramic coatings due to their columnar microstructure. Although increasing gas temperatures led to the development of alternative chemical compositions, the current material in use is still partially yttria stabilized zirconia (7YSZ). Manufacture of TBC systems is a multi-step process where every detail may influence TBC life. While TBC thickness and cycle length in testing seem to be of second order importance for life of MCrAlY based TBCs [3], EB-PVD process parameters do influence TBC life. Therefore, in the present study EB-PVD process parameters were kept in a narrow band to exclude an influence of processing on TBC life. Proper bond coat treatments prior toTBC deposition offer a large potential to economically improve TBC life [4–7]. During processing and in service, a thermally grown oxide (TGO) layer forms as a result of bond coat (BC) oxidation which usually plays the most important role for the lifetime of the TBC. Failure in EB-PVD TBCs is almost always initiated at or near the TGO, mostly along the TGO/BC interface. Although literature on individual coating types is numerous, direct comparison between them is scarce

⁎ Corresponding author. Tel.: +49 2203601 2543; fax: +49 2203 696480. E-mail address: [email protected] (U. Schulz). 0257-8972/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2008.08.056

[8–11]. It is obvious that testing temperature has a great influence on cyclic TBC life, with a lifetime reduction of about a factor of 10 for every 100 °C temperature increase under cyclic testing conditions [11]. The present paper compares the cyclic life of standard 7YSZ top coats on a variety of differently processed and treated bond coats. For selected versions the role of testing temperature was studied. Special attention was paid to the influence of rare earth elements such as Y and Hf in the bond coats on cyclic TBC life. 2. Experimental Cylindrical samples of three different Ni-base superalloys were used in the present study. Major differences involve the Ti and Mo content in IN100 while CMSX-4 and Rene 142 are additionally alloyed with W, Ta, Re and contrasting contents of Hf (see Table 1). All three alloys are used in turbine blade applications with the highest temperature capability of the CMSX-4, followed by Rene 142 and IN100. Samples of 6 to 10 mm diameter received the following bond coats and treatments (a–d) prior to TBC deposition that are additionally summarized in Table 2 (for details of sample preparation and treatments see references for each treatment): (a) 100 to 120 µm thick EB-PVD Ni-22Co–20Cr–12Al–0.1 to 0.2Y (wt.%) bond coats, designated EB-MCrAlY, were densified by ball peening and subsequently vacuum annealed at 1080 °C for 4 h at a pressure lower than 5 ⁎ 10− 5 mbar [2]. A selected version got an exceptional smooth substrate finishing prior to bond coat deposition (4000 grit finish). Only for this version, a slightly changed testing procedure was employed that reduced the arrest time in ”humid” laboratory air for visual inspections at room temperature [12].

450

U. Schulz et al. / Surface & Coatings Technology 203 (2008) 449–455

Table 1 Chemical composition of substrate materials, measured by X-ray fluorescence (wt.%, Nickel is balanced) Alloy

Co

Cr

Al

Mo

IN 100 DS CMSX-4 Rene 142

14 9 12

9 6.5 6.8

5 5.6 6.1

2.3 0.6 1.5

W 6 5

Ta 6.5 6.4

Ti 5 1

Hf 0.1 1.5

Re 3 2.8

C

B

Other

0.18

0.015

0.05 Zr, 1 V

0.12

0.015

0.02 Zr

Concentration of minor additions of C, B, and others are taken from nominal compositions of the manufacturer of the alloys.

