Application of small punch creep testing to a thermally sprayed CoNiCrAlY bond coat

High velocity oxy-fuel thermal spraying was used to prepare free-standing CoNiCrAlY (Co–31.7% Ni–20.8% Cr–8.1% Al–0.5% Y (wt%)) bond coat alloy samples approximately 0.5 mm thick. Creep tests were conducted at 750 °C on these samples using a small punch (SP) creep test method. The samples were characterised before and after creep testing using scanning electron microscopy with electron backscatter diffraction (EBSD). EBSD revealed a two phase fcc γ-Ni and bcc B2 β-NiAl microstructure with grain sizes ~1–2 μm for both phases, which did not change significantly following testing. The constant temperature SP test data were characterised by a minimum creep strain rate, View the MathML source, and a total time to failure, tf, at different applied stresses. The data are fitted to conventional power law equations with a stress exponent for creep close to 8 in the Norton power law and between 7 and 10 in the Monkman–Grant creep rupture law. Creep rupture was predominantly due to creep cavitation voids nucleating at both the γ–β interphase boundaries and the γ–γ grain boundaries leading to final failure by void linkage. However, rupture life was influenced by the quantity of oxide entrained in the coating during the spray deposition process.


Introduction
The continuing requirement to improve the efficiency of aeroengine and land based gas turbines means that there is a need to continue to increase turbine gas inlet temperatures. Thus thermal barrier coating (TBC) systems, which are used to protect turbine blades from these harsh operating environments, continue to be intensively studied so that their durability can be improved [1][2][3]. Generally a TBC system is deposited on a superalloy substrate and consists of an aluminium containing bond coat (BC) between the superalloy and the ceramic top coat, which acts as the actual thermal barrier. Additionally, a thermally grown oxide (TGO), predominantly alumina, forms at the interface between the bond coat and the topcoat during service at elevated temperature. The chemical, microstructural and mechanical characteristics of the bond coat are crucial to the durability of the overall system. This is because oxygen permeates through the ceramic top coat and oxidises the bond coat, slowly thickening the TGO during service [4][5][6][7]. This causes a progressive build-up of stress in the system and eventual spallation of the ceramic coating. It is the way the strain from TGO growth is accommodated by the creep, yield and ductility characteristics of the bond coat that will often determine the coating lifetime [8,9]. Bond coats are typically either of the diffusion aluminide type or the overlay MCrAlY-type (M=Co, Ni). Overlay coatings, deposited by low pressure plasma spraying (LPPS), vacuum plasma spraying (VPS) or, more recently by high velocity oxy-fuel (HVOF) thermal spraying, have become more widely used because of advantages such as lower cost, better control of composition and the ability to employ complex alloys very different in composition from the superalloy substrate. MCrAlY alloys are typically complex multi-phase materials and can comprise, for example, fcc -Ni and bcc B2 β-NiAl phases or , β and ordered  (Ni 3 (Al,Ti)) phases and can have a composition which is tailored to achieve specific performance needs in particular applications [10].
Furthermore, the thermally sprayed bond coat microstructure depends critically on the following factors: the alloy composition; the type of thermal spray process used; the process parameters employed and the overall coating thickness deposited. All of these factors can affect the mechanical properties of the coating. It is therefore important to distinguish between the mechanical behaviour of bulk MCrAlY alloys and that of  [18] is based on the deformation of a miniature disc-shaped test specimen typically 8-10 mm in diameter and around 0.5 mm thick. The test is carried out by applying a loaded ball to one surface of the disc which is firmly clamped around its rim. The response to applying a constant load over a period of time is then recorded to obtain creep displacements. The SP test for creep behaviour involves complex deformation processes [19]. Nevertheless, a recent review [20] has indicated that SP creep testing can lead to consistent results providing jig and specimen geometry are carefully controlled. Hitherto, the SP test has been most widely used for assessing creep behaviour of bulk steels used in nuclear and conventional power plant [21] where its ability to characterise in-service materials using only a small volume of material has been of significant benefit. More recently, it has been shown capable of analysing different zones of weldments in both creep resistant steel [22] and a nickel-based alloy [23]. However, SP testing of MCrAlY bond coats does not appear to have been previously reported.
Therefore the aims of the work reported in this paper are to determine if the SP test is a suitable method for examining creep behaviour of thermally sprayed bond coats, to investigate the microstructural changes and failure characteristics in the test and to quantify the creep behaviour in terms of creep rupture times and well known power law relationships. SP tests were carried out on samples of free-standing HVOF thermally sprayed coatings, approximately 0.5 mm thick, produced from a commercially available CoNiCrAlY powder. The sample thickness used here was thus significantly closer to that of alloy bond coats used in industrial TBC systems than has been the case for previously reported studies on MCrAlY alloys.

