PROTECTIVE COATINGS. A Dissertation. Presented to. The Graduate Faculty of The University of Akron. In Partial Fulfillment

Transcription

1 HIGH TEMPERATURE DAMAGE CHARACTERIZATION OF CERAMIC COMPOSITES AND PROTECTIVE COATINGS A Dissertation Presented to The Graduate Faculty of The University of Akron In Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy Matthew P. Appleby May, 2016

2 HIGH TEMPERATURE DAMAGE CHARACTERIZATION OF CERAMIC COMPOSITES AND PROTECTIVE COATINGS Matthew P. Appleby Dissertation Approved: Accepted: Advisor Dr. Gregory Morscher Committee Member Dr. Manigandan Kannan Committee Member Dr. Kwek Tze Tan Committee Member Dr. Craig Menzemer Department Chair Dr. Sergio Felicelli Interim Dean of the College Dr. Eric Amis Dean of the Graduate School Dr. Chand Midha Date Committee Member Dr. Alper Buldum ii

3 ABSTRACT Novel high-temperature experiments were conducted in ordered to address some of the most critical life-limiting issues facing woven melt-infiltrated, silicon carbide (SiC) fiber-reinforced SiC ceramic matrix composites (CMCs) as well as protective thermal and environmental barrier coatings (T/EBC). Heating of specimens was achieved using laserbased approaches that simulate the high heat-flux thermal gradient environments that these materials will be subjected to in service. Specialized non-destructive evaluation (NDE) and inspection techniques were developed to investigate damage modes and material response. First, in order to examine the capabilities of utilizing the emerging technique of electrical resistance (ER) measurement for use in high temperature mechanical testing in SiC/SiC CMCs, the temperature dependent ER response of several systems was determined. A model was developed to establish the contribution to overall ER from the individual composite constituents and applied thermal gradient. Then, elevated temperature tensile tests were performed to characterize the damage of composite materials to localized stress concentrations. Further experiments were done to assess the differences in damage mechanisms and retained tensile strength properties of uncoated SiC/SiC CMCs and EBC-CMC systems after prolonged exposure to high pressure, high velocity water vapor containing environments. Differences in damage modes were described using ER monitoring and post-test inspection. Localized iii

4 strain fields were measured using a novel digital image correlation (DIC) technique and stress-dependent matrix crack accumulation was monitored using in-situ modal acoustic emission (AE). Coupled AE and thermography measurements were also used to describe failure of protective ceramic coatings due to the life-limiting case of thermal cyclic loading. Due to the complex nature of T/EBC failure, the decrease in coating life and durability due to thermal stress concentrations and degradation via molten calciummagnesium-aluminosilicate (CMAS) infiltration was also examined. Finally, the use of ER measurements for damage characterization was extended to the complex case of creep and stress-rupture of damaged and undamaged composites as well as the dramatic increase in stress-rupture life to SiC/SiC CMCs from environmental barrier coatings. Post-test microscopy was performed to further explain differences in material response and damage morphology. iv

5 ACKNOWLEDGEMENTS I would like to thank my friend and academic advisor Dr. Gregory Morscher to whom I own a debt of gratitude that I fear I will never be able to fully repay. Without the gift of his knowledge and guidance none of this work could have been possible. My graduate school experience has been made all the more enjoyable by the rest of Dr. Morscher s research group at the University of Akron, of whom all have become close friends. Also, my early collaborations alongside previous group members Dr. Christopher Baker and Dr. Emmanuel Maillet have been some of the most gratifying times of my academic career and personal life. I would like to thank The Nation Aeronautics and Space Administration (NASA) for their financial support under the Graduate Student Researchers Program (GSRP) fellowship grant. My experimental work and research conducted at NASA Glenn Research Center could not have been possible without the direction and assistance from my NASA mentor and colleague Dr. Dongming Zhu. I would like to express my deepest appreciation for my loving fiancé Crystal. Her support and seemingly inhuman amount of patience has been my foundation through what has turned into a rather long and often meandering journey. Finally, I would like to thank God, he who gives me strength and through whom all things are possible. v

6 TABLE OF CONTENTS LIST OF TABLES... x Page LIST OF FIGURES... xi CHAPTER I. INTRODUCTION Ceramic Matrix Composites (CMCs) Thermal/Environmental Barrier Coatings... 6 II. MOTIVATION III. LITERATURE REVIEW Micromechanics of Tensile Damage of Ceramic Matrix Composites (CMCs) Damage monitoring and Non-Destructive Evaluation of SiC/SiC CMCs Acoustic Emission (AE) Electrical Resistance (ER) Digital Image Correlation (DIC) Thermal Cyclic Damage of Ceramic Coatings (T/EBCs) Influence of high heat-flux on coating damage vi

7 3.5 CMAS degradation of T/EBCs IV. TEMPERATURE DEPENDENT ELECTRICAL PROPERTIES OF MI-CVI SiCf/SiC CMCs Experimental Materials Experimental Procedure Results and Discussion Room Temperature Resistivity Temperature Dependence of Electrical Resistivity Isothermal Behavior and Parallel Constituent Model Effect of Thermal Gradient on Overall Electrical Response of Laser-heated Specimens Conclusions V. EFFECTS OF GEOMETRIC STRESS CONCENTRATIONS AND HIGH HEAT-FLUXES ON TENSILE DAMAGE OF CMCS Experimental Materials Monotonic tensile testing with high heat-flux capabilities Results and Discussion Mechanical Behavior AE Waveform Analysis Electrical Resistance Measurements Conclusions VI. ELEVATED TEMPERATURE, HIGH HEAT-FLUX EBC/CMC DAMAGE AND STRENGTH: POST ENVIRONMENTAL EXPOSURE vii

8 6.1 Experimental Materials High Pressure Burner Rig Exposure High-Temperature Monotonic Tensile Testing Results and discussion Retained properties ER and AE comparison AE waveform analysis Conclusions VII. COUPLED THERMOGRAPHY AND MODAL ACOUSTIC EMISSION CHARACTERIZATION OF THERMAL BARRIER COATING DAMAGE UNDER LASER HEAT-FLUX THERMAL CYCLING Experimental Materials Experimental Procedure Results and Discussion Conclusions VIII. CREEP AND STRESS-RUPTURE OF CMCS AND EBC/CMC SYSTEMS UNDER HIGH HEAT-FLUX, THERMAL GRADIENTS Experimental Materials Experimental Procedure Results and Discussion Tensile Creep and Electrical Resistance Post-test Damage Assessment of Creep Specimens viii

9 8.4 Conclusions IX. CONCLUSIONS X. FUTURE WORK BIBLIOGRAPHY APPENDICES APPENDIX A. DETERMINATION OF TEMPERATURE PROFILE USING SIMPLIFIED HEATTRANSFER MODEL APPENDIX B. ESTIMATION OF STEADY-STATE THERMAL STRESS FROM THROUGH THICKNESS TEMPERATURE GRADIENT ix

10 LIST OF TABLES Table Page 1: Overview of critical ceramic matrix composite damage mechanisms addressed : List of thermal and environmental barrier coating damage mechanisms affecting performance and life : Physical and geometrical properties of tested specimens; separated by reinforcing fiber type and test method (i.e. ZEM3 or Laser-based testing) : Room temperature electrical resistivity (ohm-mm) of specimens and corresponding average panel resistivity : SiC/SiC specimen geometry, fiber content and room-temperature electrical properties pre-tensile testing : Crack density measurements for each region (around the notch and far-field) of the gage section : SiC/SiC specimen geometry, fiber content (vf0), and room-temperature electrical properties pre-tensile testing : Conditions and mechanical properties of test samples during hightemperature tensile testing : Details of EB-PVD 7YSZ coated samples used in laser heat-flux thermal cyclic testing : Summary of results of AE analysis from thermal cyclic testing : Geometric properties of ZMI fiber reinforced SiCf/SiC test specimen gage sections, and EBC specifications x

11 LIST OF FIGURES Figure Page 1: Typical microstructures of (a) pre-preg and (b) slurry-cast MI-CMC materials [7] : Cross-section of an EB-PVD TBC on a turbine blade, superimposed with the temperature reduction provided by the coating [9] : Illustration of the volatilization of silica (SiO2) layer and corresponding SiC recession behavior in the presence of H2O containing combustion environment : A brief summarization of the design limits of high temperature aero-engine materials. The desire for increased temperature capabilities has driven the current material and coatings selection [125] : Brief outline of basic non-destructive evaluation (NDE) approaches used to address critical life-limiting behavior : Stress profile in the fibers (solid line) and the matrix (dotted line) for a cracked unidirectional composite at applied axial stress σc [28] : Stress profile in the fibers for a cracked composite, during (a) unloading from an applied axial stress σc to zero; (b) reloading from zero back to σc [28] : (a) Composite cylinder model used in analysis, along with axial fiber stress at different stages of (b) unloading from a peak stress σp, and (c) reloading when reverse slip reaching the end of the initial debond before full unloading [30] : (a) the stress around an isolated crack returning to the far-field stress over a slip length δ (by τ and stress jump via debond energy), and closely spaced cracks where the slip is limited to one half crack spacing; (b) slip regions (shaded) around multiple fiber breaks (X) in a composite. These fibers carry a pull out stress as they slip around the central cracking plane [37] : Schematic representation of the electrical model of a unit segment between two transverse matrix cracks [58] xi

12 11: SEM micrographs showing the microstructural and defect components of (a,b) APS and (c,d) EB-PVD TBC topcoats [63] : A schematic showing the general mechanisms driving thermal cyclic damage of: (a) APS and (b) EB-PVD ceramic coatings [9] : (a) ULVAC ZEM3 unit used for low/intermediate-temperature, isothermal characterization of SiCf/SiC specimens, (b) close-up of specimen configuration, including four-point probe configuration (note: entire specimen housed within furnace) : Schematic of laser-based heating apparatus used for characterization of hightemperature electrical properties of SiCf/SiC CMC tensile bars : Room temperature electrical resistivity measured from the centerline of the Hi-Nicalon Type S laser-heated sample, measurements were taken in the pretest/pristine condition (striped) and post-laser heating (solid) condition. Note that the heated region of the tensile bar is approximately ± 22 mm from centerline : Typical dependence of: (a) the carrier concentration in a doped semiconductor (constructed assuming a phosphorus-doped ND = /cm 3 Si sample). ni/nd versus T(dashed line) included for comparison purposes [97], (b) mobilities in n-type Si with different electron concentrations. Inset illustrates temperature dependence due to lattice and impurity scattering [94] : Temperature dependent electrical response of MI-CVI SiCf/SiC samples tested using the ZEM3 and laser-based heating technique respectively : The calculated electrical conductivity of the effective matrix materials (solid lines) of the (a) HNS and (b) SA reinforced ZEM3 tested samples respectively. The ZEM3 measured conductivity data (dotted lines) and fiber data [82] (dashed lines) used in the parallel circuit : The calculated longitudinal thermal gradient for the ZMI-A laser-heated sample. Note that each curve represents the temperature profile for a given hot-zone temperature to the end of the tensile specimen : Comparison of measured electrical resistance to hot-zone temperature of ZMI-A to proposed series resistance model : Notch geometry for the high-temperature tested double-notch SiC/SiC tensile bar xii

13 22: Image of tensile loading frame with incorporated high heat-flux laser heating and digital image correlation (DIC) apparatus : Schematic of laser heat-flux tensile test setup; including configuration of electrical resistance (ER) electrodes and modal acoustic emission (AE) sensors : (a) Estimate of through thickness thermal gradient based on measured front and back side temperatures of specimen heated-region. (b) Longitudinal thermal stresses in heated region as a function of sample thickness : Tensile stress-strain response of double notched specimen under high heatflux conditions. Note that the stress is the net-section stress and strain is the nominal mechanical strain recorded by the extensometer : Cumulative AE Energy of the sample: (a) normalized by the total by the total energy recorded in the notch region, and (b) also normalized by length to compare energy density : (a) Left: AE event location within the heated gage region (show as ± mm from the specimen centerline) versus net-section stress. Note that the size of the marker indicates the relative energy of the AE event. Right: The DIC image of the gage section taken just prior to failure. (b) Distribution of AE event energy along the gage length : Percent change in electrical resistance of the specimen versus nominal strain increase of the gage section during monotonic tensile loading to failure : Comparison of room temperature ER measurements taken along the length of the specimen (a) prior to any thermal or mechanical loading (i.e. pristine state), and (b) post high heat-flux tensile strength test. Note that the measurements refer to the distance in mm from the centerline of the sample : Mechanical behavior of samples under high temperature monotonic tensile loading, post environmental exposure. Strain measurement taken from high temperature extensometer, as well as from contact points of extensometer probes as observed from DIC images : Illustration of different matrix cracking extension. (a) Fiber-bridged matrix cracking, with associated debond at fiber/matrix interphase. (b) Unbridged matrix crack propagation. Note that the interphase thickness is exaggerated to show cracking within interfacial layer xiii