(b) A 75 to 95 µm thick novel EB-PVD bond coat was applied by two-source EB-PVD in order to obtain a composition of Ni– 23Co–23Cr–10.5Al–0.3 to 0.5Si–0.1Y–0.5 to 1.0Hf (designated EB-MCrAlY-X). The variation in hafnium content was caused by variable distances along the cylindrical samples from the Hfsource. Samples were subjected to the same vacuum heat treatment as the NiCoCrAlY bond coat (a). (c) Approx. 50 to 80 µm thick commercial PtAl diffusion bond coats were supplied by two different vendors (1) and (2). The standard treatment consisted of lightly grit blasting with 220 mesh alumina prior to TBC deposition. For an alternative version of PtAl(1) the annealing atmosphere at 1080 °C for 4 h was replaced by a mixture of argon and hydrogen at atmospheric pressure (designated PtAl(1)ArH) [3,10,13]. (d) NiCoCrAlY of the nominal composition Ni–20Co–12Cr–13Al– 0.7Y with traces of Hf/Si (wt.%) was deposited by Vacuum Plasma Spraying (VPS-MCrAlY) at research center Juelich FZJ. The standard treatment consisted of the common diffusion heat treatment at 1080 °C for 4 h followed by 870 °C for 20 h, both in vacuum. After smoothing the bond coat by an industrial source, grit blasting was applied prior to EB-PVD TBC deposition. An alternative version had a reversed order of processing steps: smoothing by hand polishing followed by vacuum annealing prior to EB-PVD TBC deposition [6,14]. One version on Rene 142 substrates followed the sequence diffusion heat treatment + smoothing + vacuum annealing at 1080 °C for 4 h. Versions (a) to (d) were given finally a 150 to 200 µm thick 7YSZ ceramic top coat by reactive EB-PVD in rotation mode at 12 rpm. In order to avoid an influence of EB-PVD processing parameters on cyclic TBC life, preheating of the samples, substrate temperature, chamber pressure and amount of oxygen were kept constant. (e) An air plasma-sprayed top coat (APS) was applied on the identical VPS bond coat with the common preparation steps of vacuum annealing as described above for d) followed by APS deposition [6]. While IN100 and CMSX-4 were used as base materials for all the above coatings, Rene 142 was used only for selected coating systems (versions (a), (c1), and (d) with a slightly changed composition). All samples were cyclically tested at a holding temperature of 1100 °C using 1 h-cycles (50 min heating, 10 min forced air cooling to room temperature). Unless otherwise mentioned, a minimum of three samples per condition was tested. Selected samples were additionally

Fig. 1. Cyclic lifetimes of 7YSZ TBCs at 1100 °C on various state-of-the-art bond coats using common industrial practice for bond coat treatment prior to TBC deposition.

tested at 1150 °C under the same cyclic conditions. Failure of the TBCs was defined as TBC spallation of an area with one dimension greater than 5 mm. Representative samples were cross-sectioned after deposition and after failure. They were prepared by conventional metallographic preparation techniques for examination by scanning electron microscopy and energy-dispersive X-ray spectroscopy. 3. Results and discussion 3.1. State-of-the-art bond coats In Fig. 1 cyclic TBC life is displayed for standard bond coat treatments. For all versions TBC life was longer on CMSX-4 than on IN100 with the exception of the EB-PVD bond coat that had a reversed order. A ranking of cyclic TBC life on IN100 can be given as follows: EBbond coat N PtAl(2) N PtAl(1), VPS-MCrAlY, APS-system. The ranking on CMSX-4 is similar with the only change for the EB-PVD bond coat: PtAl (2) N PtAl(1) N EB-bond coat, VPS-MCrAlY, APS-system. Note that the fully plasma-sprayed TBCs behaved nearly similar to the short life EBPVD top coats despite of the inherent lower strain compliance of this system. Cyclic TBC life of optimized versions that arose from several development projects are depicted in Fig. 2. Again, on the optimized EB-NiCoCrAlY bond coat TBC life is longer on IN100 than on CMSX-4 while for the PtAl system (1) it is the opposite. TBC life on the vacuum plasma-sprayed bond coat was nearly the same on both substrates. Note the dramatic increase in TBC life of the optimized versions by a factor of approx. 2 (EB-bond coat on both substrates and VPS-MCrAlY on CMSX-4) up to a factor of 3 and higher (PtAl(1), VPS-MCrAlY on IN100) in comparison to the standard treatments given in Fig. 1. The chemically modified EB-MCrAlY-X system is discussed separately in Section 3.2. Two samples of four selected substrate–coating systems were tested at 1150 °C, using the same equipment and time regime as for