Materials, HVOF thermal spraying and heat treatment
The coatings used in the creep experiments were prepared by HVOF thermal spraying using powder with the following nominal composition Co-31.7%Ni-20.8%Cr-8.1%Al-0.5%Y (wt%). The powder was obtained from Praxair (CO-210-24) and had a size range of -45+20 m with a chemically analysed oxygen content of 0.037 wt% O. The coatings were deposited onto mild steel substrates with dimensions 60 × 25 × 1.8 mm using a Met Jet III liquid fuel HVOF gun. The details of the spraying process and procedures are given elsewhere [24]. Two sets of coatings were produced for this study in two separate spray runs and hereafter are referred to as C1 and C2. The primary spray parameters employed were nominally identical for the two runs. Coatings were sprayed to a thickness of approximately 0.5 mm and were then debonded from the mild steel by bending around a mandrel.
Detached coating samples were vacuum heat treated at 1100 °C for 2 h followed by natural cooling, to replicate the initial heat treatment given to bond coat alloys during TBC manufacture. Previous work has shown that such a heat treatment is sufficient to reduce to a minimal level any porosity that might be present in these coatings after HVOF spray deposition [24]. Specimens 8 mm diameter in the form of discs for SP testing were cut from the heat treated coatings by electro-discharge machining. They were ground down from the as-deposited thickness and polished to a final thickness of approximately 430 m using 1 m diamond paste. The final thickness was controlled to within ± 8 m as measured by a digital micrometer.

Small punch creep test
The In the present work, SP creep tests were carried out by employing loads ranging from 67.5 to 100 N at 750 °C in air. For each creep test, the displacement of the centre of the disc, , was recorded as a function of time.
Load can be converted to stress using equation (2), derived from membrane stretching theory [25], where a p is the distance between the punch axis and clamping edge, a p =2 mm in all experiments reported in this paper. R s is the radius of the punch, which is 1.25 mm and t 0 is the thickness of the disc, which is 0.430 mm. K s is a nondimensional correction factor to take account of localized necking during large strain deformation. It is believed to be material dependent and for steels is typically around 1.2. In the present work load was converted to stress using equation (2) and taking K s to be 1.
Equation (3) [26] relates the strain, , to the displacement of the punch, . The displacement rate can be converted to strain rate using equation (4), obtained by differentiating (3).

Material characterisation
The microstructures of coating cross-sections before and after testing were examined.
Small samples were cut, mounted in conducting resin, ground and diamond polished to a 1 m finish. In the case of samples taken following testing, these were cut from different regions so that deformed and undeformed sections could be studied i.e.
reference samples were taken from the clamped region where the alloy had been at elevated temperature but had not undergone creep. The oxygen contents of the sprayed coatings were determined by inert gas fusion combustion analysis of debonded material at an external laboratory.

