Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments

Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments

ARTICLE IN PRESS SCT-21616; No of Pages 9 Surface & Coatings Technology xxx (2016) xxx–xxx Contents lists available at ScienceDirect Surface & Coati...

1MB Sizes 1 Downloads 12 Views

ARTICLE IN PRESS SCT-21616; No of Pages 9 Surface & Coatings Technology xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Surface & Coatings Technology journal homepage: www.elsevier.com/locate/surfcoat

Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments Albert Manero II a , Alex Selimov a , Quentin Fouliard a , Kevin Knipe a , Janine Wischek b , Carla Meid b , Anette M. Karlsson c , Marion Bartsch b , Seetha Raghavan a, * a

University of Central Florida, 4000 Central Florida Blvd.,Orlando, FL 32816, USA German Aerospace Center, Institute of Materials Research, Linder Höhe, Köln 51147, Germany c Cleveland State University, Washkewicz College of Engineering, 2121 Euclid Ave., Cleveland, OH 44115, USA b

A R T I C L E

I N F O

Article history: Received 8 July 2016 Received in revised form 23 September 2016 Accepted 24 September 2016 Available online xxxx Keywords: Piezospectroscopy Thermal barrier coatings Thermal gradient mechanical load Damage identification

A B S T R A C T The application of high temperature ceramic coatings has enabled aircraft and power generation turbines to run at higher inlet temperatures for greater efficiency. Their use extends the lifetime of the superalloy blades that bear thermal gradients and mechanical loads during operation. In this work, ex-situ photoluminescence spectroscopy was conducted to investigate the stresses within the thermally grown oxide of a thermal barrier coated tubular sample following complex realistic conditions, such as induced thermal gradients, and long duration aging. The resulting high spatial resolution stress contour maps highlight the development of the thermally grown oxide in response to the complex conditions. The outcomes highlight both the role of the aging process and the oxide growth’s influence on the stress profile which varies spatially across the specimen. The results further provide early detection of micro-damaged zones in the oxide layer nondestructively. Improving the understanding of the coating system’s response to loading conditions will allow for more accurate system modeling and early detection and monitoring of damage zones, which is critical for improving efficiency and longevity of aircraft and power generation turbines. © 2016 Elsevier B.V. All rights reserved.

1. Introduction As turbine inlet temperatures have increased the overall efficiency of engines, thermal barrier coatings (TBC) have been utilized to protect the load bearing superalloy substrate from the high temperatures in aircraft and power generation turbines in an effort to extend its lifetime [1–3]. It is therefore paramount to understand the mechanisms that govern the durability of these coatings, which play a critical role in increasing engine reliability and longevity [4,5]. The coating system typically comprises a ceramic top coat, adhered to the superalloy substrate via a metallic bond coat. Between the bond coat and ceramic top coat an oxide layer begins to develop, which grows rapidly due to thermal loads. For aircraft jet engines, the ceramic top coat is deposited via Electron Beam Physical Vapor Deposition (EB-PVD), whereas power generation turbines often utilize coatings applied by atmospheric plasma spraying. These techniques produce microstructures that influence the coatings’ thermal conductivity, porosity, and strain tolerance [4,6]. Coatings deposited

* Corresponding author. E-mail address: [email protected] (S. Raghavan).

via EB-PVD feature a columnar structure that has excellent strain tolerance and durability [7,8]. The oxide layer, commonly referred to as the thermally grown oxide (TGO) is comprised primarily of a phase alumina and has a central role in the failure mechanics of the system [9,10]. Considerable research has been conducted to identify how the thermally grown oxide behaves under cycling throughout its lifetime and eventual failure [11,12]. While much of the studies have been conducted under isothermal conditions, recent efforts to combine thermal gradients, mechanical loads, and in-situ measurements provide the ability to explore the resulting strain evolution under realistic loading conditions [13–15]. High resolution spatial mapping of stress within layers could improve our understanding of mechanisms leading to failure. Piezospectroscopy is a viable and effective method for investigating the ceramic top coat and oxide layer of TBC’s deposited via EB-PVD. It is a non-destructive technique which examines spectral response of the material under stress via laser excitation. The ceramic top coat’s columnar structure and translucence to the laser excitation allows for the excitation laser light to penetrate to the oxide layer below. The excitation and subsequent relaxation of the Cr3+ impurity in the material results in photon emission, which can provide insight as to the mechanical behavior and material

http://dx.doi.org/10.1016/j.surfcoat.2016.09.057 0257-8972/© 2016 Elsevier B.V. All rights reserved.

Please cite this article as: A. Manero et al., Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments, Surface & Coatings Technology (2016), http://dx.doi.org/10.1016/j.surfcoat.2016.09.057