Table 2 Summary of bond coat deposition method, applied BC treatment, and 7YSZ top coat deposition method Coating designation

Bond coat deposition method

BC treatment

Top coat

PtAl(1) and PtAl(2) PtAl(1)ArH EB-MCrAlY EB-MCrAlY-X VPS-MCrAlY VPS-MCrAlY improved APS

Electroplating Pt + Al diffusion Electroplating Pt + Al diffusion EB-PVD EB-PVD two-source VPS VPS VPS

Grit blasting 4 h at 1080 °C in ArH at ambient pressure Peening + 4 h at 1080 °C vacuum Peening + 4 h at 1080 °C vacuum Diffusion annealing, smoothing, grit blasting Smoothing, 4 h at 1080 °C vacuum Diffusion annealing

EB-PVD EB-PVD EB-PVD EB-PVD EB-PVD EB-PVD APS

U. Schulz et al. / Surface & Coatings Technology 203 (2008) 449–455

Fig. 2. Cyclic lifetimes of 7YSZ TBCs at 1100 °C on improved and novel bond coats. The improvements consisted of: annealing in ArH for 4 h at 1080 °C prior to TBC deposition (PtAl), 4000 grit polished samples and reduction of water vapour during room temperature inspection intervals (EB-MCrAlY), smoothing and reversed processing sequence (VPS-MCrAlY) and novel chemistry (EB-MCrAlY-X).

the 1100 °C tests. EB-PVD TBCs failed after the following averaged cycle numbers. Numbers in parenthesis give the 1100 °C results for comparison; variations in sample numbers are marked. Activation energies of lifetimes are given on the right.

451

of rumpling with an increasing number of cycles. Previous investigations have revealed that an initially flatter interface between the bond coat and TBC delays TGO rumpling and, hence, increases TBC life [10]. Martensite was observed in the PtAl bond coat occasionally after prolonged testing. The platinum concentration dropped below 5 at.% in the coating region due to fast diffusion of this element along the concentration gradient towards the substrate for all versions, however this was accompanied by substantial thickening of the bond coat through from interdiffusion of elements from the substrate alloy. In all cases the TGO consisted of alpha alumina with rare indications for the formation of an outer porous mixed alumina–zirconia zone. TGO thickness at failure was in the range between 5 and 7 µm with no large dependence on testing time. This can be explained by the rapid initial oxidation, where a large portion of the TGO thickness is formed during the first 100 h of testing while subsequent TGO thickening of PtAl bond coats is extremely slow [3]. The fully plasma-sprayed system had the typical rough TBC bond coat interface and showed indication of partial white failure, i.e. cracks above the interface TGO to TBC within the TBC. The two phase PtAl bond coats (1) and (2) differ in both, microstructure and cyclic life. Although nominally identical, version (1) was thicker and had a higher aluminum content while version (2) had a higher Pt content close to the BC to TBC interface. The higher amount of platinum in version (2) after bond coat deposition seems to overcompensate the lower total coating thickness and the smaller reservoir of aluminum.

a) on IN100 substrates: – PtAl(2) standard grit blasted 180 cycles (611); 397 kJ mol− 1 – EB-MCrAlY improved 472 cycles—only one sample—(1422); 358 kJ mol− 1 b) on CMSX-4 substrate: – modified EB-MCrAlY-X 1272 cycles—three samples were tested— (4693); 424 kJ mol− 1 – PtAl(2) standard grit blasted 504 cycles (960); 209 kJ mol− 1 Ranking and relative life of the coating systems are nearly unchanged if testing temperature is raised by 50 °C. The absolute life time is roughly one third at the higher temperature with the exception of PtAl on CMSX-4. Microstructure investigations revealed that TGO formation, failure location and crack pattern are also similar at both testing temperatures. Therefore, only selected cross sections are shown here. TGO formation after cyclic tests at 1150 °C is shown in Fig. 3 for the EB-NiCoCrAlY system. Significant visual differences between the standard and improved versions EB-NiCoCrAlY did not exist. As detailed in [2,15], the TGO consists of two layers. The outer mixed zone of less than 1 µm thickness is slightly porous and consists of alumina– zirconia; quite often in an off-plane arrangement of alternating areas of (i) zirconia-free alumina underneath yttrium-oxides that had formed during vacuum annealing, and of (ii) areas containing the common mixed zone arrangement of zirconia particles embedded in alpha alumina. Because diffusion processes are accelerated at the higher testing temperature, this off-plane arrangement was not clearly visible after 1150 °C testing. The thicker inner TGO zone is dense alpha alumina. The bright particles in this region (see Fig. 3) are Y-rich oxides, commonly identified as yttrium-aluminates with a majority of the YAG phase. They grew slightly larger in size at the higher testing temperature at the expense of their number. Spalling of the TBC occurred either between the bond coat and TGO or occasionally between the dense alumina TGO and mixed zone. EB-PVD manufactured NiCoCrAlY bond coats preserve a relatively flat TGO over the entire testing duration at both temperatures. Due to space limitations, the PtAl and VPS-MCrAlY version are only described verbally here. The TGO on all versions NiPtAl showed signs