Microstructure of the coatings prior to testing
The microstructures of the vacuum heat treated coatings C1 and C2 prior to testing are shown in Fig. 2. It is evident from Figs. 2(a) and (c) that there is very little porosity apparent in both of the samples. As previously reported [24], this CoNiCrAlY coating has a two phase microstructure, consisting of the dark contrast bcc-β (Co,Ni)Al phase and the light contrast fcc-γ Ni-rich matrix phase. The volume fraction of the uniformly distributed β phase was found by image analysis to be around 33% in both C1 and C2, with an average β-phase size of 1-2 m.  In the present case, data for C1 and C2 (Fig. 5(a)) are found to lie on a common Norton power law best fit line and the parameters B and n are given in Table 1.
In the case of the rupture power law, there is evidently a displacement of the data sets for C1 and C2 (Fig. 5(b)) and each coating has been separately fitted to the rupture power law (equation 6) with the parameters, M and  given in Tables 2; for an equivalent stress level, coating C1 fails at a shorter time. In the clamped region, i.e. to the right of region A, the microstructure was found to be essentially unchanged in terms of the -phase size and morphology and the : phase proportions. This indicates that exposure to 750 C for the duration of the test did not, in itself, induce significant microstructural changes. The grey contrast features in between regions A and B were found to be cavities with extensive internal oxide brought about by oxidation during testing which was undertaken in air, Fig.   7(b).

Higher magnification micrographs of the region A and B are shown in Figs. 7(c) and
(d) respectively. There is some evidence that the fine oxides formed during thermal spraying align as a result of the creep deformation as shown by the arrows. In Fig.   7(d) it is evident that local voids are initiated in region B and appear to have linked together to form rather larger cavities seen in Fig. 7(b). 8

Relationship to uniaxial creep testing
Small punch creep tests at various loads on both C1 and C2 confirm that at 750 C creep deformation occurred and produced displacement curves typical of the SP test [19]. As has been previously stated [19], these curves are only superficially similar in form to those typically seen in uniaxial tests. The tertiary creep region of a uniaxial test typically begins at a creep strain of 1-5% for most materials. In contrast, the equivalent feature in a SP test, when the minimum deformation rate occurs (at a deformation of 1 mm or more), is seen to be at strains of approximately 30% or more as indicated by equation (3). Hence, the minimum displacement rate in SP tests should not be considered to be directly related to the secondary creep region of a uniaxial creep test. The secondary region in the data from a SP test is a result of a balance between the "stiffening effects" of the reducing cone angle which develops during the test and the increasing contact area between specimen and indenter and "accelerating effects" of the material being in the tertiary creep region as well as the thinning of material in the contact region, as deformation increases. Therefore, the minimum displacement rate, , is related to the tertiary creep region of the material, but is also affected by the large strain, changing cone angle and changing contact conditions which occur in the secondary region of the SP test curve as discussed in detail in reference [19]. As can be seen in Fig. 7 (5) and (6)). The log-log plots in Fig. 5 show that the results obtained from SP testing of these HVOF CoNiCrAlY coatings also fall on straight lines, indicating that their constant temperature creep behaviour at 750 C can be described by the Norton and Monkman-Grant laws.
Hence, even if there are limitations in relating SP data to uniaxial creep test data the SP method is a valuable tool for assessing and, quantitatively comparing, creep behaviour of thin, thermally sprayed bond coat alloys.