14 32: High temperature retained tensile strength test: (a) comparison of in-situ electrical resistance (ER) and acoustic emission (AE), and (b) change of ER with nominal mechanical strain : In-situ electrical resistance versus normalized cumulative AE energy (approximate normalized matrix cracking) during high temperature tensile tests of (a) coated and (b) uncoated samples : Distribution of AE energy along the gage length and associated DIC strain mapping of gage (at failure stress) during high temperature tensile testing for the (a) coated and (b) uncoated sample respectively : Stress-dependent AE events recorded by each sensor during high temperature monotonic tensile testing, showing calculated frequency centroid for the (a) coated sample (with low frequency cluster highlighted), and (b) uncoated sample, respectively : AE events during high temperature testing of coated sample with (a) gagelength location of events identified in low frequency (270 khz 375 khz) cluster, (b) Average AE event energies with low frequency events highlighted : SEM micrographs of coating cross-section depicting typical EB-PVD TBC coating crack morphologies. (a) Wedge-shaped through thickness vertical cracking and horizontal delamination. (b) Densified columns and delamination cracking caused by CMAS degradation : Schematic of laser heat-flux testing apparatus used for characterization of thermal cyclic behavior of EB-PVD 7YSZ coated specimens. Including details of high temperature modal AE measurement setup : Data collected from the thermal cyclic testing of the as-deposited TBC sample (Sample 1). (a) Thermography data of TBC surface and substrate back side temperatures and applied laser power. (b) Evolution in overall thermal gradient (ΔT) with corresponding accumulation of AE energy (total AE energy and breakdown of contribution of each phase of the thermal cycle) : Thermal cyclic testing of the sample with substrate containing partial through holes (Sample 2). (a) Thermography data of TBC surface and substrate back side temperatures and applied laser power. (b) Evolution in overall thermal gradient (ΔT) with corresponding accumulation of AE energy (total AE energy and breakdown of contribution of each phase of the thermal cycle) : (a) Histogram of Sample 2 testing showing recorded AE energy versus location (± mm from center of sample). (b) Image illustrating coating xiv

15 spallation area event that occurred during thermal cyclic testing of Sample : Thermal cyclic testing of the sample exposed to pre-test CMAS infiltration (Sample 3). (a) Thermography data of TBC surface and substrate back side temperatures and applied laser power. (b) Evolution in overall thermal gradient (ΔT) with corresponding accumulation of AE energy (total AE energy and breakdown of contribution of each phase of the thermal cycle) : Calculated frequency centroid of AE events, separated by which phase they occurred. Note that the size of the bubble is representative of the relative energy of the event. (a) Sample 1, (b) Sample 2, (c) Sample : Room temperature mechanical response of tensile specimens from the same composite panels as the ones used in this study. ZMI-1 and 2 are from panel A, and ZMI-3 is from panel B : (a) ER response and thermography data for initial loading (thermal followed by mechanical) of uncoated ZMI-1 sample. (b) Time dependent strain and ER change due to constant stress (69 MPa) under high temperature thermal gradient conditions in air of uncoated CMC sample ZMI : (a) ER response and thermography data for thermal and mechanical loading of ZMI-2. Initial heating cycles represent laser faults resulting in pre-test, thermal stress induced damage. Damage events evident in ER spikes during cool down (X), and residual increase in room temperature ER (O). (b) ER response of ZMI-2 during initial and post-thermal-stress-damage heating. The average CMC temperature assumes linear temperature distribution between measured surfaces. (c) Time dependent strain and ER change due to constant stress (69 MPa) under high temperature thermal gradient conditions in air of uncoated CMC sample ZMI : (a) Time dependent strain and ER change due to constant stress (69 MPa) under high temperature thermal gradient conditions in air of coated ZMI-3 sample. (b) Mechanical and electrical response of post-creep, retained high temperature strength testing of ZMI : Optical microscopy images showing matrix crack morphologies of (a) ZMI-1, (b) ZMI-2, (c) ZMI-3 as well as (d) typical cracking observed in the EBC deposited on ZMI-3 surface : At top: Composite SEM micrograph of ZMI-2 fracture surface. At bottom: higher magnification SEM micrographs of specific areas of the fracture surface to show lack of fiber pullout through thickness xv

16 50: At top: Composite SEM micrograph of EBC-CMC sample ZMI-3. At bottom: higher magnification images of specific areas of the fracture surface to highlight difference in apparent fiber pullout, progressing from near the coated surface to the back (uncoated) surface : Summary plot of in-situ ER measurements collected during testing of each sample to demonstrate the increase in electrical resistance due to thermal and mechanical loading and time-dependent effects : Temperature dependencies of thermal conductivity (k), natural convection coefficient (h), and emissivity (ε) used in the numerical solution of longitudinal thermal gradient : Beam of rectangular cross-section with through thickness thermal gradient xvi

17 CHAPTER I INTRODUCTION 1.1 Ceramic Matrix Composites (CMCs) Because of their high temperature mechanical capabilities and physical properties, ceramic materials are considered as an attractive option for use in many extreme environments including turbine engine hot-sections. However, due to the brittle nature of ceramic failure, some type of reinforcement phase is required in order to increase material toughness and mechanical durability. This need has led to the development and implementation of various fiber-reinforced ceramic matrix composites (CMCs). Ideal CMCs would possess various performance improvements over presently utilized metallic super-alloy components in turbine engines, whose current structural capability is limited to ~1100 C [1, 2]. The ideal structural CMC therefore, would possess increased operating temperature capability, decreased densities and the necessary toughness to prevent catastrophic failure. These attributes would lead to higher engine operating temperatures, reduced cooling requirements, and weight savings that would all lead to increased engine efficiencies and thrust [2]. Among the many CMC systems being studied, silicon carbide (SiC) fiber-reinforced SiC matrix CMCs were identified by NASA as the leading candidate for turbine engine core structures [2]. Specifically, the NASA 1

18 High Speed Research-Enabling Propulsion Materials (HSR-EPM) Program successfully demonstrated a combustor liner for hundreds of hours at operation temperatures of ~1200 C [3]. Beginning in the early 2000s, under the Ultra Efficient Engine Technologies (UEET) Program, NASA further developed the HSR-EPM material for capabilities up to 1315 C and expanded the program to include turbine vane components with next generation protective coatings [1]. General Electric (GE) has also dedicated a considerable effort into developing SiC/SiC composites for both aero and land-based turbines. GE is planning on introducing CMCs into service as a static first stage compressor shroud in their LEAP engine and have recently completed tests on a GEnx demonstrator engine with GE9x CMC components in both the combustor and turbine. Other major aero-engine manufacturers Rolls Royce and Pratt & Whitney have invested serious resources developing static and rotating turbine engine components using CMCs as well [4]. There are several processing methods that are used in manufacturing SiC/SiC composites with the most common being polymer infiltration and pyrolysis (PIP), chemical vapor infiltration (CVI), and liquid silicon (Si) melt-infiltration (MI) techniques. While these processes differ in the way that they densify the composite matrix, all of these methods begin with tows of commercially available SiC filaments that act as the continuous fiber-reinforcement for the composite system. These SiC fibers are coated with a thin layer using chemical vapor deposition (CVD), generally pyrolytic carbon (pyc) or boron nitride (BN), to create a weak interface with the matrix material. These fibers 2

19 are then arranged into any number of architectures, consisting of either laminate or woven fiber plys, designed to fit the desired geometry and structural application. The PIP method for creating CMCs takes the fibers/fiber layups and infiltrates them with either pure polymeric ceramic precursors or a ceramic filled liquid system in the form of a melt or slurry. The infiltrated plys are assembled and cured under pressure in an autoclave, essentially creating a continuous fiber reinforced polymer. From there the composite is decomposed into a ceramic via pyrolysis in a high temperature vacuum. The newly formed ceramic matrix is often re-infiltrated and the process is repeated in order to further densify the material [5]. Unlike, PIP CMCs the CVI process uses vapor means to build up a composite matrix. Once again a vapor deposition step is performed to deposit a thin interfacial coating on the SiC fiber tows. Then, in a furnace at temperatures above 800 C the fiber preform is exposed to a gaseous mixture that infiltrates the fibers and begins to build of a layer of SiC on the surface of the coated fibers. The SiC matrix deposited using the CVI process results is a relatively homogeneous β-sic with a fine grained microstructure. Unfortunately, the CVI SiC method has a major disadvantage in that the deposition on the fiber tows eventually closes off the interior of the composite resulting in relatively high fractions of open porosity. The level of porosity in CVI SiC/SiC CMCs directly controls the matrix cracking strength of the composite and the overall scatter in composite properties [6]. In order to produce SiC/SiC composites with a more reliable structure, two separate MI techniques have been developed to in order to produce near fully-dense composite matrices. The first is an MI pre-preg approach developed by General Electric for a 3

20 variety of engine components. It begins with a CVD fiber coating consisting of a BN layer followed by a protective silicon nitride (Si3N4) over-layer. The coated fiber tows are then allied into 2D tapes using a polymer-based binder that contains particles of SiC and carbon (C). These tapes are then stacked in various orientations to form the desired laminate architecture. The laminate is then heated in a pyrolysis step to decompose the polymer binder, and final matrix densification is performed by infiltration of molten silicon (or silicon alloy). The SiC matrix if therefore formed via reaction bonding of the silicon and the residual carbon in the preform [7]. The second MI techniques was developed under NASA HSR-EPM Program and consists of a woven SiC fabric that is stacked in order to produce the fiber preform. A layer of Si-doped BN is then deposited on the fibers, followed by a thin CVI SiC layer. This thin CVI SiC matrix helps to protect the interphase and fibers from the final densification steps in which SiC particles are slurry-cast in to the preform (near room temperature), followed once again by infiltration of molten silicon near 1400 C. This resulting material made from this fabrication method is often referred to as a MI-CVI matrix. Both the pre-preg and slurrycast MI techniques result in highly densified matrices that greatly increase the matrix cracking strength of the composite. However, the low melting temperature (1400 C) of the excess silicon left over from processing, limits the maximum usage temperature of MI SiC/SiC composites [1, 2]. Images illustrating the details of the composite microstructures of the two MI techniques are shown below in Figure 1. 4

21 (a) (b) Figure 1: Typical microstructures of (a) pre-preg and (b) slurry-cast MI-CMC materials [7]. 5

22 1.2 Thermal/Environmental Barrier Coatings The desire to increase turbine engine operation temperatures requires improvements in cooling design, materials and coatings. In the past increases in operating temperatures were achieved through improved alloy design, and internal air cooling delivered by channels cast into components. However, as the maximum temperature capabilities of super-alloy materials reached their inherent limits, one solution was the development of thermal barrier coatings (TBCs) that can be deposited on the existing metallic turbine components. TBCs were designed as multifunctional thick films (typically 100 μm to 2 mm) of a refractory material to protect the underlying metallic substrate from extreme temperatures in gas [8]. Figure 2 shows a typical TBC system consisting of: (1) a refractory top coat for thermal protection, (2) a thermally grown oxide (TGO) layer which bonds the TBC to the subsequent layer and slow the oxidation process, and (3) a bond coat layer that supplies the necessary elements to produce the TGO and provides oxidation protection and adhesion to the underlying substrate [9]. Currently, the most widely utilized TBC top coat consists of various forms of yttria-stabilized zirconia (YSZ), in particular the metastable tetragonal-prime structure [10, 11]. YSZ was selected primarily due to its relatively low thermal conductivity and good coefficient of thermal expansion (CTE) match with the nickel (Ni)-based superalloys. The most common methods of TBC deposition are air plasma spray (APS) and electron beam physical vapor deposition (EB-PVD). The choice of deposition technique is determined by the nature of the high-temperature environment and loading conditions, as the two techniques result in very different microstructures. 6