Fig. 3. Cross section of TGO of version IN100 + NiCoCrAlY (improved)+ 7YSZ TBC after 472 cycles at 1150 °C. SEM micrograph overview (a) and detail of outer TGO part (b). Bright particles in this layer are Y-rich oxides.

452

U. Schulz et al. / Surface & Coatings Technology 203 (2008) 449–455

A significant result of the current work is that the bond coats perform differently on each superalloy. On the Ta free polycrystalline alloy IN 100 (see Table 1), the superiority of a clean bond coat containing metallic yttrium becomes clearly visible. Consequently, a long TBC life was achieved on IN100 with the EB-PVD bond coat. Previous work [12] has shown that among substrate element additions Ta had a negative influence on TBC life. The other substrate elements that differ between the used polycrystalline IN100 and the single crystal CMSX-4 alloy such as C, B, Re and W did not change EBPVD TBC life on model cast alloys. It remains still unclear why a reversed performance order of the coatings on the two substrates between EB-PVD and all other bond coats exists. It can be only speculated that diffusion in the EB-PVD bond coat is the fastest for all investigated bond coats due to the small grain size that favors grain boundary diffusion and clean grain boundaries because of the lowest oxygen content in the EB-PVD coating (especially if compared with the VPS version). Tantalum may hence diffuse faster from the substrate towards the bond coat/TGO interface, where it seems to degrade TGO adherence. For PtAl it is agreed that diffusion of substrate elements towards the surface is slowed down which would cause less detrimental effects of Ta from the SX alloy in the TGO or at the TGO/ bond coat interface. Hence other effects arising from the SX alloy may dominate, leading to a longer TBC life of PtAl on CMSX-4 than on IN100 substrates. While for the EB-bond coat it was not clear which factor was most important for lifetime enhancement of the improved version (a smoother substrate or reduced water vapor at room temperature during testing), the bond coat treatments on PtAl and on VPSMCrAlY did clearly increase TBC life. The underlying mechanisms are summarized shortly. For the VPS bond coat, a reversed sequence of processing steps assured formation of larger amounts of yttria on bond coat grain boundaries during vacuum annealing, giving rise for formation of a strong off-plane mixed zone [6,16]. This optimized TGO microstructure provides a longer TBC lifetime, most probably due to a higher effectiveness of the active rare earth element yttrium. The influence of annealing PtAl in ArH atmosphere is discussed in detail for the CMSX-4 substrate elsewhere [13] and operates similarly for the IN100 substrate samples. The oxygen partial pressure during this annealing is high enough for partial oxidation that leads to an oxide film consisting mainly of alpha alumina prior to TBC deposition. The EB-PVD top coat obviously adheres better on this TGO compared to the one that forms on grit blasted surfaces during EB-PVD processing, leading to prolonged TBC life. In [17] an oxidation pretreatment has also prolonged TBC life on PtAl bond coats by a factor of two. Vacuum annealing of PtAl did not provide an adequate atmosphere in order to achieve full coverage of the bond coat surface by alumina [3,10,13]. 3.2. Novel EB-bond coat The best lifetime improvements were achieved on the CMSX-4 substrate by doping standard EB-PVD MCrAlY compositions with Hf (EB-MCrAlY-X in Fig. 2). Although the measured aluminum content in the as manufactured stage was slightly lower for the NiCoCrAlY-SiHf version compared to the standard NiCoCrAlY (10.5 vs. 12 wt.%), small additions of the reactive element Hf did increase TBC life by a factor of 15. A comparative test of this coating on a single IN100 sample, however, brought about only a marginal life time improvement which is the topic of ongoing research. In the as-coated state shown in Fig. 4 hafnium has already diffused towards the bond coat surface and is present on top or underneath the TGO. Bright particles are Hf-rich; if in contact to the TGO they are oxidized and away from the interface they are still metallic. In analogy to NiCoCrAlY it can be concluded that during vacuum annealing Hf is partially oxidized. Due to the low concentration of Hf in the bond coat