ARTICLE IN PRESS 2

A. Manero et al. / Surface & Coatings Technology xxx (2016) xxx–xxx

properties of the material [16–20]. This technique has been utilized to probe the behavior of the interfaces of the oxide scale, where many failure mechanisms are observed [15,21,22]. In these efforts, the technique has contributed to both identifying damage to the TGO, and some efforts to assess its degree of severity [10,23]. As the coating ages, micro-cracking and micro-delamination occurs. Through their growth and merging, macro scale damage begins to propagate which can lead to the failure of the system [23–27]. Modeling the life expectancy for the coating structure remains an arduous process, as the complexity of the service conditions may lead to a variety of failure mechanisms. Our growing efforts to enable high-resolution spatial measurements with this technique has expanded its capabilities for stress distribution and damage mapping [28–30]. Monitoring the stress evolution in the TGO, with high spatial resolution, identifying initiation of damage to the coating under realistic loading conditions could provide a powerful advantage for understanding its response to complex loading conditions and its longevity. The ability to observe the initial onset of damage and to capture its progression will shed light on the long term evolution of the coating system, allowing for the refinement and validation of numerical models with the aim of advancing the efficiency and longevity of these high temperature coating systems. In this study, representative specimens were held under cyclic thermal gradient mechanical loading conditions. The tubular geometry of the specimens was designed for investigations of the strain response with internal forced cooling for in-situ synchrotron X-ray diffraction measurements described elsewhere [15,31,32]. The as-processed samples were cyclically aged with this complex loading and then investigated via piezospectroscopy to determine how the loading conditions develop the oxide layer (Dataset 1). Following a long duration aging under isothermal loading conditions of 1000 ◦ C, the piezospectroscopic measurements were re-conducted to investigate the response due to thermal aging and to identify any zones of preliminary micro-damage on the tubular geometry (Dataset 2). Further cyclic aging was then conducted, and a final set of piezospectroscopic measurements was evaluated for comparison (Dataset 3). 2. Experimental procedure 2.1. Experimental setup and test protocol The investigation was conducted on a TBC system deposited on an Inconel 100 superalloy substrate. The TBC was comprised of a 7 wt.% Yttria partially stabilized Zirconia (YZS) top coat and a NiCoCrAlY bond coat (BC), both deposited via EB-PVD. The deposition method provides a columnar structure of the YSZ top coat, resulting in strain tolerant behaviors [33,34]. The specimens were designed to incorporate the application of internal coolant, thermal loads, and mechanical loads and used for in-situ synchrotron X-ray diffraction measurements described elsewhere [15,31]. As such, the sample was designed with a tubular geometry with a substrate inner diameter 4 mm and outer diameter 8 mm. The YSZ ceramic top coat and NiCoCrAlY metallic bond coat had an as-coated thickness of 240 lm and 80 lm, respectively. The coated length was 102 mm of the full length of 160 mm. The design and manufacturing of specimens was conducted at the German Aerospace Center (DLR) in Cologne, Germany. A schematic of the loading conditions and diffraction measurements is presented in Fig. 1a. Specimens were heated to an exterior temperature of 1000 ◦ C with simultaneous internal cooling to impose a thermal gradient across the coating while a 30 lm beam passed through the coating. This is illustrated in Fig. 1b and c. A combination of the focusing of the heater lamps over the sample length, the internal cooling and heat flux to the specimen’s clamping grips, results in a thermal gradient which also evolves with the highest temperatures in the central section of the sample

where X-ray data was collected. This is likely to create a variation in TGO development within and outside the central heating zone. The thermal condition of external heating and internal cooling over the cross-section and length of the layers is represented in a schematic in Fig. 1d. The measurements have revealed in-situ strain evolution for the bond coat and YSZ layers. The current study compliments and extends the synchrotron measurements [15,32] by providing valuable information on the TGO layer, identifying any variations indicating non-uniform thermal gradient mechanical loading, and to prove the location of the X-ray diffraction measurements was uniform and viable for study. The stress state was influenced by the prior complex thermal loads for a high temperature duration of 17 h and the mechanical loading. Photoluminescence spectroscopy measurements were then taken to map the stress state spatially over the cylinder. To conduct the photoluminescence measurements, a green 532 nm diode laser with 19 mW of power was utilized to excite a piezospectroscopic response. An exposure time of 4 s was used to optimize the collection, and the calibration was conducted with a Neon–Argon source lamp. The expected uncertainty was ±0.030 nm root mean squared (RMS) with the calibration. A Princeton Instrument Acton Series 2150 spectrometer with a 1200 grating/mm grating was utilized in conjunction with a fiber optic probe to collect the piezospectroscopic data. A schematic of this can be seen in Fig. 2a. For the early developing oxide, four horizontal snake scans of 2000 points each were collected to produce a high resolution map. The resolution in the vertical and horizontal scan direction measured 400 lm, with 100 vertical scan rows and 20 horizontal scan columns. This covered an area of 40 mm by 8 mm and spanned the entire diameter of the sample. After each scan, the sample was turned 90◦ , which provided complete 360 ◦ of measurements plus overlap of two faces. It has been reported in literature that small spot sizes below 2 lm fall within the range of individual crystallites and can therefore result in variances due to the anisotropy of mechanical and thermal properties of a alumina [35] at that scale. In this study, the laser spot size was large enough to average the crystal grains effectively. The scan was focused on the midsection of the primary heated zone and upward to the top of sample outside of this zone. This was designed to measure the regions of variation induced by slight changes in thermal loading and interactions with the induced cooling flow on the inner wall of the tubular sample [36]. The scan was broken into four scanning face segments with overlap to reduce the error associated with the extremities of the map. After the analysis revealed that the 90 ◦ scan procedure’s overlapping data did not have significant variations in the stress profile’s mean or standard deviation due to laser incidence angle, the presented scans feature only two faces 180◦ apart which assist in clearly showing the spacial stress profile variations. For clarity, the comparison of the statistical difference between the inclusion of the overlapping spacial data and the presented circumferential axis images is provided in Fig. S1. Fig. 2b schematically illustrates the characteristic peaks R1 and R2 from Cr3+ doped alumina [37] and how they shift due to compressive stresses. To ensure that the influence of depth of focus was not significant, a separate experiment was designed by varying the depth of focus over a distance of 2400 lm at a spatial resolution of 10 lm. The measurements from this test are plotted in Fig. 2c which revealed small variation of approximately 13 MPa over the varying focus distance. This confirmed that the geometry of the sample in conjunction with the YSZ top coat does not significantly alter the alumina emissions with respect to focusing distance. As such, high resolution spectral contour maps could be produced for each loading history condition. Table 1 presents the full cycling history and when the three datasets were collected. Following piezospectroscopic measurements of the early TGO with thermal gradient and mechanical fatigue (TGMF) loading (Dataset 1), the specimen was placed in a furnace for long duration isothermal aging. The isothermal aging was

Please cite this article as: A. Manero et al., Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments, Surface & Coatings Technology (2016), http://dx.doi.org/10.1016/j.surfcoat.2016.09.057

ARTICLE IN PRESS A. Manero et al. / Surface & Coatings Technology xxx (2016) xxx–xxx

3

16-128 MPa Tensile Loading

a.

Coating Temperature Inner Substrate Temperature

b.

Cooling Air

1000°C Primary Heated Zone

Thermal drop

c.