23 Figure 2: Cross-section of an EB-PVD TBC on a turbine blade, superimposed with the temperature reduction provided by the coating [9]. While work continues on the development new TBC systems, the demand for ever increasing operating temperatures, reduced air-cooling and increased weight savings is pushing the use of Ni-based super-alloys to their fundamental limits. Therefore, as previously mentioned, this had led to the development of SiC/SiC CMCs as an attractive replacement to ceramic coated super-alloy components. One typical advantage of Sibased materials is their excellent high temperature oxidation resistance in dry air. However, in the presence of hot combustion gases containing high velocity, high pressure water vapor, the protective silica (SiO2) layer formed at elevated temperatures volatilizes leading to rapid surface recession of the material [12-15]. A schematic of the basic oxidation and volatilization principles associated with SiC/SiC CMCs is shown in Figure 3. 7

24 Combustion gas SiO 2 + 2H 2O (g) = Si(OH) 4 (g) SiO 2 SiC Figure 3: Illustration of the volatilization of silica (SiO2) layer and corresponding SiC recession behavior in the presence of H2O containing combustion environment It is clear that if SiC/SiC CMCs are going to be used in engine environments then new state-of-the-art coatings are needed to not only withstand increased temperature requirements, but high-velocity corrosive combustion gasses as well. To meet this need, a large effort has been made to develop multilayer ceramic coatings know as Environmental Barrier Coatings (EBCs) in order to protect SiC/SiC CMCs and increase component life-time and durability. While both TBCs and EBCs are refractory ceramic coatings, the design of EBC systems has been focused on providing both thermal and environmental protection. The desired qualities for an EBC is therefore: (1) environmental stability and low oxygen permeability, (2) low CTE mismatch to prevent cracking via thermal stresses (good adherence), (3) high-temperature chemical and phase stability, and (4) low thermal conductivity for increased thermal protection and reduced cooling air requirements [16]. Early CMC coating development focused mainly on protecting silicon based ceramics from corrosion by molten salts. Early coating development focused on using plasma spray coatings with mullite as a major constituent because of its close CTE match with 8

25 SiC. However, a major drawback to these early mullite coatings was the formation of surface cracks that allowed corrosive species to penetrate deep into the coating to the substrate. The coatings group at NASA Glenn determined that the plasma sprayed mullite contained large volumes of amorphous mullite that formed during solidification. The crystallization of this amorphous mullite, and its associated volumetric change, was identified as the primary cause of the coating cracking. Subsequently, a modified plasma spray process was developed that eliminated the amorphous mullite from the coating [17]. However, when it was later determined that in the presence of water vapor that mullite lost silica through volatilization, there was a major focus shift to development of coatings with increased resistance to combustion environments. This shift resulting an attempt to use YSZ as an overlay of the mullite coating to prevent recession, and was truly the first generation of what would become known as EBCs [18]. Second generation coatings, with substantially improved durability, developed under the NASA HSR-EPM program replaced the YSZ top coat with a mullite + BSAS intermediate coat and a BSAS top coat. These EBCs were proved for thousands of hours at 1250 C on the SiC/SiC CMC combustor liners of three Solar Turbines Centaur 50s gas turbine engines [19]. A few years later under the NASA UEET program, research continued on the development of EBCs that had engine temperature capabilities of 1482 C (2700 F) and greater. Unfortunately the BSAS coating system showed increased spallation behavior at temperatures as low as 1300 C due to volumetric changes caused by sintering [20]. However this study did identify some rare earth monosilicates that proved to be viable candidates for EBC top coats [21]. In a later UEET study, coatings were investigated for 9

26 use of up to 1650 C. These studies identified hafnia (HfO2), pyrochlore, and magnetoplumbite oxides as the best performing top coat candidates [22-23]. Recently, next generation EBC research has continued at NASA Glenn under the Fundamental Aeronautics (FA) Program Supersonics and later the NASA Environmentally Responsible (ERA) Projects. The emphasis for these advanced EBC systems is the desire for lower thermal conductivity materials that also demonstrate increased high temperature stability. Lower thermal conductivity will allow for the use of thinner coatings to achieve large temperature reductions even at increased engine operating temperatures. Therefore, you can maintain the appropriate CMC substrate temperature while simultaneously reducing cooling requirements and weight. The most recent state-ofthe-art EBCs developed under these programs have focused around the success of HfO2/ZrO2 + rare earth silicate multilayer compositions as both a thermally and environmentally protective coating [24-26]. Figure 4 briefly describes the development of Ni-based super-alloys, CMCs and T/EBCs for use in aero turbine engines. The desire to increase engine output pushes the technology forward to achieve even higher engine component temperature capabilities. 10

27 Figure 4: A brief summarization of the design limits of high temperature aero-engine materials. The desire for increased temperature capabilities has driven the current material and coatings selection [125]. 11

28 CHAPTER II MOTIVATION As previously mentioned, because of their distinct high temperature capabilities CMCs are being considered as an attractive alternative to Ni-based super-alloy materials for use as aero-engine hot-section components. In order to further increase thermal capabilities and ensure environmental durability, the next generation CMCs will require sophisticated high-temperature ceramic coatings known as EBCs. Due to the complicated nature of ceramic coating and composite failures, a comprehensive understanding of the dominate damage mechanisms is of critical importance. Therefore, the motivation for this work is to provide insight into the damage of T/EBCs and SiC/SiC CMCs via novel thermo-mechanical testing, health monitoring and inspection, and modeling of material response. Due to the high level of difficulty and cost, at the time of this publication much of the research done on tensile damage and failure of MI SiC/SiC CMCs has been conducted under room temperature conditions. The work described herein furthers the scientific field in that it describes damage of ceramic coatings, SiC/SiC CMCs, and coated CMCs under simulated engine environments. To achieve this goal, experimental work 12

29 was performed utilizing novel laser-based high heat-flux testing apparatus as well as a high pressure burner rig (all unique to the NASA Glenn Research Center) to more accurately simulate thermal and mechanical loading and the extreme environments that these materials will be subjected to in service. This work goes beyond typical room temperature characterization of mechanical properties and attempts to address the critical life limiting issues to SiC/SiC CMCs and their requisite EBCs that need to be addressed before they can be considered for service in turbine engines. The experimental work presented here has been constructed to simulate actual engine environment by introducing high temperature thermal gradients, environmental exposure, thermal cyclic and time dependent damage mechanism otherwise unaddressed by current research. A summary of the life-limiting damage behavior investigated in the following chapters is presented below in Tables 1 and 2. 13

30 Table 1: Overview of critical ceramic matrix composite damage mechanisms addressed. LIFE LIMITING DAMAGE BEHAVIOR OF SIC/SIC CERAMIC MATRIX COMPOSITES MATRIX CRACKING - Matrix cracking decreases load carrying capabilities and open crack allow for environmental access to the interior of the composite leading to oxidation degradation. Macro-cracking Large scale crack development from excessive overstressing of components. Causes: High mechanical loading, from simple tensile loading and complex multiaxial loading conditions. Excessive thermal stresses due to extreme thermal gradients and/or thermal shock. Because these materials are intended for use at extremely high temperatures, it is likely that some portion of the component is susceptible to cracking from creep related phenomena. Micro-cracking Small scale cracking can for in the vicinity of stress concentrations even at low applied loads. Causes: Unlike metals that can tolerate local overloads via local deformation to accommodate and redistribute stress, ceramics relieve stress in the form of cracking. ENVIRONMENTAL DEGRADATION Level of life limiting degradation dependent on composite composition, temperature and oxidation products. Dry Air Undamaged composite: The interior of an undamaged composite is sealed off from the outside environment, leaving it susceptible only to broad surface recession (generally slow, but increased at intermediate temperatures). Damaged composite: Matrix cracks (even micro-cracks) allow of ingress of oxidizing environment to interior of composite, resulting in possible degradation of matrix, fibers, and interphase. Wet Air Rapid increase in recession of SiC through volatilization of normally protective SiO 2 layer. CHAPTER 5, , 6 5,

31 Table 2: List of thermal and environmental barrier coating damage mechanisms affecting performance and life. LIFE LIMITING DAMAGE BEHAVIOR OF PROTECTIVE CERAMIC COATINGS COATING CRACKING Coating cracking allows for oxidizing environments and excessive temperatures to reach underlying component. Causes: Applied mechanical loading above the strength of the coating and/or 6, 7 CMC matrix crack propagation through bonded interface. Cracks generating by high temperature coating sintering and creep. 7 Transient thermal stresses arising from large thermal gradients and 7 CTE mismatch between coating layers and between coating and substrate. Internal stresses caused by TGO growth and buckling. 7 HIGH TEMPERATURE CORROSION Degradation of coating effectiveness from ingestion of foreign substances into high temperature engine environment. Causes: Infiltration of the coating by molten substance (CMAS) that leads to damage from thermomechanical and thermochemical processes CHAPTER 7 The experimental work detailed in the following chapters also introduces some very novel NDE techniques including the use of high temperature digital image correlation and electrical resistance monitoring that to date has not been demonstrated by previous researchers. High temperature acoustic emission is also incorporated as a monitoring technique, and analysis is presented for damage assessment and comparison to other NDE results. An emphasis is placed on extending the use of the emerging technique of electrical resistance measurements as an effective monitoring and inspection tool for SiC/SiC CMCs. However, as electrical resistance is dependent on thermal and mechanical loading as well as time/temperature dependent effects, a systematic approach was employed to describe the contribution of each. This investigation into the use of electrical resistance measurements is once again pioneering 15

32 since the bulk of research on this topic up to this point has focused on room temperature testing. Therefore, this study provides an in-depth examination of the capabilities of electrical resistance measurements for use under high temperature testing conditions. Figure 5 outlines the methodologies that have been employed in testing and their relationship to characterization of critical ceramic matrix composite and protective coating life-limiting damage behaviors. Damage Characterization NDE Approach Acoustic Emission (AE) Digital Image Correlation (DIC) Electrical Resistance (ER) Measurements Monitoring Inspection AE signals associated with elastic stress waves from fracture events Tracks local displacements of specimen surface Dependent on strain (elongation, matrix cracking, associated fiber sliding) Residual effect of microstructural changes and damage Coating Cracking Matrix Crack Accumulation Creep related damage and time-dependent microstructural changes Differences in damage behavior due to Environmental Degradation Figure 5: Brief outline of basic non-destructive evaluation (NDE) approaches used to address critical life-limiting behavior. 16

33 CHAPTER III LITERATURE REVIEW 3.1 Micromechanics of Tensile Damage of Ceramic Matrix Composites (CMCs) Damage mechanics can be summarized as the study of how damage affects the overall strain response of a material. For homogeneous materials on a structural scale, often a macro or continuum mechanics approach is taken for deformation analysis. However, in the case of composite materials, a micromechanics methodology is often adopted because it describes the material on the scale of damage. Therefore, the intent of this review is to describe the damage mechanisms associated with the redistribution of stress and accumulation of strain due to tensile loading of brittle composites. The basic micromechanical principles of brittle composite fracture were first described in the seminal work of Aveston, Cooper and Kelly (ACK) in 1971 [27]. In general, this paper refers to analysis of fracture in a two-phase fiber-reinforced composite systems (i.e. a matrix material reinforced with continuous fibers in the loading direction). It is worth noting that in practice there exists an interphase material that separates the fiber and matrix, however while this fiber coating material is often not explicitly mentioned the engineered properties of this interphase are generally just 17

34 referred to as fiber/matrix interfacial properties. The authors begin by describing the difference between single and multiple fracture of composites. The paper defines single fracture of a composite as failure of a composite due to the failure of a single constituent and the inability of the remaining constituent to carry the load. Such is the case for high strength, high modulus fibers in a weaker matrix where the failure strain of the matrix far exceeds that of the fibers. This behavior is commonly seen in materials like carbon fiber reinforced polymer matrix composites. However, when the non-broken constitute is in fact able to support the load upon failure of the other, a tensile specimen will experience multiple fracture of the more brittle phase until the ultimate strength of the stronger phase is exceeded. The case of multiple fracture is observed in ceramic matrix composites where periodic fracture of the matrix occurs when the fibers have a higher failure strain and the fiber volume fraction is above a critical level. Multiple fracture of the matrix is beneficial to the design of CMCs because the ability to share the load between the matrix and fiber is the mechanism that results in the increased toughness of the composite. In terms of composites this phenomenon is often referred to as transverse cracking or matrix microcracking. In the elastic region an undamaged composite is considered as a continuum which consists of bonded fibers and matrix that extend under an iso-strain condition (i.e. constituent strain is equal to the overall composite strain), where the composite stress σc can be determined by a force balance the constituents to develop a typical rule of mixtures relationship: 18