Fig. 4. Cross section of CMSX-4 with novel EB-PVD MCrAlY-X bond coat after TBC deposition. a) overview, b) higher magnification of TGO. Bright particles are Hf-rich; if in contact to TGO it is oxidized and far away still metallic.

(≤1 wt.% ), surface coverage with HfO2 is incomplete. The preferred Hf diffusion path is along grain boundaries between the two phases of the bond coat that are based on beta-NiAl and gamma-Ni solid solution, as can bee seen from the position of bright particles in Fig. 4. In analogy to the mechanism in yttrium dominated bond coats such as vacuum annealed EB-PVD [16] and VPS NiCoCrAlY [6], an off-plane mixed zone has formed on EB-PVD NiCoCrAlY-SiHf. Underneath the upper bright hafnia particles pure alumina forms (Fig. 4). Adjacent to these particles, a typical mixed zone consisting of zirconia particles within alumina is present. Its formation is assumed to follow the same reaction scheme as in NiCoCrAlY where solution of zirconia in transient alumina or in spinel type oxides occurs with subsequent precipitation of zirconia when the TGO transforms into the stable alpha alumina polymorph [16,18,19]. TGO formation during cyclic testing and after TBC failure is for the novel bond coat quite different from the Hf-free counterparts (Fig. 5). The major differences comprise (i) a broccoli-structured wavy interface between TGO and bond coat accompanied by a large and varying TGO thickness of 12 to 30 µm, (ii) bright and mostly rounded hafnia particles spread throughout the whole TGO thickness, (iii) occurrence of short-distance cracks in the TGO visible parallel to the interface, and (iv) NiCoCrAl-spinel oxides spread in the TGO (medium gray areas). The TGO consists of an approximately 1 µm thick outer porous mixed zone and a thick inner dense TGO underneath. The undulated TGO bond coat interface is caused by oxide pegs of alumina that surround Hafnium oxide particles. Cracks appear already

U. Schulz et al. / Surface & Coatings Technology 203 (2008) 449–455

453

Fig. 5. CMSX-4 substrate with modified EB-PVD MCrAlY-X bond coat after a)–c) intermediate testing interval of 1584 cycles at 1100 °C and d) after TBC failure at 1490 cycles at 1150 °C. Bright particles in TGO are Hf-oxides, medium grey particles are NiCoCrAl-spinel oxides.

after intermediate testing times in cross sections. Although very carefully prepared, it can be never conclusively declared that all cracks were already present prior to sample preparation. Cracks occur mainly at the TGO bond coat interface where they bridge the TGO in areas of oxide pegs that have a larger thickness, but also at various distances from the TGO–TBC interface (Fig. 5b and c). The cracks cannot follow the wavy interface TGO/bond coat. It remained unclear what the mechanism for crack arresting is, but all cracks are only a few 100 µm long in maximum. Cracks cross hafnia particles or go around them. From cross sections one would deduce that crack density increases within the TGO at the higher testing temperature, thereby creating a blocky TGO appearance. Although still speculative, this specific short-distance cracked TGO microstructure is believed to be beneficial for the dramatically prolonged TBC life. One possibility is that stresses within the TGO are lowered and therefore the crack driving force stays below the critical one for TBC spallation over prolonged times. Although quite often believed to be detrimental in a TGO, spinel type oxides embedded in alumina were not harmful in the present investigation as they do not influence cracks or diffusion paths of e.g. oxygen. Neither silicon nor yttrium was identified in the TGO region in any enriched phases, although they were both present in minor concentration in the bond coat. Silicon was still present within the bond coat after testing while Y has most probably off-diffused into the TGO, but did not form larger oxide particles as common for NiCoCrAlY bond coats.

