Residual stress measurement of an EB-PVD Y2O3-ZrO2 thermal barrier coating by micro-Raman spectroscopy

Residual stress measurement of an EB-PVD Y2O3-ZrO2 thermal barrier coating by micro-Raman spectroscopy

Surface & Coatings Technology 204 (2009) 657–660 Contents lists available at ScienceDirect Surface & Coatings Technology j o u r n a l h o m e p a g...

295KB Sizes 19 Downloads 108 Views

Surface & Coatings Technology 204 (2009) 657–660

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 ev i e r. c o m / l o c a t e / s u r f c o a t

Residual stress measurement of an EB-PVD Y2O3-ZrO2 thermal barrier coating by micro-Raman spectroscopy M. Tanaka, R. Kitazawa, T. Tomimatsu, Y.F. Liu, Y. Kagawa ⁎ Research Center for Advanced Science and Technology, The University of Tokyo, Tokyo 153-8904, Japan

a r t i c l e

i n f o

Article history: Received 15 April 2009 Accepted in revised form 31 August 2009 Available online 9 September 2009 Keywords: Thermal barrier coating Stress measurement Y2O3-ZrO2 Raman spectroscopy Piezo-spectroscopic coefficient

a b s t r a c t Residual stress distribution in an EB-PVD 4 mol% Y2O3-ZrO2 thermal barrier coating (TBC) layer coated on a superalloy substrate has been measured by micro-Raman spectroscopy. Piezo-spectroscopic coefficient was independently calibrated on a freestanding TBC layer. The coefficient for uniaxial stress is Πuni = 5.43 cm− 1GPa− 1. The stress measurement through the TBC thickness shows compressive stress distribution from small to an almost large constant value. Such a distribution agrees with theoretical consideration since the small stress correctly reflects the free edge effect and the large constant stress is closely related to TBC bulk stresses. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Electron-beam physical vapor-deposited thermal barrier coatings (EB-PVD TBCs) have been widely used for gas turbine to protect superalloy components from high temperature environment [1–3]. The total performance of coated superalloy components depends on various factors, such as properties of constituents and their change of properties during service. The residual stress in the TBC is one important factor to affect overall performance of the coating system and has been measured by different experimental techniques. The X-ray diffraction technique [4–6] has been widely applied as a non-destructive method. The method is effective for measuring averaged stress in the TBC layer because the resolution of stress is ~ 20 MPa and spatial resolution is larger than ~ 100 μm and ~ 20 μm for in-plane and through-the-thickness directions, respectively. The curvature method [6–8] can detect the curvature change of the TBC layer coated on a substrate. This technique is destructive and allows only macroscopic in-plane stress measurement; the spatial resolution for the through-the-thickness direction being 10 ~ 50% of the total specimen thickness. The Raman spectroscopy is an effective method for measuring stresses in ZrO2 and has been applied to local stress measurement in plasma sprayed coatings [9–11] as well as in EB-PVD coating layer [12]. However, the piezo-spectroscopic coefficient was not measured for EB-PVD YSZ, because various literature papers list piezo-spectroscopic coefficients of bulk and plasma-sprayed YSZ. The present paper focuses on Raman spectroscopic techniques for measurement of residual stresses in an EB-PVD Y2O3-ZrO2 thermal ⁎ Corresponding author. Tel.: +81 3 5452 5086; fax: +81 3 5452 5087. E-mail address: [email protected] (Y. Kagawa). 0257-8972/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2009.08.042