35 σ c = σ m V m + σ f V f (1) According to the ACK model, assuming the fibers have a higher strain to failure than the matrix, multiple fracture of the matrix will occur when: σ fu V f > σ mu V m + σ f V f (2) where σfu and σmu are the ultimate strengths of the fiber and matrix, and Vf and Vm are the respective fiber volume fractions of the fiber and matrix. The term σf represents the stress on the fibers that is required to produce a breaking strain on the matrix. As previously mentioned, most CMCs exhibit multiple matrix cracking where the matrix crack is bridged by the reinforcing fibers. That is, at low tensile stresses microcracks initiate in the matrix and extend across the material. Due to the ability of the fiber and matrix to debond, the matrix cracks do not penetrate the fibers. Subsequently, these bridging fibers carry all of the applied load across a matrix crack and limit the crack opening. ACK suggests that the load carried by these bridging fibers is transmitted back into the matrix over a distance x on each side of the matrix crack by a constant interfacial shear stress τ between the fiber and matrix. By using a global force balance, a value of the load transfer length x can therefore be determined in terms of the interfacial shear stress and the load on the fibers. ACK identified that the basic mechanisms leading to inelastic strains in CMCs are matrix cracks and fiber/matrix interfaces that debond and slide. 19

36 A number of micromechanics based damage models have been developed around the concept of a global energy balance as described in ACK. Most models are also similarly based on the assumption that load is transferred from debonded fiber and matrix via a constant τ. Furthermore, a common simplification in many micromechanics models is the use a single fiber approximation in which a multifilament composite is modeled as a single fiber inside a matrix at the same volume fractions as a multi-fiber material. An influential model developed by Pryce and Smith [28] uses some basic micromechanical assumptions to model the stress-strain behavior of a cracked laminate composite under monotonic and unload/reload testing. Other than the inclusion of unloading and reload, Pryce and Smith also consider the effect of thermal residual stresses on fibers. In the plane of a matrix crack, obviously the matrix can carry no load and therefore all of the stress is on the bridging fiber. This load is transferred between the fiber and matrix resulting in a length dependent axial fiber distribution stress which can be written as: σ f = σ c 2τx V f r (3) where r is the fiber radius and x is the distance from the matrix cracking plane. Moving away from the matrix crack, the stress on the fiber decreases by the interfacial shear stress τ until the load transfer distance x is reached. At this point no more sliding between the fiber and matrix occurs and the fiber and matrix reach strain compatibly. Figure 6 shows the stress profile on the fiber and matrix on one side of a matrix crack. 20

37 Figure 6: Stress profile in the fibers (solid line) and the matrix (dotted line) for a cracked unidirectional composite at applied axial stress σc [28]. By including the effects of thermal residual stresses, the stress on the fiber at this point becomes: σ f = σ c E f E c + σ f T (4) where σ f T is the thermal stress in the fiber and Ef and Ec are the modulus of the fiber and the undamaged composite which can be related by a rule of mixtures approximation as: E c = E m V m + E f V f (5) where Em is the matrix modulus. Equation (3) and (4) lead to an expression for x in terms of fiber radius r, elastic properties, and applied and residual stresses: x = r 2τ [σ c V m E m V f E c σ f T ] (6) Once the load transfer length has been determined the authors were able to develop an expression for the mean strain on the fibers (same as mean composite strain) in the debonded region based on the average fiber stress and uniform crack spacing s. The 21

38 total mean composite strain is then considered as the sum of the mean strain in the debonded region plus the strain in the composite in the bonded strain compatible region. Referring to Figure 6 these are the regions AB and BC respectively. Notably, the equation that was derived for average composite strain did not include the contribution of the crack bridging fibers, as it was determined that this was negligible due to small scale of crack opening displacements. The paper goes on to describe the state of stress in the fiber and matrix upon unloading the composite from an axial stress σc to some intermediate stress of σ. As the composite is being unloaded, a portion of the matrix will begin to slip relative to the fiber. Once again this slippage is controlled by the interfacial shear stress opposing the slipping direction of the fiber. This results in a length of reverse slip y as the composite is unloaded to the intermediate axial stress σ. The length dependent stress on the fiber is shown by the dotted line in Figure 7a. The Pryce & Smith model assumes that upon full unload of the fiber the matrix will have slipped a total distance of (x + x T )/2, resulting in the residual stress state on the fibers shown be the solid line in Figure 7a. Because a portion of the strain is still sustained by the interfacial shear stress, x remains constant upon subsequent unload and reload steps (assuming the original peak stress σc is not exceeded. If the composite is then reloaded from this residual stress state to some stress σ, the slippage distance will once again vary with length according to the value of τ opposing the relative movement of the fiber. This results in a fiber stress profile represented by the dotted line in Figure 7b. The equation for the average composite strain can once again be calculating by defining the mean fiber strain based on the fiber 22

39 stress profile. The stress-strain behavior of the composite can therefore be determined by developing the fiber stress profiles for given points in the unload/reload history. (a) (b) Figure 7: Stress profile in the fibers for a cracked composite, during (a) unloading from an applied axial stress σc to zero; (b) reloading from zero back to σc [28]. While Pryce and Smith are able to describe the fundamentals of micromechanics with the basic principles of load sharing described in ACK, a very influential article by Hutchinson and Jensen (H&J) [29], later expanded upon by Marshall [30] and 23

40 Parthasarathy et al [31], proposed a more sophisticated model that provides a closer examination of the role of interfacial properties on overall mechanical behavior. Hutchinson & Jensen consider an axisymmetric cylindrical model of a fiber surrounded by matrix having a residual compressive stress at the fiber/matrix interface. A schematic of the single fiber model is shown in Figure 8a. Note that the nomenclature used in the cylindrical model uses σf + and σf to refer to the axial fiber stress above and below the debond crack tip respectively. The solution for the stresses and strains above the fiber/matrix debond region are based on the widely used solution for this classical axisymmetric Lamé problem that was first developed around The Lamé solution is valid if the axial stress vary slowly over the debond length (i.e. τ is small compared to the stress on the fiber σf ). The paper assumes two boundary conditions for the external surface of the outer cylinder: (1) there are no normal or shear forces acting on the surface, and (2) there are no shear forces but the outer cylinder is constrained in the radial direction. These boundary conditions represent the cases of a single fiber surrounded by matrix (similar to a pushin or pull-out specimen) and a multifilament composite where multiple bridging fibers are pulled evenly. One major consideration of the H&J model that separates it from the simpler models (like that of Pryce and Smith) is that the model has been extended to include the Mode II fracture energy Gc at the tip of the debond crack. The influence of the crack tip debond energy was first described by Marshal et al [32], when modeling fiber sliding during push-in tests. For debonding and relative fiber/matrix sliding to occur (rather 24

41 than brittle fracture) the debond energy Gc must be less than an upper limit dictated by the fiber fracture energy. Because most composites are designed to have a very weak interface to facilitate fiber/matrix debonding, often a small debond energy (SDE) assumption is made in which the debond energy is considered very small or in the case of some models negligible. However, a significantly large debond energy (LDE) results in a drastic jump in stress at the crack tip. Depending on the magnitude of the debond energy, neglecting it (e.g. models like Pryce and Smith) can lead to significant over estimations of strain accumulation during unload reload tests. This is because when the debond energy is large, reverse sliding reaches the end of the debond length at some point upon unloading. A parameter that specifies whether or not the debond energy is small or large with respect to its effect on mechanic response during loading and unload has been identified [29, 30]. Therefore, in cases where debond energy is not negligible the mechanical properties of fiber/matrix interface are characterized by both fracture (debond energy) and sliding (interfacial shear stress). The H&J model continues to investigate the contribution of interfacial properties by comparing two types of sliding resistance: (1) the previous constant τ assumption (i.e. constant friction), and (2) a Coulomb friction case in which the frictional stress is proportional to the normal stress σr (which is negative) on the fiber/matrix interface as: τ = μσ r (7) where μ is a constant coefficient of friction. The Coulomb friction case in interesting because it considers the state of radial stress in the fiber. In a bonded composite the radial stress at the interface is generated by the thermal misfit strains between the 25

42 constituents that are produced upon cooling from the processing temperature. However, in the debonded region the radial stress is a function of the applied axial stress on the fibers by Poisson effects. Since the axial fiber stress along the debond is length dependent, it follows that the radial stress and therefore the frictional stress are is well. In the case of rough fibers, Eq. 7 can be updated to include the addition of a constant term τ0 that accounts for increase in interfacial shear stress due to fiber surface roughness. The axial stress profiles on the fiber using the axisymmetric cylindrical model including debond energy and Coulomb friction for both the unloading and reloading cases are shown in Figure 8b and 8c. Note the differences between these cases and the more simplified cases shown in Figure 7a and 7b that neglect Gc and consider a constant τ (i.e. effect of length dependent radial stresses at the interface is neglected). (a) 26

43 Unloading Reloading (b) (c) Figure 8: (a) Composite cylinder model used in analysis, along with axial fiber stress at different stages of (b) unloading from a peak stress σp, and (c) reloading when reverse slip reaching the end of the initial debond before full unloading [30]. Note that the unloading shown in Figure 8a is for various stages of unloading, where σ3<σa<σp and γ the difference in axial fiber stress behind and ahead of the debond crack tip. The relation for γ is approximated by H&J as a function of Gc, elastic properties, and fiber radius and volume fraction. As before, the composite strain can be calculated by determining the mean stress on the fiber for a given axial loading history. Matrix cracking and fiber breakage, like all ceramics, are based upon the statistical distribution of pre-existing flaws within the material. As previously discussed the mechanical decoupling of the fiber and matrix leads to an increase in mechanical toughness (reduction in brittle monolithic ceramic behavior). However, the overall accumulation of inelastic strain is still based on the flaws in the matrix and fibers. Therefore, research has been performed to include the statistical fracture of fibers and matrix under uniaxial tensile loading to the established micromechanics model [33-37]. Because the ultimate strength of a CMC is determined by the strength of the fibers, 27

44 work was first conducted on single filament composite (s.f.c) tests to describe the random distribution of fiber strengths [33, 34]. In these tests an s.f.c. is loaded in tension and experiences multiple fiber breaks as the stress is increased. Though the fiber sustains multiple breaks, the composite is held together by a ductile matrix material. It was shown that the distribution of fiber strength is of the form of a Weibull distribution of fiber breakages versus stress. This distribution of fiber strength can then be used in modeling the overall behavior of materials with that reinforcing fiber type. Further work was done to include multiple stress-dependent matrix cracks by presenting a similar consideration of matrix flaw distribution [35-37]. Unlike the previously described micromechanics models that assume an average crack spacing that occurs at a defined matrix cracking stress, these models determine matrix crack evolution via statistical methods. In general, these models are based upon a minimum matrix cracking stress and a reference stress for matrix cracking centered upon a Weibull strength distribution [35, 36]. One major consideration of the stochastic multiple matrix crack model is that of matrix crack interactions. That is, for closely spaced cracks it is possible to encounter the case of crack interaction where debonds interact with each other and the assumption of a crack spacing greater than twice the slip length is no longer valid. The maximum stress on the fiber still decreases via interfacial shear stress as you move away from the crack face, but reaches the debonded area of an adjacent crack before returning to the far-field fiber stress value. Figure 9a illustrates the case of crack spacing resulting in isolated cracks and multiple crack interaction. Continued increase in matrix crack density results in the limiting case 28