3.3. TBCs on Rene 142 substrates On the Hf-containing superalloy Rene 142 selected bond coats were applied and the following lifetimes at 1100° were recorded: (i) 2327 cycles for VPS-MCrAlY of a slightly changed composition Ni–38Co–19Cr–10Al–0.4Y (measured values of Amdry 995, supplied by an industrial vendor) with grit blasting treatment and 1138 cycles with vacuum heat treatment, both treatments were applied after the smoothing step; (ii) 2000 cycles for standard EB-NiCoCrAlY (test has been stopped after this time without TBC failure); (iii) 988 cycles for PtAl(1) using standard grit blasting prior to TBC deposition. TBC lifetimes are dramatically longer for both MCrAlY versions tested and about twice that for PtAl(1) on this substrate compared to IN100 and CMSX-4 substrates and identical bond coat treatments. For the VPS system results prove that a more conventional sequence of processing steps (diffusion annealing, smoothening, followed by vacuum annealing) is less favorable compared to grit blasting since smoothing removes most of the yttria formed during vacuum diffusion annealing [6]. It remains for further research to confirm that a reversed sequence of processing steps as shown for CMSX-4 and IN100 (Fig. 2) would give a longer life of VPS bond coats on Rene 142 as well.

454

U. Schulz et al. / Surface & Coatings Technology 203 (2008) 449–455

Fig. 6. SEM cross section of Rene 142 + VPS-MCrAlY + EB-PVD TBC (vacuum annealed, after 1463 cycles at 1100 °C). White arrows indicate hafnium oxides.

Pertinent characteristics of all MCrAlY coatings on Rene 142 are i) presence of hafnia particles in the TGO and ii) a very wavy interface between TGO and bond coat, accompanied by large variations in TGO thickness and apparently entrapped metallic bond coat particles that are in reality a consequence of the three-dimensional undulations of the TGO and a 2D-cut through three-dimensional oxide fingers. An example of a VPS-MCrAlY is shown in Fig. 6 that also discloses significant large spinel oxides within the TGO. This TGO microstructure is quite similar to the TGO of the EB-MCrAlY-X version. Obviously the source of hafnium doesn't matter, and diffusion of this element towards the surface is fast enough in both cases. Since the diffusion distance is larger for the Hf-containing substrate Rene 142 compared to an Hf-containing bond coat, it develops the hafnium-rich oxides slightly later and not during vacuum annealing. Accordingly, these oxides are found mainly in the lower TGO region, accompanied by larger undulations TGO/bond coat, while in the Hf-containing bond coat they are more frequently observed in the upper TGO region. Consequently, an off-plane mixed zone is missing here. Diffusion of Hf in MCrAlY is much faster than for Y, and hence the tendency to surface enrichment is higher for Hf compared to Y as shown in [20] by experimental evidencing a sputtered Hf interlayer. As detailed in previous papers [2,3,21], the fact that TGOs on samples with the longest lifetimes showed a very rough BC/TGO interfaces does not match models that have been developed for TBC spallation suggesting that imperfections in that area might be initiation sites for large scale buckling and subsequent spallation of the TBC. In the case of MCrAlY on Rene 142 and EB-MCrAlY-X on CMSX-4, hafnia containing oxide pegs and the large undulations may act as crack stoppers since an interface crack has to change its propagation direction repeatedly if it follows intimately the BC/TGO interface that possesses the lowest interface toughness. The TGO is tightly bonded to the metallic bond coat by these pegs, with a positive effect of the rare earth element hafnium. In [22] plasma-sprayed NiCoCrAlY-SiHf + EB-PVD top coat systems developed similar oxide pegs rich in Hf an Y suggesting that presence of mainly Hf always leads to oxide pegs and a typical undulated TGO bond coat interface. Same observations have been described for Hf-doped PtAl coatings [23]. Surprisingly, a similar undulated interface containing oxide pegs was observed for the VPSMCrAlY systems on the Hf-free alloys despite of only traces of Hf in the bond coat. The TGO of version PtAl on Rene 142 shows only limited signs of rumpling on this alloy but it contains also large Hf-oxide pegs at the TGO bond coat interface [10]. 3.4. Activation energy of cyclic lifetime data Activation energies can provide indirect evidence to decode TBC failure mechanisms. If compared with activation energies of basic