16-128 MPa Tensile Loading

d.

Time Thermal loading representation for each cycle TGO Coolant Flow

X-Ray Scan Zone

<1000°C

YttriaStabilized Zirconia (YSZ)

Incident X-ray Bond Coat(BC)

1000°C

Substrate TGO

YSZ BC Substrate Fig. 1. Schematics of cyclic thermal gradient and mechanical loading conditions for X-ray diffraction experiments. a. Loading methods for diffraction measurements, b. thermal cycle loading with internal cooling imposing a thermal gradient across the coating system, c. X-ray scan location across the layers moving from tangential toward the substrate, and d. the influence of complex temperature field on the thermally grown oxide development.

conducted at 1000 ◦ C with no mechanical load and without induced internal cooling. After 300 h of isothermal aging, piezospectroscopic measurements were conducted to acquire a stress profile for the now more developed thermally grown oxide layer (Dataset 2). For the

additional isothermal aged oxide scale, again four snake scans of 2000 points each were collected to produce a high resolution map. The resolution in the vertical and horizontal scan direction measured 400 lm, with 100 vertical scan points by 20 horizontal scan points collected. This covered an area of 40 mm by 8 mm and spanned the

a.

Intensity (Arbitrary)

b.

R1 Δv

Unstressed State Stressed State R2 Δv

c.

Centerline Stress (GPa)

-1200 µm

Schematic of R-Line shift due to compression

1200 µm

4.7

0

4.65 4.6 4.55 4.5 4.45

Average = 4.55GPa Stdev = 12.98 MPa

4.4

Wavenumber(cm-1)

-1500

-1000

-500

0

500

1000

1500

Depth of Focus (µm)

Fig. 2. a. Schematic of experimental setup with spectroscopy equipment for nondestructive measurement, b. schematic of piezospectroscopic shifts for a-alumina under compressive loads, and c. measurements taken at varying depth of focus showing relatively unchanging stress calculated from the R1 peak location.

Please cite this article as: A. Manero et al., Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments, Surface & Coatings Technology (2016), http://dx.doi.org/10.1016/j.surfcoat.2016.09.057

ARTICLE IN PRESS 4

A. Manero et al. / Surface & Coatings Technology xxx (2016) xxx–xxx

Table 1 Loading history for all three datasets. Sample

Loading history

Early TGO with TGMF Aging Dataset 1

Additional Isothermal Aging Dataset 2 Additional TGMF Cycling Dataset 3





Ramp up: 50 C per min to 1000 C. Hold: 1000 ◦ C for 40 min. Ramp down: 1000 ◦ C with −50 ◦ C ramp per min to room temperature Hold: 1000 ◦ C Ramp up: 50 ◦ C per min to 1000 ◦ C. Hold: 1000 ◦ C for 20 min. Ramp down: 1000 ◦ C with −50 ◦ C ramp per min to room temperature

entire diameter of the sample. Each scan was turned 90◦ which provided 360◦ of data [36]. This is highlighted in Fig. 3. After the analysis revealed that the 90◦ scan procedure’s overlapping data did not have significant variations in the stress profile’s mean or standard deviation due to laser incidence angle, the presented scans feature only two faces 180◦ apart which assist in clearly showing the spacial stress profile variations. For clarity, the comparison of the statistical difference between the inclusion of the overlapping spacial data and the presented circumferential axis images is provided in Fig. S1 as well as the raw stress profile data with the piezospectroscopic conversion equations in Fig. S2. Further thermal exposure was added during subsequent TGMF testing at Argonne National Laboratory. To probe the condition after additional TGMF cycling, further piezospectroscopic measurements were conducted denoted as Dataset 3. This is included in the complete loading history presented in Table 1. For the collection of this dataset, the experimental setup was considerably improved and as such allowed for an extensive scan with additional accuracy with continuous rotation of the sample. A Thor Lab’s rotational stage was used to rotate the specimen after a 200 point vertical line scan, rotating 5.5◦ for a total of 65 line scans. Fig. 3 shows the configurations of photoluminescence data mapping for the three datasets. 2.2. Analysis methods For analysis a pseudo-Voigt curve-fitting algorithm was employed and used to identify the properties of the spectral peaks which allows for an understanding of the TGO mechanics [30,38]. The resolved peaks are shifted to the left (smaller wavenumber) indicating a compressive stress in the oxide layer. This value is compared to the known reference for stress free a-alumina R1 and R2 peaks [39,40]. Studying the contour maps of the peak number

Mechanical load [MPa]

Internal cooling [SLPM]

Time at temp

16 to 128

0 to 100

17 h total

N/A 16 to 128

N/A 0 to 100

304 h 17 h total

and its associated shifts provide an examination of the stress profile variation of the coating, and can be a powerful tool for qualitative investigations. The relationship between the change in wavenumber and its corresponding stress value magnitude is represented by following tensorial piezospectroscopic equation [41,42]. Dm = Pij • sij

(1)

Here P ij represents the piezospectroscopic coefficients, and Dm represents the change in wave number from the peak shift. Here, the oxide scale has variable thickness over its lifetime ranging from 0.5 to 5 lm and as such the thickness values were idealized as thin films due to the geometry. This allowed the use of the biaxial assumption, and led to the modified piezospectroscopic equation [43] presented in Eq. (2). Dm = (2/3) • (P11 + P22 + P33 ) • sb

(2)

In Eq. (2), s b is defined as the biaxial stress average and is found by comparing the wave number shift for the characteristic peak with the corresponding piezospectroscopic coefficient. A piezospectroscopic coefficient of 7.59 cm −1 /GPa for R1 and 7.61 cm −1 /GPa for R2 was used in conjunction with the zero stress reference [39,40,44]. The fitting procedures are accurate to very small wavenumber shifts, yielding high precision with standard deviations of less than 0.01 cm −1 [38]. 3. Results and discussion The photoluminescence measurements were taken following each loading application outlined in Table 1, and the results are

Dataset: 3

Datasets: 1 & 2

65 steps (from 0 to 354.6°) D 40mm

Bond Coat

Substrate

8 mm

A

C

B

200 vertical points

Bond Coat

8 cm

20 Horizontal scan locations

5.54°

400 µm

100 Vertical scan locations

Substrate

TGO YSZ

TGO YSZ

Fig. 3. Data collection methods for Datasets 1 and 2, and advanced setup for Dataset 3.