45 where the applied load reaches the matrix crack saturation stress (i.e. the matrix can no longer sustain a high enough stress to initiate further cracking). The common assumption amongst these models if that at matrix crack saturation, the slip length is limited by one-half the crack spacing. (a) (b) Figure 9: (a) the stress around an isolated crack returning to the far-field stress over a slip length δ (by τ and stress jump via debond energy), and closely spaced cracks where the slip is limited to one half crack spacing; (b) slip regions (shaded) around multiple fiber breaks (X) in a composite. These fibers carry a pull out stress as they slip around the central cracking plane [37]. The work of Curtin et al [37] goes on to model the stress-strain behavior combining both the effects of matrix cracking and fiber breakage. The case of multiple matrix cracks centered about a single central crack plane is shown in Figure 9b. The combination of 29

46 both deformation mechanisms helps to explain the transition from brittle to tough behavior seen experimentally in CMC systems. Fiber breaks that occur within a slip length of the central plane still carry a reduced load. The pullout of these broken fibers are still resisted by the interfacial shear stress at the fiber/matrix interface, and it is shown that because of this the load carried by these broken fibers at the central plane is significant. Clearly this complicates the calculation of composite strain as it requires consideration of the elastic strain of the unbroken fibers as well as the pullout of the broken fibers. This model developed by Curtin et al is applied to the experimental data of tensile behavior of SiC/SiC minicomposites from Lissart and Lamon [36] with good agreement. However, application of the model does require previous data on single fiber strength distribution and average crack spacing (provided by Lissart and Lamon) along with the knowledge of fiber and matrix elastic and interfacial properties also required by previous models. All of the micromechanics based-models described in this review clearly show the importance that interfacial properties play on the mechanical behavior of CMCs. Interfacial behavior is dominated by the fiber coating properties and fiber morphology (e.g. fiber roughness) and unfortunately cannot not be known precisely a priori. In fact, because they are all unknown parameters that affect inelastic behavior there is no unique solution of thermal residual stress, Gc and τ for modeling CMC strain response during monotonic tensile loading. Therefore, approximations for values of these properties are often determined experimentally using a variety of methods. One of the most cited experimental methodologies for characterizing the inelastic strain of CMCs 30

47 was presented by Vagaggini et al (1995) [38]. In general, the procedures outlines in this methodology show how the properties of the interface and thermal residual stresses relate to the hysteresis and residual strain behavior of CMCs during unload/reload tests. The hysteresis tests presented in this work have several advantages over previously utilized push-in/pull-out and minicomposite testing. One major reason is that, because of their size and number of fibers, testing on a macrocomposite (as opposed to an s.f.c. or a minicomposite) better accounts for fiber-to-fiber variations in τ and Gc, as well as fiber misalignment, etc. Furthermore, these procedures are often far less complicated and easier to interpret than singe fiber testing. The test method suggested by Vagagginni et al requires that a tensile specimen be loaded in tension above the matrix cracking stress, but below matrix crack saturation. Any number of unload/reload (to slightly higher stress) are performed and each hysteresis loop is analyzed to evaluate constituent properties. The equations established in H&J for the composite strain behavior are used to determine relationships between inelastic strains and constituent properties for a give set of mechanical data. In general matrix crack density is required for determination of physical parameters, however the authors also propose a procedure for estimating the matrix cracking stress based on the decrease in stiffness on successive loading cycles. Vagagginni et al present all of the basic formula for evaluating τ, Gc and thermal residual stress from hysteresis measurements [38] and demonstrated the accuracy of their methodology on experimental data from unidirectional SiC/CAS and SiC/SiC composites [39] with 31

48 excellent agreement. The methodology proved to be exceptionally straightforward when determining the properties of LDE composites like the SiC/SiC sample tested. 3.2 Damage monitoring and Non-Destructive Evaluation of SiC/SiC CMCs While many techniques have been developed for non-destructive evaluation (NDE) and health monitoring of composite systems this review will focus mainly on currently utilized techniques for assessing damage in SiC/SiC CMCs. These techniques include modal acoustic emission (AE) monitoring, electrical resistance (ER) measurements, and digital image correlation (DIC) Acoustic Emission (AE) Acoustic Emission (AE) refers to a monitoring technique that operates by obtaining acoustic signals generated from a localized source within a material. Transient acoustic signals are generated when the fracture energy of solids, associated with material damage, is rapidly released in the form elastic stress waves that can be detected by the use of piezoelectric AE sensors. AE monitoring techniques can therefore be used to detect the sound emitted during various forms of plastic deformation and give information regarding the initiation and accumulation of material damage. Traditional AE (TAE) monitoring consisted of processing AE event data using narrow band or resonant frequency AE transducers that filter the AE waveform into a damped sine wave. This was done to due to the limitations of early sensors as well as the limited data processing and analysis capabilities of early computer systems. The TAE method only captures certain aspects of the acoustic signal (often referred to as AE parameters or features) which include number of AE counts, peak amplitude, energy, etc. [40]. 32

49 The rapid advancement in computing and sensing technologies has enabled modern AE systems to capture and analyze the full AE waveform and frequency spectrum. Therefore, it is now possible via the use of wide band AE sensors and modern AE acquisition and analysis to characterize the nature of AE sources from the waveform captured. The digital capture of full frequency spectrum AE waveform has come to be known as Modal AE (MAE) [41, 42]. Typical materials testing coupons can be considered as thin plates (i.e. the thickness is much smaller than the other two dimensions), and the thickness direction is generally smaller than the wavelength of a traveling acoustic wave. In a thin plate type specimen, the waves are known as plate waves and are dominated by two modes of propagation. These modes are referred to as extensional and flexural modes [43]. Extensional waves are always of higher frequency and velocity, and are essentially nondispersive (velocity not very sensitive to frequency) in CMC materials. Flexural waves on the other hand are of lower frequency and velocity and can be highly dispersive [44]. By digitizing the entire waveform MAE is able to distinguish the extension and flexural modes from the reflections of the waves produced by the boundary conditions. Event-based Modal acoustic emission testing has been used to explore the stress dependent damage accumulation of SiC/SiC CMCs under mechanical loading. Eventbased AE acquisition collects the waves from each of the AE sensors used simultaneously. This allows for synchronization between the testing channels for every damage event. By comparing the times of arrival of the extensional wave mode from different sensors it is often possible to determine the location and velocity of each AE 33

50 event. Once the locations of the AE events that occur during testing are determined, it is possible to discard any events that did not initiate from a designated area of interest. This results in a more accurate correlation between AE events and localized damage. Early work by Morscher et al [45-47] was successfully able to use MAE to monitor and locate damage accumulation during testing of SiC/SiC CMCs. These studies showed that a dramatic increase in accumulated AE energy correlates well with the nonlinearity (caused by decreasing stiffness) seen in experimental uniaxial tensile test data. This decrease in stiffness results from an increase in transverse matrix cracking with increasing load [45, 46]. In fact, Morscher [47] was able to show that the accumulated AE energy is directly proportional to the increase in stress-dependent transverse matrix cracks obtained via microscopy studies. The natural of cumulative AE energy was also used to identify matrix microcracks initiation (found to be at a stress well below the proportional limit seen in the mechanical data) and matrix crack saturation. Recent studies have attempted to further analyze MAE waveform characteristics in order to identify the specific damage modes in CMCs under mechanical loading. The principle is based on the assumption that different damage modes generate specific features that can be determined via waveform analysis techniques. Clustering techniques have been applied to AE signals generated during tensile tests of SiC/SiC [48], and SiCf/[Si-B-C] composites [49, 50]. One key issue in this type of analysis is the fact that composite systems are dispersive by nature and therefore, many AE signal properties are greatly affected by how the wave propagates. Therefore, it can be quite difficult to compare results from different composite systems because many studies fail 34

51 to account for the effects of wave attenuation on the propagation of the AE signal. To illustrate effects of signal attenuation, research was performed that utilizes the energy recorded from AE sources during testing. By monitoring the change in attenuation coefficient with increased inelastic strain, the researchers were able to evaluate energy attenuation in real-time [51, 52]. Additional work by Maillet et al uses this energy-based approach to account for attenuation and investigates the various limitations associated with using MAE for damage mode identification of SiC/SiC CMCs [53] Electrical Resistance (ER) Due to its inherent self-sensing capabilities and ease of experimental implementation, electrical resistance (ER) measurements have been utilized for monitoring damage and strain accumulation in a number of material systems containing electrical conductive constituents. Several polymer matrix composites have been investigated using ER monitoring, in particular carbon fiber reinforced polymer composites (CFRP). In the case of CFRP where the fibers are conductive, but the matrix is insulating, change in electrical resistance during mechanical loading is dependent on fiber behavior and fracture. ER has been used to monitor the increase in material resistance caused by fiber elongation and breakage during quasi-static and fatigue loading [54, 55]. The electrical response of SiC/SiC composites is quite different from CFRPs because both the fibers and matrix are semiconductor materials that are capable of carrying electrical current. Therefore, the ER change during mechanical testing of SiC/SiC could be controlled by damage to both the matrix and the fibers. However, the conductivity of 35

52 the matrix material is highly dependent upon the processing technique utilized to create the composite. For instance, for CVI SiC/SiC composites, the electrical resistivity of the CVI matrix may be on the same order of magnitude as that of the fibers. However, due to the significant presence of free silicon in the matrix, melt-infiltrated SiC/SiC can be orders of magnitude more conductive than CVI SiC/SiC systems. Recent works have successfully been able to correlate increases in ER with stress-dependent matrix cracking and damage accumulation in room-temperature tensile testing of CVI [56] and MI-CVI [57] woven SiC/SiC CMCs. Preliminary efforts by Morscher et al [57] have developed a simple resistor circuit model consisting of matrix crack segments and bridging fiber regions that carry the current across the crack. The matrix crack segments are assumed to be of uniform length and are represented by a parallel circuit of resistors that attempt to account for the current carrying capabilities of the fiber, silicon in the matrix and resistance at the fiber/matrix interface. That is, at a given damage state the model consists of resistors that represent: (1) undamaged segments composite, (2) a region of fiber/matrix debond, and (3) crack bridging fibers. Furthermore the model relies on an assumed decouple length parameter in order to account for the electrical behavior at the interface at high stresses. However, it was seen that the model vastly under predicts the electrical behavior at high stresses and at failure. A recent expansion on this model by Baker [58] attempts to describe the unload/reload hysteresis behavior of MI-CVI SiC/SiC CMCs. A schematic representing the work by Baker is shown in Figure 10. Similar to the model by Morscher, Baker uses a simple resistor circuit to model the regions of undamaged composite (modeled as the 36

53 contribution to resistance of the fiber Rf x and matrix Rm x ), unbonded fiber/matrix interface Rc, and crack bridging fibers Rf u. Figure 10: Schematic representation of the electrical model of a unit segment between two transverse matrix cracks [58]. However, unlike Morscher et al, this work takes a more sophisticated micromechanics approach in order to determine the stress state on the interface. Also this model attempts to account for the contribution of wear on the fiber/matrix interface by accumulated relative fiber/matrix sliding and its relation to the overall change in electrical response via Rc. While this model is capable of showing the general trends associated with the electromechanical response of the experimental data shown, it still relies on many empirically based parameters and lacks the fidelity to truly identify the contributions of stress and damage dependence. The above described works have been successful in determining a correlation between increasing ER and room temperature damage. However, because CMCs are intended for use at high temperatures, investigation of electrical response of CMCs under high temperature thermomechanical loading conditions is required. 37

54 3.2.3 Digital Image Correlation (DIC) Digital image correlation is optical measurement technique based on image analysis and numerical computing. DIC relies on analysis of some sort of surface contrast (usually produced by a speckle pattern on the surface of the test coupon) in order to track displacements within the area of interest of a set of images. Although it is sometimes referred to by different names (computer aided speckle interferometry CASI, digital speckle correlation method DSCM, etc.) the principles behind DIC were developed in the early 1980s by a group of researchers at the University of South Carolina [59]. The basic working principle of DIC is the tracking of specific points (or pixels) between two or more images recorded before and after deformation. With DIC it is possible to generate a full-field displacement or strain mapping of a test coupon by comparing images of a test specimen surface in the unloaded (reference) state to images taken of the loaded (deformed) state. DIC has been utilized and optimized for thermal and mechanical testing extensively over the last several years. The fact that this technique is non-contact and provides localized strain mapping, makes DIC especially attractive for use in CMC testing were damage and failure are controlled by localized stress fields. However, some difficulties have recently been explored when using DIC on low failure strain materials and at elevated temperatures. One case study demonstrates the capability of using DIC for strain mapping of a woven SiC/SiC composite under tensile loading, but illustrates the complications associated with retaining high-fidelity results post matrix crack initiation [60]. Further work shows this increased difficulty of using imaging system for high temperature CMC testing [61]. 38