barrier coating. Direct calibration of the piezo-spectroscopic coefficient was implemented on a freestanding coating and the piezospectroscopic coefficient was then applied to a full TBC system. 2. Experimental Procedure 2.1. EB-PVD TBC system 4 mol% Y2O3-ZrO2 TBC layers with two thicknesses of ~ 200 and ~500 μm were coated on a bond coat layer 150 μm thick using EB-PVD process. The bond coat layer was coated by the low-pressure plasmaspray process. The CoNiCrAlY bond coat layer had a chemical composition (wt.%) of: 32.0-Ni, 21.0-Cr, 8.0-Al, 0.5-Y, and the remainder Co. The substrate was ≈3 mm thick Inconel 738LC and was comprised of a Ni-base superalloy (16.0-Cr, 8.5-Co, 3.4-Al, 3.4-Ti, 1.7-Mo, 2.6-W, 1.75-Ta, 0.9-Nb, 0.11-C, balance-Ni). Hereafter, the entire coated material is referred to as the “TBC system”. Fig. 1 shows a typical example of the polished transverse section of the as-deposited EB-PVD TBC system and the definition of x, y and z coordinate system: three different layers, i.e., the TBC, TGO and bond coat layers, are clearly distinguished in the section. The TBC layer consists of columnar grains with a rough feather-like surface appearance (Fig. 2). This surface appearance is typical for the EB-PVD coating, as reported elsewhere [13]. The coated material was supplied by Japan Fine Ceramics Center (JFCC, Nagoya, Japan). The as-deposited 500 μm-thick TBC system was cut into adequate size, then embedded in epoxy resin for polishing process. Thereafter, cross sectional surfaces of the TBC specimen were carefully polished up to 1 μm diamond paste finish. Special care was taken to achieve a parallel contact between loading and support surfaces. The embedded


M. Tanaka et al. / Surface & Coatings Technology 204 (2009) 657–660

Fig. 1. A typical polished section of thermal barrier coating system used in the present study (secondary electron SEM micrograph).

TBC system was dipped into dichloromethane to dissolve epoxy resin. To obtain a freestanding TBC layer, the TBC system was dipped into 50 wt% ferric chloride solution at a temperature of 300 K and the substrate superalloy was slowly etched. Delaminated TBC layer was carefully cleaned with distilled water and dried by air blow. The final freestanding TBC layer is rectangular 1.0 mm long, 1.0 mm wide and 500 μm thick. Raman peak shift measurement was done on a 200 μm-thick topcoat sample with the 4 mol% Y2O3-ZrO2 TBC layer on the bond coat and superalloy substrate. Size of the measured specimen was 3.5 mm by 4.5 mm and ~3.2 mm thick, adequate to assume bi-axial in-plane stress condition of a TBC layer on a superalloy substrate. Transverse polished section was used for measurement. 2.2. Raman Spectroscopy The freestanding TBC specimen was fixed to a specially designed homemade compressive loading device. The device was placed on a confocal type Raman measuring equipment (NRS-1000: Special Version, JASCO Co., Tokyo, Japan). Special care was taken to align

the loading axis. The loading set-up and details of the spectroscopic system for the measurement of a Raman spectrum were reported elsewhere [9]. Raman spectroscopic measurements were carried out under a stepwise loading mode. Prior to measurement, the specimen was loaded at a rate of 30 μm/min to a set point load and then Raman spectrum was measured, a typical time required for measurement of a spectrum was 30 s. Load drop due to the effect of holding time on strain relaxation in the specimen could be neglected because of a short time measurement. Raman spectroscopic measurements were done at a fully controlled temperature (297 ± 1 K) in ambient air. A green laser (wavelength: λ = 532 nm) was used as an excitation source of Raman spectroscopy and the laser beam was incident parallel to the z-axis direction in Fig. 1 from the top surface of the free-standing TBC specimen through an objective lens (×20). Diameter of the incident green laser beam at the specimen surface was estimated to be ≈4 μm, and nominal resolution of the spectrometer was less than 0.2 cm− 1 throughout the measured Raman shift range (from 580 to 690 cm− 1). The Raman peak in this frequency range corresponds to a peak from a tetragonal (t) ZrO2 [9]. The beam spot size at a surface was less than the diameter of one column, allowing measurement in individual columns. The Raman frequency shift was obtained at the same location on TBC surface independent of the applied stress. This procedure is effective to reduce scattering associated with the location of the measured area. Care was taken to adjust the incident laser beam at a center of randomly selected individual columns. The Raman peak around 640 cm− 1 was used in peak wavenumber determination for the stress derivation because this peak has been used for stress measurement of Y2O3 stabilized ZrO2 [6]. In addition, high S/N ratio helps rapid measurement to avoid heating due to laser irradiation. The obtained Raman spectra were analyzed using GRAMS software (Thermo Electron Corp., Philadelphia, PA), which uses a Gaussian function via computer curve fitting. The Raman spectroscopy for cross sectional surface of full TBC system is conducted by using the objective lens (×100), which condition gave the laser beam size on specimen surface of 1 μm. The other spectroscopic conditions for the full TBC system are same as that for freestanding coating. The specimen was placed directly on a stage of the Raman spectroscopic equipment. Care was taken to place the outermost surface of the TBC layer on the focusing point of the incident laser beam. This focusing was achieved by maximizing Raman spectrum. The measurement was done with a green laser from a free surface. 3. Theoretical considerations regarding TBC residual stresses Due to free surface effect near the edges, σx ≈ 0; σy ≈ 0; for x ≈ wtbc = 2; y→0 and z→htbc ; σz ≈ 0:


where σx, σy and σz are the stress components of x, y and z directions, respectively (Fig. 3). The average residual stress in EB-PVD TBC layer is given by R



σtbc ≈ σtbc + σtbc + σtbc

Fig. 2. Surface morphology of columnar structure (secondary electron SEM micrograph).


R B is the residual stress associated with processing, σtbc is the where σtbc T stress caused by overall bending of the sample, and σtbc is the thermal stress generated during cooling from deposition temperature to room R , originates during processing of temperature. The residual stress, σtbc the layer and columnar microstructure (cf. Fig. 2). In the present case, the geometry ratio htbc/hs = 1/15, bending effect is negligible so B T ≅ 0 [14]. Thermal stresses, σtbc , in the bulk of TBC layer, far that σtbc from the edge surface, is approximately equi-biaxial. Defining σxin and

M. Tanaka et al. / Surface & Coatings Technology 204 (2009) 657–660


Fig. 3. Definition of coordinates and stress components in TBC layer.

σyin as in-plane thermal stress components in the x and y direction, respectively, it follows [15,16] that in



σx = σy = σtbc = −

Etbc ΔT½Etgo htgo ðαtbc −αtgo Þ + Es hs ðαtbc −αs Þ ; Etbc htbc + Etgo htgo + Es hs ð3Þ

where ΔT(=1248 K) represents the temperature change from processing to ambient. E, α and h are Young's modulus, thermal expansion coefficient and thickness, respectively. Subscripts “tbc”, “tgo” and “s” refer to the TBC layer, TGO layer and substrate, respectively. Calculation was done using materials properties listed in Table 1. Young's modulus of overall TBC layer is assumed to be 44 GPa [16]. σxin, σyin and σzin ≈ 0 are related to the residual stress in the vicinity of the free edge in the following way [10]: edge



≈ ð1−νtbc Þσx :


where σxedge is the stress in the free edge, and νtbc (=0.2) is Poisson's ratio of TBC layer. For x ≈ wtbc/2, y → 0 and z → 0, the local residual stresses are singular σyedge ~ r-λ, where r is the distance from TBC/TGO interface, and λ singularity order [17,18]. However, the singular stresses are usually very localized and difficult to be detected by the current experimental technique, and therefore are out of current scope of this work.

Fig. 4. Example of Raman peak from unstressed and stressed freestanding TBC.

around 640 cm- 1. If the TBC layer obeys fully elastic deformation, Raman shift associated with stress is written as ν ≈ ν0 + Π⋅Δσ


where ν0 is the wavenumber defined on the free-standing sample when σa = 0, Π the piezo-spectroscopic coefficient and Δσ the applied stress. Dependence of the peak position at 640 cm- 1 upon uniaxial applied in-plane stress obtained from a freestanding TBC layer is shown in Fig. 5. The line drawn in the figure was obtained by least square fitting of the plots. Here, the stress, σa, is defined as σa ≡