Please cite this article as: A. Manero et al., Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments, Surface & Coatings Technology (2016), http://dx.doi.org/10.1016/j.surfcoat.2016.09.057

ARTICLE IN PRESS A. Manero et al. / Surface & Coatings Technology xxx (2016) xxx–xxx

specimen with the least thermal exposure as seen in Supplementary Fig. S3. This suggests that the center region with highest thermal exposure consists of mostly transformed a alumina. The magnitude of biaxial stresses fall within the range of similar TBC configurations on flat samples after isothermal aging in Gell et al. [49] The slightly lower values obtained in our data highlight that internal cooling in TGMF cycling has an impact on the development of biaxial stress in the TGO.

presented in the following sections. First, the results and findings unique to this study linking loading history and corresponding stress, determined from photoluminescence measurements, are discussed considering the following factors in combination including: tubular sample geometry, spatial mapping, and measurement techniques. During damage evolution a unique bimodal peak distribution, characterized by the appearance of two distinct peaks observed within the R1 peak, develops when the measurement zone (or laser spot size) captures spatial points that have multiple stress states [27,45]. Here, the implications of highly localized bimodal spectra, observed after long duration isothermal aging and subsequent thermal gradient mechanical fatigue loading, from Datasets 2 and 3, are further investigated.

3.2. Photoluminescence results (Dataset 2) — after additional isothermal aging for 300 h Dataset 2 represents the measurements conducted on the specimen following the additional isothermal aging in the range of 300 h at high temperature without mechanical loading or internal cooling. TGO growth is known to evolve in three main phases [50], where initial growth is rapid and non-linear due to exposure to thermal cycling. Compared with Dataset 1, the TGO layer thickness has increased significantly; however both Datasets 2 and 3 share similar TGO thicknesses. Here, the stress distribution map of about 4000 points for the half-sample is shown in Fig. 5. As the isothermal loading conditions were nearly uniform, the previous effects of stress variation, due to the temperature difference between the primary heating zone and the adjacent region, appear to be removed. The centerline stress was plotted and showed a biaxial stress average of 2.70 GPa with a standard deviation of 516.04 MPa. A line further to the top of the sample shows a similar average of about 2.71 GPa. Comparing the centerlines of the specimen before and after the isothermal aging (between Dataset 1 and Dataset 2) shows an increase in biaxial average stress of approximately 0.4 GPa. All stress maps have been generated using the R1 spectral peaks, due to the higher intensities and subsequently diminished influence of bimodal stress on the fitting procedures. An example comparison of contour plots from R1 and R2 peaks for Dataset 2 is presented in Supplementary Fig. S2 which show minimal variation in trends for the two maps.

3.1. Photoluminescence results (Dataset 1) — early TGO with TGMF cycling for 17 h The cycle configuration preceding the collection of Dataset 1 was approximately 17 h of TGMF cycles as indicated in Table 1. These maps represent 4000 points in total over half of the sample length as shown in Fig. 4. For the half-sample, the center of the primary heated zone is located at the bottom of the scan zone. From the map, a distinct variation in the stresses between the regions within the heat zone (closer to the center line) and outside of it (closer to the top of the sample) can be identified. The stresses for the primary heated zone, which was held near uniform at 1000 ◦ C, are appreciably lower than in the rest of the map. The centerline scan showed a biaxial stress average of 2.38 GPa with a variation of 373.1 MPa. In comparison, a line outside the heat zone showed an average stress of about 2.78 GPa. It can be deduced that this is the result of the larger thermal drop across the layers due to heater focusing, induced thermal gradients due to internal cooling, and the influence of heat flux to the cooled specimen clamps. This is most significant for this sample configuration due to the mechanisms of oxide growth, where it is known that early cycling is accompanied by phase transformation [46] and rapid growth of the TGO [47,48]. The map in Fig. 4 suggests that the initial growth of the oxide scale generates a relaxation in stress likely due to volume reduction that accompanies phase transformation [46], as seen in the reduction of stresses in the center of the specimen. This was supported by the fact that while the entire map features a phase alumina, the only discernible levels of h-alumina appeared in the Dataset 1 near the upper part of the

3.3. Photoluminescence results (Dataset 3) — additional TGMF cycling for 17 h Following the additional TGMF cyclic loading, photoluminescence measurements were conducted with a high-resolution rotational stage. The 13,000 point map is presented in Fig. 6. The additional

Stress Map Top of specimen

Spatial Position (mm)

4

0.5 0

8

-2.5

12 16

-2 20

-1.5

24 28

-1

32

-0.5 -1 -1.5 -2 -2.5 -3

-0.5

36

Center line

Mean = 2.38 GPa Standard Deviation= 373.1 MPa

-3

Biaxial stress (GPa)

Out of heat zone line

Centerline Stress 1

-3.5

0

5

-3.5

8

16

24

Spatial Position (mm)

0

-4 0

5

10

15

20

25

30

35

40

Pixel index along centerline circumference

Fig. 4. Early cycled stress map for Dataset 1 presenting scanned region along with the variation of centerline stress.

Please cite this article as: A. Manero et al., Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments, Surface & Coatings Technology (2016), http://dx.doi.org/10.1016/j.surfcoat.2016.09.057

ARTICLE IN PRESS 6

A. Manero et al. / Surface & Coatings Technology xxx (2016) xxx–xxx

Centerline Stress

Stress Map -3.5

Out of 4 heat zone 8 line

-3

1 0.5

Mean = 2.70 GPa Standard deviation =516.0 MPa

0

-2.5

12 16

-2 20

-1.5

24 28

-1

Biaxial stress (GPa)

Spatial Position (mm)

Top of 0 specimen

32

-0.5 -1 -1.5 -2 -2.5 -3

-0.5

36

-3.5

Center line

8

16

-4

0

24

0

5

Spatial Position (mm)

10

15

20

25

30

35

40

Pixel index along centerline circumference

Fig. 5. Stress distribution maps for Dataset 2, presenting scanned region along with the variation of centerline stress.