55 Imaging of a C/SiC composite was demonstrated up to 1500 C using a specialized laserheating setup. Acquisition of usable images required the use of high temperature oxide paints for speckling as well custom optics, filter and lighting schemes. Clearly, DIC can be a useful tool for use in thermomechanical testing and characterization of CMCs, but many obstacles still exist in the experimental applications. 3.3 Thermal Cyclic Damage of Ceramic Coatings (T/EBCs) During regular thermal cyclic engine operation, several factors lead to the cracking and ultimate spallation failure of ceramic coatings. The specific mechanisms that lead to microcracking and eventual failure are dependent upon coating composition and environmental/loading conditions. While several coating microstructures can be produced through the use of different deposition techniques, this review will focus on two common deposition techniques: (1) Air Plasma Spray (APS), and (2) Electron Beam- Physical Vapor Deposition (EB-PVD). T/EBCs produced by APS and EB-PVD are so different in their: microstructure, morphology and material properties that different failure mechanisms are dominant. The EB-PVD is a process in which the chosen coating material is evaporated from an ingot and the vapor is directed onto a preheated substrate. EB-PVD creates a columnar structure that simultaneously provides strain tolerance and decreased thermal conductivity. APS is generally a lower cost alternative to EB-PVD and results in inter-splat porosity that provides less significant strain tolerance and thermal conductivity reduction [62, 63]. Scanning electron microscope (SEM) micrographs illustrating some of the major microstructural aspects of APS and EB- PVD topcoats are shown in Figure 11 (a,b) and Figure 11 (c,d) respectively. 39

56 Figure 11: SEM micrographs showing the microstructural and defect components of (a,b) APS and (c,d) EB-PVD TBC topcoats [63]. The durability of coatings is governed by a sequence of crack nucleation, propagation and coalescence events that accumulate prior to final failure by large scale buckling and spalling. The details of the damage mechanics amongst coating systems is slightly different, however the growth of the thermally grown oxide (TGO) layer is often the most critical phenomenon responsible for coating failure under thermal cycling[9, 62, 63]. The TGO is a thin oxide layer that forms between the bond coat and ceramic top coat via bond coat oxidation during high temperature engine cycles. For example, the stresses that drive cracking mechanisms in APS coatings can be summarized by the behavior of the TGO layer. It has been shown that two types of cracking modes dominate coating failure: (1) separation of bond coat/tgo interface at bond coat undulation peaks, and (2) cracking within the top coat layer associated with troughs of TGO/top coat interface [64, 65]. First, undulations created by the bond coat surface 40

57 roughness result in tensile stresses at the peaks of the bond coat/tgo interface and compressive stresses at the troughs. Note that for APS coatings these undulations can be significant due to deliberate roughening of the bond coat surface prior to top coat deposition in order to improve adhesion [65]. During repeated thermal cycling the TGO with grow in thickness due to high temperature oxidation causing an increase in the tensile stresses at the bond coat/tgo interface resulting in crack initiation at the peaks. Secondly, coefficient of thermal expansion (CTE) mismatch at the interface between the top coat and TGO results in out of plane tension in the topcoat above the undulation peaks and compression in the troughs. The resulting tension creates cracking at the TGO/top coat interface as well as radial cracking within the top coat in the vicinity of the peaks. If the TGO grows thick enough, the CTE mismatch of the bond coat/tgo becomes lower than the top coat and bond coat which switches the stress at the top coat undulation troughs from compression to tension resulting in crack propagation in the valleys between peaks. Due to the increased strain tolerance resulting from the columnar structure of the coating created by the deposition process, the cracking behavior under thermal cyclic load in EB-PVD topcoats is driven by slightly different mechanisms. In fact, the damage of these coatings is quite complex and still not entirely understood. Therefore this review will include only a cursory explanation of the crack driving mechanisms. Separation at the bond coat/tgo interface is similar to that described in APS coatings, however the undulation peaks in this case are caused by the inherent imperfections of the deposited bond coat layer and are considerably less severe that those created 41

58 during surface preparation of APS coatings. A second mechanisms seen in EB-PVD coatings is cracking at the TGO/top coat interface and increasing penetration of the TGO into the bond coat resulting in increased roughness. This behavior is thought to be caused by bond coat creep [66] and plasticity [67]. This increase in roughness/undulations associate with EB-PVD failures is often referred to as ratcheting [68], rumpling and/or buckling of the TGO/bond coat (nomenclature used typically depends on exactly how the mechanism is described). The cracks generated by this mechanism eventually coalesce into macro-cracks resulting in large scale spallation. For EB-PVD coatings with very smooth interfaces, the increased compression in the TGO with successive thermal cycling can result in large scale buckling of the coating as well. Figure 12: A schematic showing the general mechanisms driving thermal cyclic damage of: (a) APS and (b) EB-PVD ceramic coatings [9]. Fracture of both APS and EB-PVD ceramic coatings can also be accelerated by the effects of sintering. At elevated temperatures sintering of the top coat can lead to volumetric 42

59 changes and closures of coating micro-porosities. However, this often leads to increased coating cracking by making the top coat far less strain tolerant. Furthermore, wedgeshaped surface cracking generated by this sintering can result in an increase in temperature at the bond coat and substrate. These elevated temperatures lead to increases in bond coat oxidation (TGO growth) and creep [69, 70]. 3.4 Influence of high heat-flux on coating damage As mentioned above, thermal and environmental barrier coatings are applied to components that typically experience active back side air-cooling during engine operation. This results in large steady-state thermal gradients being generated across the coating and substrate materials. Furthermore, during engine starts/shutdowns, the coated surfaces heat/cool very rapidly resulting in additional transient thermal gradients. As detailed in the above section, the thermal cyclic damage of T/EBCs is driven by mechanisms such as TGO growth, strains generated by CTE mismatch upon cooling, bond coat rumpling and top coat sintering. Many of this mechanisms can be affected by increased heat fluxes and corresponding thermal gradients. It therefore becomes important to consider the effect that this type of thermal loading will have on the acceleration of coating damage initiation and accumulation. Experimental testing of coatings under high thermal gradients and heat fluxes presents some unique challenges. One approach is to utilize a flame rig in which heat is applied to surface of a coating by a high temperature burner and applied forced cooling air to the back of the sample. An alternative is to use a high heat-flux laser rig such as the one developed at the NASA Glenn Research Center for testing coating durability and 43

60 thermal conductivity [69-72]. Initial studies were among the first to characterize high temperature laser sintering and creep behavior of T/EBCs. Also, one novel aspect of this laser technique is the ability to directly determine the in-situ thermal conductivity of the test specimen. The evolution of the thermal conductivity can therefore be used to deduce the nature of sintering and delamination behavior under high heat-fluxes. This rig was also used in one study to determine the possible difference in coating damage accumulation between isothermal (furnace) and high heat-flux thermal cyclic testing [73]. In their study, Eldridge et al extended an upconversion luminescence imaging technique to be used in monitoring delamination progression in coatings under thermal gradient conditions. For this test two heat fluxes of 95 and 125 W/mc 2 were chosen to create surface temperatures of 1290 C and 1345 C respectively, and bond coat interface temperatures of 1140 C and 1175 C respectively. These samples were compared to an isothermal furnace cyclic test conducted at 1163 C. At various intervals in the testing process, luminescence images where captured from the samples tested under heat fluxes and compared to those taken under furnace conditions. It was shown that the high heat-flux (125 W/mc 2 ) tested sample exhibited pronounced differences in delamination behavior to both the lower heat-flux test and furnace tested sample. For example, the high heat-flux sample experienced accelerated vertical cracking due to increased sintering at the high temperature (1345 C) coating surface. Furthermore, it was determined that heat fluxes can also promote high degrees of bond coat rumpling thereby increasing the driving mechanism of delamination progression. 44

61 Various modeling techniques have been also been utilized to describe the effect of thermal gradients on coating damage mechanics. A highly cited paper by Evans and Hutchinson [74] examines two separate thermal loading conditions often observed in aero-turbine engines that generate thermal gradients within coatings. They include: (1) the steady-state gradient that is generated via heating of the coating surface by the combustion environment and internal cooling of the underlying substrate, and (2) the transient gradient developed by cooling of the component during engine shutdown. After a detailed stress analysis of each case, energy release rates and parameters to determine mode mixity for delamination cracking within the coating are presented. Based upon this rather sophisticated mechanical model, the authors present delamination maps for coatings based on the severity of the thermal gradient and coating thickness. A threshold (allowable) thermal gradient to prevent delamination can then be calculated for coating design purposes. A small number of similar laser-based high heat-flux rig have been developed by other groups to investigate thermal fracture of ceramic coatings. For example, a study by Choules et al [75] investigated the damage morphology of coatings of various thicknesses subjected to a single, transient (5 sec.) high heat-flux. The research developed correlations between damage modes, coating thickness and surface temperature. A more recent study by Tan et al [76] compares model-derived thermal conductivity and through-thickness temperature distributions with experimental data obtained through laser-based high heat flux testing of plasma spray coatings. Essentially the model provides an understanding of sintering behavior observed under thermal 45

62 gradients testing in terms of isothermally determined experimental data. In this way the model attempts to describe the effect of temperature dependent sintering behavior on the effective thermal conductivity of the coating. 3.5 CMAS degradation of T/EBCs Recent observations of aero-turbine engine components removed from service have revealed the existence of protective coating damage by calcium-magnesium aluminosilicate (CMAS) via ingestion of materials such as sand, volcanic ash, etc. into engine hot-sections. This CMAS debris becomes molten at high temperatures and can infiltrate the T/EBC microstructure. It has been shown in a number of studied that CMAS infiltration can be deleterious to ceramic coatings by both thermomechanical and thermochemical processes. One experimental study performed by Mercer et al [77] proposes a mechanism to describe the decrease in EB-PVD coating durability by CMAS exposure. The researchers concluded that when the coating surface Tsur exceeds the melting temperature of CMAS (TM CMAS 1240 C), the wetting characteristics of CMAS allow it to penetrate the coating to a depth where Tsur = TM CMAS. Then, upon rapid cooling from engine operating temperatures, the CMAS solidifies forming a highly densified phase resulting in an overall stiffening of the coating. This increase in stiffness generates large stresses upon rapid cooling leading to mode I delamination within the coating. A study by Krämer et al [78] studied the susceptibility of thick APS coatings to CMAS infiltration produced delamination. They found that molten CMAS had penetrated to a depth of about half of the top coat and filled all of the open areas. This lead to subsequent channel cracks and delaminations within the coating. 46

63 Depending on the composition of the coating, thermochemical changes resulting from CMAS exposure have been shown to lead to coating degradation as well. In particular Krämer et al [79] investigated the effect of CMAS exposure on an EB-PVD yttria-stabilized-zirconia (YSZ) based TBC system. The researchers observed a change in the columnar microstructure of the coating, densification, and destabilization of the t - zirconia phase. A more recent study by Ahlborg and Zhu [80] investigated the thermochemical effect of CMAS on several advanced rare-earth silicates and rare-earth oxide-doped HfO2 and ZrO2 advanced EBC systems. It was found that while the rareearth doped ZrO2 remained relatively unreacted, the majority of the coating systems suffered severe preferential attack at the grain boundaries. 47

64 CHAPTER IV TEMPERATURE DEPENDENT ELECTRICAL PROPERTIES OF MI-CVI SiCf/SiC CMCs Among the various CMC systems being investigated for use in turbine engine hot sections, silicon carbide (SiC) fiber-reinforced melt-infiltrated (MI) SiC matrix composites have been shown to possess increased high temperature capabilities up to 1315 C [1, 2, 7]. These properties are attributed to the increased densification of MI composites over competing SiCf/SiC CMC processing methods, such as polymer infiltration and pyrolysis (PIP) and chemical vapor infiltration (CVI). The decrease in matrix porosity helps to decrease paths for environmental attack on the reinforcing fibers as well as increasing the load carrying capability of the matrix. If these CMCs are to be used in structural applications, it becomes crucial to investigate and understand the critical damage mechanisms associated with their proposed high temperature operating environments. As previously mentioned, due to the semiconducting nature of their constituents, electrical resistance (ER) measurements have been shown to be effective in sensing tensile damage in several SiCf/SiC CMC systems. Recent work has been performed using electrical resistance monitoring to successfully detect damage accumulation, in the form of transverse 48