processes such as diffusion, oxide growth, phase transformation, crack propagation etc., they can indicate potential mechanisms responsible for failure [20,25]. Furthermore, if the activation energy changes over a larger temperature interval of e.g. 1000–1100–1200 °C, this is an indication of a change in failure mechanism. Although the current data are by far not robust due to the small sample numbers, which immediately gets evident if the previous novel bond coat data of 483 kJ mol− 1 based on two failed samples [20] and the latest 424 kJ mol− 1 based on meanwhile three tested samples are compared, the following conclusions can be drawn. The lowest activation energy achieved with PtAl on CMSX-4 suggests that this system degrades slowly if the temperature is raised. The calculated activation energy is consistent with 225 kJ mol− 1 from data given for PtAl on Rene N5 between 1100 and 1150 °C in [24], whereas other investigations for PtAl systems revealed higher activation energies of 356 to 520 kJ mol− 1 especially for the temperature interval 1100 to 1200 °C [9,20]. The apparent failure mechanism of rumpling of the CMSX-4 + PtAl system investigated here seem to have a low activation energy. Intermediate values of activation energies in the range of 350 to 390 kJ mol− 1 compare well to inward-growing oxidation of alumina-forming alloys. The current values for EB-PVD NiCoCrAlY and PtAl on IN100 fall into that regime and agree well with literature data that have been put together in [20]. The highest activation energy of 424 kJ mol− 1 was found for the chemically modified EB-MCrAlY-X bond coat system. The arrangement of cracks may play a decisive role for failure, as derived from the micrographs in Fig. 5 and detailed in [20]. The high activation energy indicates a large temperature dependence of the failure mechanism and hence a large drop in cyclic life if the temperature is raised. But the dramatic gain in lifetimes still outrules the deficiencies envisaged at high temperatures. Interestingly, a similarly high activation energy of 418 kJ mol− 1 can also be determined for an MCrAlY + EB-PVD TBC system given in [26] in the temperature range of 1150 to 1190 °C. Finally, the potential of a coating system can be realized more reliably in view of associating absolute cyclic life and activation energy data. 4. Conclusions Cyclic oxidation experiments of 7YSZ EB-PVD top coats on various substrate–bond coat combinations revealed the following conclusions. – The longest lifetimes were achieved on a Hf-containing EB-PVD bond coat on CMSX-4 followed by coatings on the Hf-containing substrate superalloy Rene 142. Whenever hafnium is involved, hafnia as well as oxide pegs are present in the TGO leading to large undulations at the interface TGO bond coat that prolong TBC life. Local spinels in the TGO are not harmful for TBC life. – The substrate alloy plays a decisive role for TBC life. There is no universal bond coat for all substrates, instead an optimized bond coat treatment tailored for each system offers potential for life time improvements. Treatments such as annealing of the bond coat or surface smoothing can prolong TBC life. – Analysis of activation energies offers a valuable information to ascertain possible failure mechanisms. Acknowledgements The authors gratefully acknowledge careful manufacture and testing of the coatings by J. Brien, C. Kröder, H. Mangers, and D. Peters. O. Bernardi, W. Braue, H. Lau, U. Kaden, and B. Baufeld contributed some of the TGO and lifetime investigations. The provision of the VPS bond coats and several discussions with R. Vaßen from FZ Juelich is gratefully acknowledged. Special thanks to M. Peters for reviewing the manuscript.