2.70 GPa for Dataset 2 and 4.57 GPa for Dataset 3. The evolution of the TGO stresses has been characterized as consisting of three stages over the lifetime; an initial increase in magnitude, stabilization and a sharp drop in the magnitude of stress [50]. However, the rate and cyclic age defining the stages that make up this typical behavior varies with the TBC configuration including factors such as bond coat type and surface preparation. Similar configurations of MCrAlY bond coat samples cycled with maximum temperatures up to 1121 ◦ C for 40 min have shown that the biaxial stress increases to a range of 2.5 to 4.5 GPa even up to 200 thermal cycles before stabilization and a sharp decline [50]. The results presented indicate that our datasets are within the first stage of increasing magnitude of biaxial stress. It further highlights that the thermal gradient mechanical loading has a significant effect on the TGO development from Dataset 3. Further investigation on an As-coated sample without aging as well

TGMF loading is seen to change the stress distribution from the near uniform condition following isothermal loading in Fig. 5 and presents increased magnitude of biaxial stress in the heat zone. The centerline data shows a very uniform biaxial average stress of 4.57 GPa with a standard deviation of only 78 MPa. A line outside the heat zone reveals a lower magnitude of stress in the range of 3.52 GPa. At this stage, the TGO is generally composed of a-alumina. The increase in biaxial stress with thermal gradient mechanical loads highlights the significant influence of loading and cycling on the TGO development. A selected characteristic peak for each dataset has been presented in Supplementary Fig. S4 showing distinct R-lines used for the analysis of peak positions leading to stress determination. Comparing all three datasets over the history of the specimen shows an increase in magnitude of biaxial stress over the aging history with centerline average stresses at 2.38 GPa for Dataset 1,

Stress Map

Out of heat zone line -5 -4.5

16

-4

24

-3.5

32

-3

40

-2.5

48

-2

56

-1.5

64

Center line

-0.5

80 16

-4.3

-4.4

-4.5

-4.6

-1

72

8

Mean = 4.57 GPa Standard deviation = 78.03 MPa

-4.2

Biaxial stress (GPa)

Spatial Position (mm)

8

Centerline stress

-4.1

-4.7

0

24

-4.8

Spatial Position (mm)

0

10

20

30

40

50

60

70

Pixel index along centerline circumference

Fig. 6. Stress map of Dataset 3 along with the variation of centerline stress.

Please cite this article as: A. Manero et al., Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments, Surface & Coatings Technology (2016), http://dx.doi.org/10.1016/j.surfcoat.2016.09.057

ARTICLE IN PRESS A. Manero et al. / Surface & Coatings Technology xxx (2016) xxx–xxx

as studies of additional aging of the sample will enable a validation of lifetime trends as the specimens experience further stages of TGO growth for comparison. Considering the stress variation in the heat zones among the three datasets, it was observed that long duration isothermal aging removes the stress variation over the axial direction that was attributed to complex thermal gradients. However, this type of loading which is more representative of in-service conditions, will generate stress profile variations that in turn affect damage mechanisms. Sharp gradients in stress are seen at highly localized regions in maps of Figs. 5 and 6, which were then further investigated in Section 3.4. Photoluminescence measurements on an actual blade geometry have shown that curvature can trigger changes in the location at which failure initiates, within or at the interface of the TGO [51]. The effects of the geometry for the tubular specimens used in this experiment are not significant due to the larger radius of the specimens. For specimens with a radius of curvature in excess of 4 mm, minimal variations in axial, circumferential, and radial stress measurements have been observed in literature [51]. Other studies investigating the influence of concave and convex geometries have also demonstrated that trends follow flat plate specimens as the radius of curvature increases [52]. 3.4. Micro-damage identification Of particular interest were data points that showed bimodal stress response, thought to be indicators of preliminary damage, presented in the raw data of Fig. 7a. This phenomenon has been reported in literature and the response is expected as the specimens progress

Contribution from undamaged TGO sections

in age [27,45]. Bimodal stress states occur when some of the alumina crystals shift to a zero stress state while the overall probed volume still shows the presence of a high biaxial stress state. These correspond to stress states in flakes of spalled alumina [50] and alumina still attached to the bond coat, respectively. The regions featuring presumed micro-damage is called out as Location 1 in the spatial map of Fig. 7b from regions marked in Fig. 5. A region of higher stress can be identified on one side of the stress free location, accompanied by a reduction in stress on the other front from the local spatial average. Close by, a pocket of considerably higher stress is observed marked as Location 2. While the majority of the map in Fig. 7b is fairly uniform at approximately 2.75 to 3 GPa of stress, the localized relaxation associated with the beginning of micro-spallation and damage is visible from this mapping. This type of relaxation has also been attributed to potential rumpling effects, which can cause a bending state, which then precedes damage [49]. The bimodal peaks indicate multiple stress states as observed through the broadening of the spectral peaks and presence of smaller secondary characteristic spectral peaks for a-alumina. A closer examination of the raw spectra in Fig. 7a, shows the presence of a quartet of peaks. It appears that as the deviation between the stress states is so large and localized, a fit of the standard doublet indicates approximately 3 GPa difference between two dominant stress states of 0 and 3 GPa, respectively. The measurement location in the laser scanning diameter of 0.4 mm, which is sufficiently large to capture small zones of damage in tandem with healthy zones. Fig. 7c demonstrates schematically how the superposition of the bimodal spectra results in characteristic signs of a split stress state, including the leftward peak split and trough deviation. Fig. 7d from the location marked in Fig. 6 shows similar variations at locations in Dataset 3. Capturing this image in

Loc. 1

Unstressed peak position

-3.5

b)

-3

Raw

-1 -0.5

14400

14450

Dataset 2

4

8x 10

High stress peaks Low stress peaks Quartet

7

Intensity (A.U.)