65 matrix cracking, in both CVI [56] and MI-CVI SiCf/SiC CMCs [57] under room temperature tensile loading. Morscher et al were able to show that the increase in stress-dependent matrix crack accumulation to rupture increases the ER of the test coupon by hundreds of percent. This work also showed that when performing successive unload/reload cycles, a residual increase in ER at complete unload was related to the increased strain at zero stress caused by matrix cracking and associated matrix/fiber debonding [57]. This finding leads to the possibility of using ER not only as a room temperature in-situ strain monitoring technique, but also as a post-test inspection technique as well. However, because these CMC materials are intended to be used at elevated temperatures, understanding the high temperature damage mechanisms becomes the critical focus. Recent investigations have been done using ER to characterize the high temperature creep properties of MI materials under both furnace-heated (quasiisothermal) and laser-heated high heat-flux (thermal gradient) conditions [81]. While this work points out an obvious correlation between strain and ER increase, it also points out the fact that the electrical response under these conditions can be quite convoluted and is not entirely understood. That is, when performing thermomechanical testing on SiCf/SiC CMCs in oxidizing environments there are several potential factors that can influence ER response. Therefore, if ER is to be used for health monitoring, a systematic approach must be taken to understand the contributions of thermal and mechanical loading as well as environmental effects. The present work attempts to describe the electrical response of MI-CVI SiCf/SiC composites, and the contribution from their various constituents, under thermal loading 49

66 conditions only. Some limited literature data does exist on the temperature dependence of electrical resistivity for various polycrystalline SiC fiber types which extends into the temperature range of interest [82], as well as some CVI-SiCf/SiC composite systems [83, 91]. However, the ER response of MI composites to thermal loading is presumed to be significantly different due to the nature of the siliconized silicon carbide (Si-SiC) matrix material created during the molten silicon infiltration process. The relative high volume fraction of free silicon (Si) left in the matrix after processing (5-15%), will tend to make the matrix material considerably more conductive than that of a purely CVI matrix. Some interesting work has been performed characterizing the temperature dependent ER response of Si-SiC with various dopant types and concentrations [84]. The experimental data illustrates the significant contribution of the excess silicon to the overall electrical response of the system. Therefore, because the electrical resistivity is dependent on both temperature and microstructure, in order to expand ER monitoring to high temperature thermomechanical testing of MI SiCf/SiC CMCs it first becomes necessary to characterize the temperature dependent response of these composites and their constituents. Furthermore, the electrical data taken from this study will prove invaluable to future efforts in modeling the high temperature ER response to damage mechanisms. This paper describes two separate experimental techniques for measuring the temperature dependent electrical resistivity of silicon melt-infiltrated SiCf/SiC composite systems of various reinforcing fiber type: Hi-Nicalon Type S, Tyranno SA and ZMI. The second of which is a novel laser-based heating technique that utilizes standard tensile 50

67 bar specimens used in mechanical testing. The physical mechanisms that dominate the electrical response in different temperature regions are generalized, and when available experimental data from literature is used in order to understand the contribution of the different composite constituents to electrical conduction. To demonstrate this, a simple parallel resistance model is presented that can be used in order to generalize the contribution of each constituent to the overall electrical behavior of the composite system. The results of this model confirm that even at elevated temperatures the electrical current flow through the composite is dominated by the matrix material. Since this is the main operating principle of ER monitoring of room temperature damage, this work confirms the applicability of extending ER measurements for elevated temperature damage characterization. Finally, as the experimental setup utilized in the high temperature laser-heating approach results in the generation of a thermal gradient across the measured length of the sample, the effect of the thermal gradient on the overall electrical response of the material is discussed. A model composed of a series resistance model of temperature dependent elements is presented and compared to experimentally measured ER data. 4.1 Experimental Materials All of the composites tested in this study consisted of 8 plies of balanced 0 /90, 5 harness-satin woven fiber preforms of either: Hi-Nicalon Type S, Tyranno SA or ZMI reinforcing fibers. Using a chemical vapor deposition process the fiber preforms were then coated in a boron nitride (BN) interphase, followed by a layer of CVI SiC, and final densification by a slurry cast molten silicon melt-infiltration process (creating a Si-SiC 51

68 matrix). The samples consisting of Hi-Nicalon Type S (HNS) fibers were manufactured by Hyper-Therm HTC, Inc. (Huntington Beach, CA, now part of Rolls-Royce), while the composites containing the Tyranno SA and ZMI fibers were manufactured by the Goodrich Corporation (Brecksville, OH, now part of United Technologies). The study of various fiber types and material manufacturers provides valuable insight into potential variations in electrical response due to differences in both constituent content and matrix microstructure. 4.2 Experimental Procedure Two separate experimental techniques were utilized to characterize the temperature dependent electrical response of the CMCs mentioned above. The first method involved the use of a commercially available ULVAC-ZEM3 unit [85]. The ZEM3 is a popular instrument in the field of thermoelectrics which uses a four-point probe method to measure the electrical resistance of a small test sample. In a four-point probe method a constant electrical current is applied to the ends of a specimen via two outer electrodes, while two inner leads are used to measure the voltage drop caused by the resistivity of the material. The entire test sample is heated in a low-pressure helium environment by an infrared gold image heating furnace. Figure 13 shows (a) an image of the ZEM3 unit with a (b) close-up of a typical specimen and the ER measurement configuration (all of which is housed inside the furnace). The ZEM3 accommodates only small prismatic samples (6 to 22mm in height), and has is limited in its heating capabilities to approximately 900 C (below the desired usage temperature for many MI- SiCf/SiC CMC structural applications). 52

69 Furnace (a) Power supply Multimeter (b) 20 mm Figure 13: (a) ULVAC ZEM3 unit used for low/intermediate-temperature, isothermal characterization of SiCf/SiC specimens, (b) close-up of specimen configuration, including four-point probe configuration (note: entire specimen housed within furnace). Details on the physical and geometrical properties of the CMC test samples measured using the ZEM3 can be found in Table 3. For all of the test specimens in this study the fiber volume fraction in the longitudinal direction, v f0, has been determined by dividing the cross-sectional area of the test coupon by the fiber area in the longitudinal direction. The fiber area can be estimated by multiplying the area of a single fiber by the number of fibers in a tow, number of tows per ply and number of plies in the lay-up [86]. 53

70 Table 3: Physical and geometrical properties of tested specimens; separated by reinforcing fiber type and test method (i.e. ZEM3 or Laser-based testing). Cross-section* Avg. Fiber radius Fiber volume fraction (mm x mm) (μm) (v f0 ) ZEM3 Samples ZMI 4.00 x SA 2.80 x HNS 4.10 x Laser Samples ZMI x SA x HNS x *dimensions refer to average gage width and thickness In order to overcome the sample geometry and temperature limitations of the ZEM3 unit, a second experimental technique was developed and used to determine the high temperature electrical properties of these CMC systems. This technique utilizes a 3.5 kw CO2 laser to heat the front side of a 44 mm (1.75 inch) gage-section of a standard 152 mm (6 inch) tensile bar specimen (considerably larger than the samples that can be accommodated in the ZEM3 unit). Figure 14 shows a schematic of the laser-based heating setup with integrated ER measurement. This laser setup was specifically developed to demonstrate the capabilities of using ER measurements to monitor CMC tensile coupons up to their maximum operating temperatures. Therefore, this unique setup provides a useful temperature-dependent database as well as insight into the capability of utilizing ER monitoring for future characterization of CMCs under thermomechanical loading conditions. Table 3 outlines some of the relevant properties of the test specimens used in this laser-heating technique. 54

71 CO 2 laser beam y IR Pyrometer FLIR Camera Laser Aperture SiC/SiC tensile bar x y z Heated Region Inner ER leads Thermocouple Outer ER leads IR Pyrometer Figure 14: Schematic of laser-based heating apparatus used for characterization of hightemperature electrical properties of SiCf/SiC CMC tensile bars. The samples containing the Tyranno fibers had been machined into dog-bone specimens (approximate tab and gage widths of 12.7 mm and 10 mm respectively), while the test specimen containing the HNS fibers was a straight sided bar (i.e. no variable cross section). A steel aperture plate containing a 44mm (1.75 inch) diameter circular opening is used to assure a consistent laser-heated area, and the surface temperature of the hot-zone was monitored using an IRCON Modline 5 Series infrared 55

72 pyrometer and FLIR thermal imaging system. This setup is capable of delivering a very stable beam profile, leading to a near constant temperature in the heated region [70]. The electrical resistance of the test specimen is monitored using an Agilent 34420A digital multimeter utilizing a four-point probe method. To avoid damage to the ER system caused by the high temperatures in the heated area, the outer ER electrodes are attached at the specimen ends and the inner electrodes attached at 15 mm from the specimen ends (resulting in an inner probe separation distance of approximately 122 mm). Because the specimens are relatively thin, the surface heating generates a negligible through thickness thermal gradient. However, because only the gage section of the sample is heated, a longitudinal thermal gradient is generated from the edge of the heated region to the specimen end. Implications of the resulting thermal gradient on the sample response are discussed in the proceeding sections. 4.3 Results and Discussion The following section presents the electrical resistance properties of the various composite samples used in this study. First, variations in room temperature ER will be discussed, followed by the temperature dependent electrical properties Room Temperature Resistivity The as-produced room temperature electrical resistivity values for the specimens used in this study are listed in Table 4. The volumetric electrical resistivity, ρ, was calculated from the equation: ρ = RA l (8) 56

73 where R is the electrical resistance determined from the four-point probe method described above, l is the inner probe distance and A is the average cross-sectional area (listed in Table 3) of the sample. The results of these measurements yield some interesting trends. First, the resistivity of the much larger (laser-heated) samples are considerably higher than the ZEM3 test specimens. While this is potentially coincidental, there is the possibility of size effects produced by an insulating effect at the grain boundaries of the constituents (especially the small grain size fibers [87]). That is, assuming a constant grain size between samples of different lengths, a smaller sample would contain less grain boundaries and hence a lower resistivity. A similar size effect could be caused by porosity such that, the electrical resistance of a smaller sample is going to be much more sensitive to porosity than that of a larger sample. Furthermore, it has been shown that the electrical resistivity measured via the ZEM3 unit contains an uncertainty of ± 7% across any temperature measurement [110]. However, by far the most dominate mechanism controlling differences in sample resistivity is likely the consequence of processing variations. Though they may contain the same fiber types, the samples used for each type of test were from different processing batches (i.e. from separately produced CMC panels). The difference in free silicon content in the reaction bonded Si-SiC matrix resulting from the MI process could be significant, which due to the high electrical conductivity of silicon at room temperature (many orders of magnitude higher than SiC), would result in a considerable batch-variation effect on composite resistivity. This can lead to the considerable differences between different panels of the same composite architecture. However, a much less significant variation is 57

74 seen when comparing the ER of each sample used in this study to other specimens machined from the same CMC panel. The room temperature resistivity of the samples tested are listed in Table 4. To determine an average panel resistivity, the resistivity of a number of tensile specimens taken from the same panel as the test specimen were averaged in order to define a nominal panel resistivity. This average is then compared to the resistivity in each sample (see Table 4). Note that the table lists: the average panel resistivity calculated, the number of specimens used to determine this average, and the corresponding scatter in measured values. In general, the resistivity of each sample is in good agreement with its respective nominal panel value. Table 4: Room temperature electrical resistivity (ohm-mm) of specimens and corresponding average panel resistivity. ZEM3 Avg. [# of specimens] Avg. [# of specimens] Laser (scatter) (scatter) Tyranno ZMI [3] (+0.056/-0.083) [3] (+0.208/-0.088) Hi-Nicalon Type S [3] (+0.097/-0.070) [3] (+0.123/-0.117) Tyranno SA [6] (+0.078/-0.076) [7] (+0.468/-0.393) To further investigate the scattering in ER within a panel, discrete measurements along the length of the larger samples can be taken to illustrate the order of anisotropy. For example, Figure 15 shows room temperature ER values of the pre-tested, asproduced (pristine) sample, calculated for small increments along the length (± distance from the center of the sample) of the HNS-reinforced tensile specimen used in the laserbased heating test. The figure shows that there is a similar variation in ER within a sample as variation between samples in the same panel. This supports the previous 58