Pa htbc wtbc


where htbc is the thickness, wtbc the width of TBC layer and Pa is the applied compressive load (Pa < 0). Here, the peak wavenumber when Pa = 0, ν0 is set to 643.78 cm- 1. Least squares fitting of the plots (Fig. 5) gives Πuni = 5.43 cm− 1GPa− 1. It was reported that the

4. Results 4.1. Stress-Raman Shift Fig. 4 is a typical example of Raman shift of a freestanding TBC layer without applied stress (σa = 0) and uniaxially compressed with applied stress σa = –110 MPa. It should be noted that the spectra are processed using GRAMS; this process does not affect peak location of the spectra. The peak wavenumber from the freestanding TBC layer is ν0 = 643.78 cm-1 and that stressed one is 644.36 cm- 1. This result clearly demonstrates that the Raman peak position shifts with applied stress, i.e., stress is measurable using Raman shift at a wavenumber

Table 1 Material properties for calculation of thermal stress. Material

Layer thickness (μm)

Young's modulus (GPa)

Coefficient of thermal Expansion (10- 6K- 1)

TBC TGO Substrate

200 0.5 3000

44 380 200

11 8 15

Fig. 5. Plots of Raman peak versus applied compressive stress.


M. Tanaka et al. / Surface & Coatings Technology 204 (2009) 657–660

stresses in the TBC layer, in contrary to negligible bending stresses in the current experiment for much thinner specimens. (4) As shown in Fig. 2, the TBC layer consists of rough feather-like columnar grains more than 4 μm wide, larger than the incident laser spot area. The large scattering in the measured stresses is believed to be reasonable because the experimental data points capture the local microstructural variation. This suggests that the local stress inside the columnar grain can be obtained by the present method, thus providing a new perspective to investigate TBC failure mechanisms through mapping Raman stresses in the TBC layer in order to investigate the presence of an inhomogeneous stress distribution inside the columnar crystals and/or in different columns. 5. Summary Fig. 6. Residual stress distribution in through-the-thickness direction of the TBC layer.

piezo-spectroscopic coefficient of dense 7 mol% Y2O3-ZrO2 processed by hot pressing is Πuni = 0.96 cm− 1GPa− 1 [19]. The coefficients Πuni ~ 1.0 cm− 1GPa− 1 of undoped and Y2O3 doped ZrO2 also were reported elsewhere [20–22]. However, it should be pointed out that the coefficient has different meaning from that reported in the literatures because the freestanding TBC layer has porous and columnar microstructure. In addition, in previous studies on PVD and plasma-sprayed TBC, appreciably large coefficients have been observed. The coefficient for PVD TBC was found to be 4.55 cm− 1GPa− 1 [6]. The coefficient for plasma-sprayed TBC was reported to be much larger 25 cm− 1GPa− 1 probably due to a higher porosity [9]. The coefficient obtained from EBPVD TBC in this study is close to that from PVD TBC.

Raman piezo-spectroscopic coefficient of a freestanding 4 mol% Y2O3-ZrO2 EB-PVD TBC was obtained by applying uniaxial compressive stresses. The coefficient Πuni = 5.43 cm− 1GPa− 1 was then used to measure residual stresses in the TBC layer of a full TBC system. The stress in the as-deposited TBC layer was found to be within − 270 ~ −120 MPa. Near the TBC top surface the stress approaches to zero due to the free edge effect. The difference between theoretical thermal stresses and measured residual stresses at a distance larger than about 40 μm from the TBC top surface may be attributed to existence of some intrinsic stress generated during processing and an initial stress in the calibrated sample that was not stress-free. This study showed the effectiveness of micro-Raman spectroscopy for measurement of local residual stress in an EB-PVD TBC layer using the calibrated piezospectroscopic coefficient.

4.2. Residual Stress in TBC Layer and Discussion


Fig. 6 shows a typical stress distribution in the TBC layer along the through-the-thickness direction by using the piezo-spectroscopic coefficient obtained previously. It is clear that the stress is compressive with large scattering. The solid line indicates the average. Assuming that the TBC layer is homogeneous and isotropic, thermal stresses in the TBC layer under the plane stress condition are calculated using Eq. (3). σxedge is then estimated by Eq. (4) and shown by the broken line in Fig. 6. These results will be discussed below.