-2 -1.5

Absolute Wavenumber (cm-1)

c)

-2.5

Loc. 2

6

d)

0.5

Loc. 1 -5 -4.5 -4

Loc. 2

5

0

-3.5

Biaxial stress (GPa)

1200 1000 800 600 400 200 0 -200 14350

Intensity (A.U.)

a)

7

-3

4

-2.5

3

-2

2

-1.5

1

-1

0 14300 14350

14400 14450

14500 14550

Dataset 3

-0.5 0

Absolute Wavenumber (cm-1) Fig. 7. a. Bimodal stress state identified in raw Dataset 2, b. marked subsection of Dataset 2 (in Fig. 5) showing location with unstressed or low stress peaks (denoted as Loc. 1) and high biaxial stress (denoted as Loc. 2), c. schematic representation of bimodal quartet, and d. marked subsection of Dataset 3 (in Fig. 6) showing location with unstressed or low stress peaks (denoted as Loc. 1) and high biaxial stress gradient from primary heated zone (denoted as Loc. 2). Note: for clarity, the locations highlighted in b) and d) are not spatial equivalent.

Please cite this article as: A. Manero et al., Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments, Surface & Coatings Technology (2016), http://dx.doi.org/10.1016/j.surfcoat.2016.09.057

ARTICLE IN PRESS 8

A. Manero et al. / Surface & Coatings Technology xxx (2016) xxx–xxx

the map provides a look into the mechanics of thermally grown oxide micro-spallation, whereby a small segment of approximately 500 lm appears to show signs of micro-debonding. Such spectroscopic evidence of local debonding, seen in the results herein, have been observed by researchers and presented in literature [23-25,35,44,45,49-51]. When the examination laser spot size is larger than the local anomaly, broadening and convolution of the peaks that represent superposition of the luminescence over an area exhibiting multiple stress states have been reported [27,45]. Over the lifetime of the turbine blade, these local damaged zones are thought to propagate and merge which can lead to the eventual failure of the coating. Another possible feature that could cause the variation in the optical peaks is that of the formation of porosities [53]. In addition, studies on curved blade samples with PtAl bond coats have revealed mixed oxides of a and h-alumina remaining even in cycled samples that lead to variation in stress states and corresponding bimodal spectra [51]. This possibility can be ascertained by examining the spectra to identify any h-alumina peaks, which were not evident in our Datasets 2 and 3. While the circumstances leading to the development of the micro-damage have been attributed to several factors including chemical composition, interface morphology, curvature and microstructural changes in the TGO [51], the non-destructive measurements in this work have shed light on the presence of such damage in our specimen with thermal aging and complex loading history. Modifications to enable fitting of multiple peaks will be pursued to enable accurate assessments of deviations within the multiple stress states. Further investigations with additional hours of aging can be used to monitor the observed micro-damage and its evolution to understand the effects of thermal gradient mechanical loads on formation and development of the micro-spallation for the corresponding TBC configuration.

4. Concluding remarks This study demonstrated the ability to conduct high resolution stress distribution maps for EB-PVD coating systems on tubular specimens under ex-situ conditions at various stages of cycling and was conducted to compliment previous in-situ synchrotron studies with realistic operational environments [15,32]. The specimen history at each stage of investigation was documented including 1) as-coated with few thermal gradient mechanical loading cycles, 2) isothermal aging in the range of 300 h at 1000 ◦ C, and 3) further aging under thermal gradient mechanical loading. For the case of thermal gradient mechanical loading, the piezospectroscopic studies validated the uniformity of stress development in the TGO within the central location of the primary heat zone where diffraction measurements were taken. However, regions outside the heat zone were distinctly different. Isothermal aging was seen to remove the deviation of residual stresses that exists due to proceeding complex loading conditions. In a real turbine blade such non-uniform loading conditions are expected and such deviation of stresses across the blade will likely be present. Following the aging process, evidence of micro-damage was observed in the thermally grown oxide. These zones with relaxed stresses were seen to have an effect on the coating around them, as large and localized variations in stress were observed. The anomalies identified were smaller than the 0.4 mm laser excitation spot size, resulting in a bimodal measurement for the collected area. These deviation zones at or near zero stress were accompanied by a surrounding area where the biaxial stress was found to be 3 GPa. They are thought to be micro-damage zones. As cycling continues, such zones with such stress gradients are expected to increase and eventually coalesce to form damage zones accompanied by relaxation of stresses. The data presented here shows that throughout the loading history, biaxial average stresses at centerline increased

from approximately 2 GPa toward 4 GPa suggesting that the aging remains within the first stage of coating life [50] and further aging will likely result in relaxation of stresses. Further, the methods used for this study showcase a way to identify the preliminary signs of micro-damage revealing that spallation is the final result of a slow process where micro-scale damage eventually expand and coalesce over long duration aging. The mapping of the TGO biaxial stress allows a unique insight into its evolution. These techniques could advance a non-destructive method to assess damage and predict life. Acknowledgments This material is based upon work supported by the National Science Foundation grants (Grant Nos. OISE 1460045,DMR 1337758, and CMMI 1125696), the German Science Foundation (DFG) grant (Grant No. SFB-TRR103), Project A3, and the Fulbright Academic Grant (Grant No. 34142765). Appendix A. Supplementary data Supplementary data to this article can be found online at http:// dx.doi.org/10.1016/j.surfcoat.2016.09.057.