U. Schulz et al. / Surface & Coatings Technology 203 (2008) 449–455

References [1] U. Kaden, C. Leyens, M. Peters, W.A. Kaysser, in: J.M. Hampikian, N.B. Dahotre (Eds.), Elevated Temperature Coatings: Science and Technology III, 1999, p. 27. [2] U. Schulz, M. Menzebach, C. Leyens, Y.Q. Yang, Surf. Coat. Technol. 146–147 (2001) 117. [3] U. Schulz, H. Lau, H.-J.R. -Scheibe, W.A. Kaysser, Zeitschrift f. Metallkd. 94 (2003) 649. [4] U. Schulz, K. Fritscher, in K. Kokini (ed.) ASME, New York, 1993, p. 163–172. [5] T.J. Nijdam, W.G. Sloof, Surf. Coat. Technol. 201 (2006) 3894. [6] U. Schulz, O. Bernardi, A. Ebach-Stahl, R. Vassen, Surface and Coatings Technology (in press), doi:10.1016/j.surfcoat.2008.08.071. [7] M. Matsumoto, Surf. Coat. Technol. 202 (2008) 2743. [8] H.M. Tawancy, N. Sridhar, N.M. Abbas, D. Rickerby, J. Mater. Sci. 35 (2000) 3615. [9] N.M. Yanar, G. Kim, S. Hamano, F.S. Pettit, G.H. Meier, Mater high temp. 20 (2003) 495. [10] H. Lau, C. Leyens, U. Schulz, C. Friedrich, Surf. Coat. Technol. 165 (2003) 217. [11] G.M. Kim, N.M. Yanar, E.N. Hewitt, F.S. Pettit, G.H. Meier, Scri. Mater 46 (2002) 489. [12] U. Kaden, Phd thesis, Werkstoffwissenschaftliche Schriftenreihe, Vol. 56. 2003, RWTH Aachen. ISBN 3-86130-094-X. [13] B. Baufeld, U. Schulz, Surf. Coat. Technol. 201 (2006) 2667. [14] O. Bernardie, diploma thesis, 2005, TU Darmstadt. p. 62.

455

[15] W. Braue, K. Fritscher, U. Schulz, C. Leyens, R. Wirth, Mat. Sci. Forum 461–464 (2004) 899. [16] W. Braue, U. Schulz, K. Fritscher, C. Leyens, R. Wirth, High Temp. Mater. 22 (2005) 393. [17] V.K. Tolpygo, D.R. Clarke, Surf. Coat. Technol. 200 (2005) 1276. [18] M.J. Stiger, Metall. Mater. Trans. A 38 (2007) 848. [19] C.G. Levi, E. Sommer, S.G. Terry, A. Catanoiu, M. Rühle, J. Am. Ceram. Soc. 86 (2003) 676. [20] K. Fritscher, U. Schulz, C. Leyens, Mater.wiss. Werkst.tech. 38 (2007) 734. [21] U. Schulz, K. Fritscher, and W.A.Kaysser, COST 2002, Liege, J. Lecomte-Beckers, M. Carton, F. Schubert, P.J. Ennis, 2002, p. 483–492. [22] C. Mercer, S. Faulhaber, N. Yao, K. McIlwrath, O. Fabrichnaya, Surf. Coat. Technol. 201 (2006) 1495. [23] J. Nesbitt, B. Nagaraj, J. Williams, in: N.B. Dahotre, J.M. Hampikian, J.E. Morral (Eds.), Elevated Temperature Coatings: Science and Technology IV, TMS, 2001, p. 77. [24] M.J. Stiger, N.M. Yanar, M.G. Topping, F.S. Pettit, G.H. Meier, Z. Met.k.d. 90 (1999) 1069. [25] R.M. German, Sintering Theory and Practice, Wiley, 1996. [26] D.S. Rickerby, Advanced Coatings for High Temperatures, Nice, Forum of Technology, 2002.