The authors are grateful to an anonymous reviewer for many constructive suggestions.

(1) Close to the free surface near the edges, the stress follows Eq. (1) and approaches to zero value asymptotically. is ~135 MPa, which is about 60 % smaller than (2) Estimated σedge x that far from the TBC top surface. When the distance from the TBC top surface is larger than 40 μm, the stress level falls within a range of − 270 ~ −180 MPa with an almost constant average. The stress difference between the solid and broken lines at a distance larger than 40 μm from the TBC top surface may be attributed to the following two factors: (i) existence of the intrinsic stress σRtbc that has not been taken into account in Eqs. (3) and (4); and (ii) existence of some initial stress in the calibrated sample for σa = 0, that is, the free-standing sample was not stress-free and thus a systematic error in the calibration existed to shift all data points parallelly. (3) The stress dependence on the distance from the TBC top surface is different from that reported in Ref. [12]. The sample used in Ref. [12] was much thicker leading to substantial bending

References [1] M.J. Stiger, N.M. Yanar, M.G. Toppings, F.S. Pettit, G.H. Meier, Z. Metall. 90 (1999) 1069. [2] A.G. Evans, D.R. Mumm, J.W. Hutchinson, G.H. Meier, F.S. Pettit, Prog. Mater. Sci. 46 (2001) 505. [3] D.R. Clarke, C.G. Levi, Annu. Rev. Mater. Sci. 33 (2003) 383. [4] A.H. Bartlett, R.D. Marschio, J. Am. Ceram. Soc. 78 (1995) 1018. [5] M. Levit, I. Grimberg, B.Z. Weiss, Mater, Sci. Eng. A206 (1996) 30. [6] V. Teixeira, M. Andritschky, W. Fischer, H.P. Buchkremer, D. Stöver, J. Mater. Proc. Tech. 92–93 (1999) 209. [7] M.K. Hobbs, H. Reiter, Surf. Coat. Technol. 34 (1988) 34. [8] J. Matejicek, S. Sampath, Acta Mater. 51 (2003) 863. [9] M. Tanaka, M. Hasegawa, A.F. Dericioglu, Y. Kagawa, Mater. Sci. Eng. A419 (2006) 262. [10] S. Krämer, S. Faulhaber, M. Chambers, D.R. Clarke, C.G. Levi, J.W. Hutchinson, A.G. Evans, Mater. Sci. Eng. A490 (2008) 26. [11] A. Scrivani, G. Rizzi, U. Bardi, C. Giolli, M.M. Miranda, S. Ciattini, A. Fossati, F. Borgioli, J. Therm. Spray Technol. 16 (2007) 816. [12] T. Tomimatsu, Y. Kagawa, S.J. Zhu, Metall. Trans. A 34A (2003) 1739. [13] V. Lughi, V.K. Tolpygo, D.R. Clarke, Mater. Sci. Eng. A368 (2004) 212. [14] C.H. Hsueh, J. Appl. Phys. 91 (2002) 9652. [15] S.S. Kim, Y.F. Liu, Y. Kagawa, Acta Mater. 55 (2007) 3771. [16] Y.F. Liu, Y. Kagawa, A.G. Evans, Acta Mater. 56 (2008) 43. [17] N.Y. Cao, Y. Kagawa, Y.F. Liu, Surf. Coat. Technol. 202 (2008) 3413. [18] D.B. Bogy, J. Appl. Mech. 38 (1971) 377. [19] A.M. Limarga, D.R. Clarke, J. Am. Ceram. Soc. 90 (2007) 1272. [20] P. Bouvier, G. Lucazeau, J. Phys. Chem. Solids 61 (2000) 569. [21] B. Alzyab, C.H. Perry, R.P. Ingel, J. Am. Ceram. Soc. 70 (1987) 760. [22] J. Cai, Y.S. Raptis, E. Anastassakis, Appl. Phys. Lett. 62 (1993) 2781.