References [1] R. Miller, Current status of thermal barrier coatings an overview, Surf. Coat. Technol. 30 (1987) 1–11. [2] N.P. Padture, M. Gell, E.H. Jordan, Thermal barrier coatings for gas-turbine engine applications, Science 296 (2002) 280–284. [3] D. Zhu, R.A. Miller, D.S. Fox, Thermal and environmental barrier coating development for advanced propulsion engine systems, Tech. rep., Glenn Research Center. 2008. [4] D.R. Clarke, M. Oechsner, N.P. Padture, Thermal-barrier coatings for more efficient gas-turbine engines, MRS Bull 37 (10) (2012) 891–898. [5] B. Bewlay, M. Jackson, P. Subramanian, J.-C. Zhao, A review of very-high-temperature Nb-silicide-based composites, Metall. and Mater. Trans. A. 34 (10) (2003) 2043–2052. [6] S. Meier, D. Gupta, The evolution of thermal barrier coatings in gas turbine engine applications, J. Eng. Gas Turbines Power 116 (1) (1994) 250–257. [7] S. Sampath, U. Schulz, M.O. Jarligo, S. Kuroda, Processing science of advanced thermal-barrier systems, MRS Bull. 37 (10) (2012) 903–910. [8] M. Stiger, N. Yanar, M. Topping, F. Pettit, G. Meier, Thermal barrier coatings for the 21st century, Z. Met. 90 (12) (1999) 1069–1078. [9] A. Evans, M. He, J. Hutchinson, Mechanics-based scaling laws for the durability of thermal barrier coatings, Prog. Mater. Sci. 46 (2001) 249–271. [10] M. Gell, E. Jordan, K. Vaidyanathan, K. McCarron, B. Barber, Y. Sohn, V. Tolpygo, Bond strength, bond stress and spallation mechanisms of thermal barrier coatings, Surf. Coat. Technol. 120-121 (4) (1999) 53–60. [11] A.M. Karlsson, J.W. Hutchinson, A.G. Evans, A fundamental model of cyclic instabilities in thermal barrier systems, J. Mech. Phys. Solids 50 (2002) 1565–1589. [12] A.G. Evans, M.Y. He, A. Suzuki, M. Gigliotti, B. Hazel, T.M. Pollock, A mechanism governing oxidation-assisted low-cycle fatigue of superalloys, Acta Mater. 57 (2009) 2969–2983. [13] M.T. Hernandez, A.M. Karlsson, M. Bartsch, On TGO creep and the initiation of a class of fatigue cracks in thermal barrier coatings, Surf. Coat. Technol. 203 (2009) 3549–3558. [14] R.O. Diaz, In-Situ Stress Measurements of EB-PVD Thermal Barrier Coatings Using Synchrotron X-ray Diffraction Under Thermo-Mechanical Loading, University of Central Florida. 2010. Master’s thesis [15] K. Knipe, A. Manero, S.F. Siddiqui, C. Meid, J. Wischek, J. Okasinski, J. Almer, A.M. Karlsson, M. Bartsch, S. Raghavan, Strain response of thermal barrier coatings captured under extreme engine environments through synchrotron x-ray diffraction, Nat. Commun. 5 (2014). [16] D. Lipkin, D. Clarke, Sample-probe interactions in spectroscopy: sampling microscopic property gradients, J. Appl. Phys. 77 (1995) 1855–1863. [17] S.E. Molis, D.R. Clarke, Measurement of stresses using fluorescence in an optical microprobe: stresses around indentations in a chromium-doped sapphire, J. Electron. Struct. Ceram. 73 (11) (1990) 3189–3194. [18] S.Margueron, D. Clarke, Effect of residual stress on the luminescence lifetime of r-line emission from polycrystalline alumina formed by oxidation, J. Am. Ceram. Soc. 90 (2007) 1798–1801. [19] R. Christensen, D. Lipkin, D. Clarke, Nondestructive evaluation of the oxidation stresses through thermal barrier coatings using Cr3 + piezospectroscopy, Appl. Phys. Lett. 69 (1996) 3754–3756.

Please cite this article as: A. Manero et al., Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments, Surface & Coatings Technology (2016), http://dx.doi.org/10.1016/j.surfcoat.2016.09.057

ARTICLE IN PRESS A. Manero et al. / Surface & Coatings Technology xxx (2016) xxx–xxx [20] W.A. Ellingson, R.J. Visher, R.S. Lipanovich, C.M. Deemer, Optical NDT methods for ceramic thermal barrier coatings, Mater. Eval. 64 (1) (2006) 45–51. [21] C. Giolli, A. Scrivani, G. Rizzi, F. Borgioli, G. Bolelli, L. Lusvarghi, Failure mechanism for thermal fatigue of thermal barrier coating systems, J. Therm. Spray Technol. 18 (2009) 223–230. [22] B. Heeg, V.K. Tolpygo, D.R. Clarke, Damage evolution in thermal barrier coatings with thermal cycling, J. Am. Ceram. Soc. 94 [S1] (2011) S112–S119. [23] X. Peng, D. Clarke, Piezospectroscopic analysis of interface debonding in thermal barrier coatings, J. Am. Ceram. Soc. 83 (2000) 1165–1170. ´ [24] G. Moskal, L. Swadzba, ´ B. Mendala, M. Góral, M. Hetmanczyk, Degradation of the TBC system during the static oxidation test, J. Microsc. 237 (3) (2010) 450–455. [25] K. Ogawa, T. Shoji, H. Aoki, N. Fujita, T. Torigoe, Mechanistic understanding for degraded thermal barrier coatings, JSME Int. J. Ser. A 44 (4) (2001) 507–513. [26] Y. Zhang, P. Withers, M. Fox, D. Knowles, Damage mechanisms of coated systems under thermomechanical fatigue, Mater. Sci. Technol. 15 (9) (1999) 1031–1036. [27] B. Jayaraj, Correlating Microstructural Development and Failure Mechanisms to Photo Stimulated Luminescence Spectroscopy & Electrochemical Impedance Spectroscopy in Thermal Barrier Coatings, University of Central Florida, Orlando, Florida, 2011. (Ph.D. thesis) [28] G. Freihofer, L. Poliah, K. Walker, A. Medina, S. Raghavan, Optical stress probe: in-situ stress mapping with Raman and photo-stimulated luminescence spectroscopy, J. Instrum. 5 (2010) 12003. [29] G. Freihofer, D. Fugon-Dessources, E. Ergin, A.V. Newkirk, A. Gupta, S. Seal, A. Schülzgen, S. Raghavan, Piezospectroscopic measurements capturing the evolution of plasma spray-coating stresses with substrate loads, ACS Appl. Mater. Interfaces 666 (2014) 1366–1369. [30] G. Freihofer, A. Schülzgen, S. Raghavan, Multiscale mechanics to determine nanocomposite elastic properties with piezospectroscopy, Acta Mater. 81 (2014) 211–218. [31] S.F. Siddiqui, K. Knipe, A. Manero, C. Meid, J. Wischek, J. Okasinski, J. Almer, A.M. Karlsson, M. Bartsch, S. Raghavan, Synchrotron X-ray measurement techniques for thermal barrier coated cylindrical samples under thermal gradients, Review of Scientific Instruments 84 (8) (2013) 083904. [32] A. Manero, II, S. Sofronsky, K. Knipe, C. Meid, J. Wischek, J. Okasinski, J. Almer, A.M. Karlsson, S. Raghavan, M. Bartsch, Monitoring local strain in a thermal barrier coating system under thermal mechanical gas turbine operating conditions, JOM 67 (7) (2015) 1528–1539. [33] N.P. Padture, K.W. Schlichting, T. Bhatia, A. Ozturk, B. Cetegen, E.H. Jordan, M. Gell, S. Jiang, T.D. Xiao, P.R. Strutt, E. Garcia, P. Miranzo, M.I. Osendi, Towards durable thermal barrier coatings with novel microstructures deposited by solution-precursor plasma spray, Acta Mater. 49 (12) (2001) 2251–2257. [34] A. Maricocchi, A. Bartz, D. Wortman, PVD TBC Experience on GE aircraft engines, J. Therm. Spray Technol. 6 (2) (1997) 193–198. [35] X. Wang, A. Atkinson, Piezo-spectroscopic mapping of the thermally grown oxide in thermal barrier coatings, Mater. Sci. Eng. A 465 (1) (2007) 49–58.