Creep fracture of the bond coat samples in the SP test
The log-log plot of minimum strain rate versus stress (Fig. 5(a)) shows that results from both coatings fall on the same straight line, indicating very similar material behaviour during creep deformation. This is not, in itself, surprising given the essentially identical microstructures of the two coatings. However, some difference in behaviour of the two coatings is seen in the plot of stress versus time to failure ( Fig.   5(b)) where coating C1 fails at shorter times for a given stress. Coating C1 was found to have a marginally higher overall oxygen content by chemical analysis, 0.40 compared with 0.36 wt% O and, as can be seen from Fig. 2(a), there is a visible band of material in the middle of C1 which was particularly rich in oxide stringers which would account for the higher oxygen level in chemical analysis. The formation of such a band is possibly due to a temporary perturbation in the HVOF process. It seems possible therefore that this microstructural difference between the two coatings is somehow responsible for the earlier creep failure of C1 as discussed below.
Consider first the lack of any shape change of the monocrystalline β phase particles and the retention of the fine equiaxed grain structure in the -matrix. These features indicate that deformation in the SP test of the CoNiCrAlY HVOF coating is dominated by creep of the γ-matrix (Figs. 7 and 8). Extensive tertiary creep as seen here is generally due to a high density of mobile dislocations which increases with strain. However, the alloy eventually fractures because of the gradual evolution of grain boundary cavitation. Indeed Figs. 7(c) and 8(b) provide evidence for creep cavitation voids nucleating at γ -β interphase boundaries as well as at  - grain boundaries. It is likely that once nucleated the voids continue to grow leading to final failure by void linkage. The extensive internal oxidation of voids seen in Fig. 7(b) supports this view. It is important to recall that at 750 C the -NiAl phase has little ductility [28] and brittle second phase particles have previously been noted as preferential sites of void nucleation during creep of, for example, duplex stainless steel [29], Nimonic 80A [30] and SiC reinforced aluminium [31].
Although it is proposed that creep rupture is primarily initiated by voids forming at β or - interfaces, the fine scale oxide particle and stringers may also have a role in initiating voids. The fine oxide particles are generated during thermal spraying and tend to form at interparticle boundaries ie where feedstock powder particles come into contact. It may thus be expected that these fine oxides decrease creep rupture time as they are also essentially hard, undeformable particles but on a finer scale than the phase. Creep failure would then be expected to be accelerated in a coating with a higher concentration of such particles (particularly if somewhat localised) as is the case in sample C1. The net effect would be a shorter creep life as is indeed observed in C1. This seems intuitively correct since a larger volume fraction of hard phase particles gives rise to a greater particle/matrix interfacial area providing more nucleation sites for creep cavitation voids.

Comparison with previous studies
The limited data available from the literature appear to show that creep parameters of MCrAlY materials depend very significantly on alloy composition, test temperature and the method used to fabricate the test samples. A ductile-to-brittle transition behaviour is consistently observed; it is composition dependent but typically occurs around 450 to 600 C. Between about 650 and 1050 C, creep appears to follow Norton and Monkman-Grant power law behaviour [11][12][13][14][15][16][17]. It is clear that in the present alloy the ductile-to-brittle transition is below 750 C, the fracture surface seen in Fig. 3 shows no evidence for brittle fracture and the fine grained fracture surface is entirely consistent with creep failure. With regard to the values of the stress exponent, n, in the Norton power law equation the available data are limited but values in the range 3-5 are suggested [13][14][15] compared to the present value 7.9 .
However, previous studies have generally either been conducted at temperatures above that used here (750 C) or involved different MCrAlY compositions. Hebsur and Miner [12] suggested that power law (dislocation) creep is the governing mechanism at lower temperature in these type of alloys and the stress exponent is usually predicted to vary from 3 to about 10 for such behaviour so the present results could be consistent with this mechanism. However, the grain size of previous samples was not reported and the very fine -phase grain size of the present materials could significantly influence the creep mechanisms in the test and the resultant values of the power law parameters. Although it is clear that further work is needed with the SP test to establish with greater certainty power law parameters for specific alloys, it is evident that the test is valuable as a tool for ranking the behaviour of different MCrAlY alloys in the form of thin, thermally sprayed bond coats.      Typical small punch test displacement curve at 750 C of coating C1 for three different loads

Figure 5
Log-log plots of (a) minimum strain rate versus stress and (b) stress versus time to failure for C1 and C2 in the small punch creep test. Experiemental data are represented by symbols and best fitting lines obtained using equations (5) and (6)        Fig. 7. The phase is coloured red in (b) and there is no evidence of preferred orientation in (a). a b Table 1 Norton power law parameters for C1 and C2 (combined data sets) B (s -1 MPa -n ) n C1 6.0 10 -20 7.9 C2 Table 2 Rupture power law parameters for C1 and C2 (individual data sets) 3.3 10 -20 10.5 C2 1.2 10 -15 7.6