9

[36] A. Manero, II, Nondestructive Analysis of Advanced Aerospace Materials via Spectroscopy and Synchrotron Radiation, The University of Central Florida. 2014. Master’s thesis [37] A. Misu, Emissions of ruby, J. Physical Soc. Japan 19 (12) (1964) 2260–2270. [38] A. Stevenson, A. Jones, S. Raghavan, Stress-sensing nanomaterial calibrated with photostimulated luminescence emission, Nano Lett. 11 (2011) 3274–3278. [39] C. Baudín, J. Gurauskis, A.J. Sánchez-Herencia, V.M. Orera, Indentation damage and residual stress field in alumina-Y2 O3 -stabilized zirconia composites, J. Am. Ceram. Soc. 92 (1) (2009) 152–160. [40] D.C.J.A. Nychka, Damage quantification in TBCs by photo-stimulated luminescence spectroscopy, Surf. Coat. Technol. 146-147 (2001) 110–116. [41] L. Grabner, Spectroscopic technique for the measurement of residual stress in sintered Al2 O3 , J. Appl. Phys. 49 (5) (1978) 580–583. [42] J. He, D.R. Clarke, Determination of the piezospectroscopic coefficients for chromium doped sapphire, 78 (5) (1995) 1347–1353. [43] J.A. Nychka, D.R. Clarke, Damage quantification in TBCs by photo-stimulated luminescence spectroscopy, Surf. Coat. Technol. 146-147 (2001) 110–116. [44] S. Sridharan, L. Xie, E.H. Jordan, M. Gell, Stress variation with thermal cycling in the thermally grown oxide of an EB-PVD thermal barrier coating, Surf. Coat. Technol. 179 (2004) 286–296. [45] V.K. Tolpygo, D. Clarke, K. Murphy, Evaluation of interface degradation during cyclic oxidation of EB-PVD thermal barrier coatings and correlation with TGO luminescence, Surf. Coat. Technol. 188 (2004) 62–70. [46] B.W. Veal, A.P. Paulikas, P.Y. Hou, Tensile stress and creep in thermally grown oxide, Nat. Mater. 5 (2006) 349–351. May. [47] M.T. Hernandez, D. Cojocaru, M. Bartsch, A.M. Karlsson, On the opening of a class of fatigue cracks due to thermo-mechanical fatigue testing of thermal barrier coatings, Comput. Mater. Sci. 50 (2011) 2561–2572. [48] J. Toscano, D. Naumenko, A. Gil, L. Singheiser, W. Quadakkers, Parameters affecting TGO growth rate and the lifetime of TBC systems with MCraly-bondcoats, Mater. Corros. 59 (6) (2008) 501–507. [49] M. Gell, S. Sridharan, M. Wen, E.H. Jordan, Photoluminescence piezospectroscopy: a multi-purpose quality control and NDI technique for thermal barrier coatings, Int. J. Appl. Ceram. Technol. 1 (4) (2004) 316–329. [50] Y. Sohn, K. Vaidyanathan, M. Ronski, E. Jordan, M. Gell, Thermal cycling of EB-PVD/MCraly thermal barrier coatings: II. Evolution of photo-stimulated luminescence, Surf. Coat. Technol. 146 (2001) 102–109. [51] X. Wang, G. Lee, A. Atkinson, Investigation of TBCs on turbine blades by photoluminescence piezospectroscopy, Acta Mater. 57 (1) (2009) 182–195. [52] W. Mao, J. Jiang, Y. Zhou, C. Lu, Effects of substrate curvature radius, deposition temperature and coating thickness on the residual stress field of cylindrical thermal barrier coatings, Surf. Coat. Technol. 205 (8) (2011) 3093–3102. [53] W. Quadakkers, V. Shemet, D. Sebold, R. Anton, E. Wessel, L. Singheiser, Oxidation characteristics of a platinized MCraly bond coat for TBC systems during cyclic oxidation at 1000 ◦ C, Surf. Coat. Technol. 199 (1) (2005) 77–82.

Please cite this article as: A. Manero et al., Piezospectroscopic evaluation and damage identification for thermal barrier coatings subjected to simulated engine environments, Surface & Coatings Technology (2016), http://dx.doi.org/10.1016/j.surfcoat.2016.09.057