75 conclusion that, while there is likely to be considerable batch-variation between CMC panels, the resistivity within a panel is considerably more uniform. Another interesting trend seen in the room temperature ER values listed in Table 4 is the consistency between sample types (ZEM3/laser) in terms of reinforcing fiber. That is, for both sets of samples the CMC resistivity is highest in the ZMI reinforced system and lowest in the SA. This trend seems to be consistent with that of recorded fiber resistivity that list these fibers in order of most-to-least resistive as: ZMI, HNS and finally SA [9, 14]. This suggests to some extent the possible contribution of the fiber resistivity to the overall room temperature composite electrical response. However, it is difficult to determine the significance of fiber contribution since (as previously mentioned) the room temperature composite resistivity is largely dominated by the volume of Si in the MI matrix material. Finally, the values of room temperature ER measured post laser-heating indicate an interesting phenomenon resulting from the high temperature heat treatment of the CMC samples. In all cases, the effect of heat treatment on the specimens was to permanently increase the room temperature electrical conductivity (decrease resistivity) of the material. The ER values of the laser-heated samples decreased from the values listed in Table 4 (1.361, 0.931, ohm-mm) to 1.263, and ohm-mm for the ZMI, HNS and SA samples respectively. Similar behavior was not evident however in the ZEM3 tested samples, presumably due to the lower maximum test temperatures to which they were exposed. To further investigate the effect of high temperature exposure on the electrical properties of the laser-heated samples, post-heating room temperature ER measurements were taken at the same locations as the pre-test 59

76 measurements. Figure 15 shows the variation along the length of the HNS-reinforced tensile bar after being tested via the laser-heating technique. 25 to 50 mm 10 to 25 mm ± 10 mm -10 to -25 mm -25 to -50 mm Figure 15: Room temperature electrical resistivity measured from the centerline of the Hi-Nicalon Type S laser-heated sample, measurements were taken in the pretest/pristine condition (striped) and post-laser heating (solid) condition. Note that the heated region of the tensile bar is approximately ± 22 mm from centerline. It is clear that the heated area of the sample saw the most significant residual increase is room temperature electrical conductivity, while the area outside the heated zone (± 25 to 50mm) saw only a negligible change. It is therefore concluded that exposure to high temperatures (>1000 C) effectively changes the microstructure of the composite, resulting in a residual increase is room temperature conductivity. The electrical resistivity of semiconductors like SiC are controlled by various microstructural attributes including: the concentration and type of chemical impurities (dopants) within the material and at the grain boundaries, as well at the grain boundaries themselves 60

77 [93]. Heat treatment of bulk SiC can precipitate these impurities, greatly changing the electrical properties of the material. Also, heat treatment of fine grained polycrystalline SiC fibers has been shown to lead to grain coarsening (grain growth) leading to a decreased number of grain boundaries [87, 95]. Fewer grain boundaries leads to a lower barrier to electron movement and in turn an increase in electrical conductivity. An additional mechanism aiding in increasing conductivity is the possibility that hightemperature exposure decreases the residual thermal stresses within the matrix that were developed by cooling the material from processing temperature. This decrease in residual thermal stress could have a pressure effect on the free silicon within the Si-SiC matrix that results in an increase is conductivity. While it is evident that heating MI-CVI SiCf/SiC composites to elevated temperatures increases their room temperature electrical conductivity, the mechanism that is most significant to the residual change observed in these materials is not known at this time. Looking ahead to applications in CMC damage characterization, the technique demonstrated in Figure 15 could be adopted to quantify post-test mechanical damage of CMCs by looking at the residual increase in ER at room temperature. This would be particularly helpful in identifying areas of increased localized damage associated with stress concentrations and local strain fields. Note that if this technique is being used for inspection following hightemperature mechanical testing, care must be taken to recognize the competing mechanism of conductivity increase due to high temperature exposure and increased resistivity due to accumulated composite damage. 61

78 4.3.2 Temperature Dependence of Electrical Resistivity When considering the temperature dependent response of complex semiconductor materials like MI-CVI SiCf/SiC CMCs, it is first important to consider the mechanisms controlling electron transport in semiconductor materials in general. Therefore, in terms of transport mechanics, electrical resistivity ρ can be expressed as the following: ρ = 1 σ 1 neμ (9) where the electrical response of the semiconductor is governed by the temperature dependencies of both the carrier concentration n and mobility μ, and where the electron charge e is considered a physical constant that does not depend on temperature. In terms of energy band theory, for an intrinsic semiconductor (i.e. a material that does not contain any charge donor or acceptor impurities/dopants) electrical conductivity is dictated purely by the number of electrons shifted from the valence to the conduction band. As temperature T increases the number of electrons excited to the conduction band increases, leaving a proportional number of holes/vacancies in the valence band. The intrinsic carrier concentration ni has be shown to be proportional to T 3/2 exp(-eg/2kt), where Eg is the semiconductor band gap energy (i.e. the activation energy required to move an electron to the conductive band). Since intrinsic materials are free from any dopant atoms there is no effect of impurity scattering of electrons only thermally induced lattice scattering, resulting in electron and hole mobilities proportional to T -3/2. The temperature dependencies of carrier concentration and mobility therefore result in an electrical resistivity controlled by an exponential relationship to the energy band gap exp(eg/2kt) [94]. The natural log of 62

79 resistivity versus the reciprocal of absolute temperature plots are therefore often used to represent electrical data since such plots are indicative of the energy distribution of charge carriers in relation to thermally induced excitations. The inclusion of small amounts of impurity atoms to a semiconductor can greatly affect the electrical properties. The temperature dependence of electron transport of extrinsic semiconductors (i.e. semiconductors containing impurity atoms) demonstrate a slightly more complicated behavior than pure semiconductor materials. In the temperature range corresponding to the extrinsic or saturation region, the carrier concentration n is approximately equal to the impurity concentration (dopant density) ND thereby making carrier concentration effectively independent of temperature. This behavior corresponds to the temperature range labeled as Extrinsic Region in Figure 16a depicting the temperature dependence on carrier concentration for moderately doped n-type Si [97]. (a) 63

80 (b) Figure 16: Typical dependence of: (a) the carrier concentration in a doped semiconductor (constructed assuming a phosphorus-doped ND = /cm 3 Si sample). ni/nd versus T(dashed line) included for comparison purposes [97], (b) mobilities in n-type Si with different electron concentrations. Inset illustrates temperature dependence due to lattice and impurity scattering [94]. Also, for low and moderate impurity concentrations, lattice scattering dominates and electron mobility decreases with temperature as T -3/2. Figure 16b illustrates the nature of the temperature dependencies of carrier mobility for varying dopant concentrations. Therefore, due to the inverse proportionally of resistivity to mobility (Eq. 9), materials at moderate dopant levels and temperatures will show an increasing resistivity with temperature. Resistivity will continue to increase with temperature until a transition temperature is reached at which carrier concentration begins increasing due to the significant number of bonds being ruptured at high temperature. The semiconductor then behaves essentially as an intrinsic material and resistivity decreases (conductivity increases) and follows the intrinsic curve. Increasing of the dopant concentration (number of conductive impurities) will effectively decrease the resistivity of the 64

81 material, thereby shifting the ln(ρ) versus inverse absolute temperature curve down the resistivity axis and move the onset of the intrinsic behavior to a higher temperature. That is, an increased number of bonds are required to be broken in order to increase the number of charge carriers to overcome the large number of dopant atoms. If dopant concentration is very high the effect of impurity scattering increases and at some dopant level will cancel out the effect of lattice scattering, making carrier mobility completely temperature independent. If the material is in the extrinsic temperature region where carrier concentration is also temperature independent, resistivity becomes effectively insensitive to temperature. Hence, heavily doped semiconductors show very little temperature dependence of electrical resistivity. The electrical resistivity of MI-CVI SiCf/SiC CMCs containing various fiber types was investigated up to approximately 900 C and 1300 C using a commercially available ZEM3 unit and novel laser-heating technique respectively. The results of the testing are shown in Figure 17. As with typical semiconductor materials, the initial increase in resistivity with temperature is the result of a reduction in carrier mobility caused by the lattice scattering of valence electrons. For these materials, this behavior appears to change at temperatures ranging from 900 C C (depending on the composite microstructure and testing type) when presumably the increase in carrier concentration becomes more dominant and the samples start to become increasingly more conductive with temperature. 65

82 Figure 17: Temperature dependent electrical response of MI-CVI SiCf/SiC samples tested using the ZEM3 and laser-based heating technique respectively. While all of the samples show trends of this nature, there are apparent differences in activation energies and transition temperatures caused by microstructural differences between specimens. It could therefore be possible to quantify differences in material microstructure based on their individual material response Isothermal Behavior and Parallel Constituent Model If electrical resistance monitoring is to be extended for use in elevated temperature mechanical testing, it is critical that the contribution of each composite constituent at high temperature is well understood in order to correlate observed changes in ER with specific damage mechanisms. Therefore, if the composite is considered as two separate and continuous phases of: (1) fibers in the longitudinal direction v f0, and (2) an effective matrix material meff consisting of the fiber/matrix BN interphase, CVI SiC 66

83 matrix, Si-SiC matrix, and transverse fiber tows (i.e. everything that is not fibers in the longitudinal direction), it is possible to model the composite as a parallel circuit in order to determine the contribution of each constituent phases to the overall electrical conductivity. This parallel circuit model should be a reasonable assumption since both of the assigned phases forms a continuous material throughout the composite. When parallel processes of electrical conduction exist, the total conductivity is the sum of the individual contributions. Therefore, the route with the highest conductivity dominates the conductivity of the system. With respect to resistivity, we can express this parallel circuit in terms of the reciprocal resistivity of each phase and their respective volume fractions. Using the known fiber resistivity and measured composite resistivity it is therefore possible to calculate the contribution of the effective matrix material. v f0 + v m eff = 1 (10) σ c = v f0 σ f + v m eff σ m eff (11) σ m eff = σ c v f0 σ f 1 v f0 1 ρ c = v f 0 ρ f + v m eff ρ m eff (12) (13) ρ m eff = (1 v f0 ) ( 1 v 1 (14) f 0 ) ρ c ρ f The parallel circuit assumption represented in Eq. 13 shows that the reciprocal of the composite resistivity is the weighted average of the reciprocal of the constituents, where the weights following a rule of mixtures assumption representing the respective 67

84 volume fractions of the fibers v f0 and effective matrix vm eff. To illustrate the use of this model to determine the temperature dependence of each composite phase (i.e. fibers and effective matrix), Figure 18 compares the calculated effective matrix conductivity σm eff for the (a) HNS and (b) SA reinforced composites using the isothermal ZEM3 data for each composite system and the fiber conductivity curves reported in Scholz et al. [82]. 68

85 (a) (b) Figure 18: The calculated electrical conductivity of the effective matrix materials (solid lines) of the (a) HNS and (b) SA reinforced ZEM3 tested samples respectively. The ZEM3 measured conductivity data (dotted lines) and fiber data [82] (dashed lines) used in the parallel circuit. The results of the parallel circuit model clearly show that the matrix material not only dominates the room temperature electrical conductivity, but continues to 69

86 dominate with increasing temperature. While the fiber contribution to composite conductivity continues to increase with temperatures, so does the conductivity of the matrix as it transitions into the apparent intrinsic semiconductor region. Therefore, because the matrix is so much more conductive than the fibers across the entire range of temperatures (and continues to increase with temperature), it is believed that the matrix will persist in dominating the electrical response of the composite as it reaches its maximum operating temperature of approximately 1315 C. As previously mentioned, the fact that the matrix controls the conductivity of the CMC is the basic principle used in room temperature ER monitoring to correlate change in the resistance to composite damage accumulation via matrix cracking and fiber sliding [56-58]. The results of this simple parallel circuit model confirms that this is also the case at elevated temperature. Therefore, similar damage mechanisms as those described in room temperature tensile tests should also correlate to change in ER during high-temperature thermomechanical testing Effect of Thermal Gradient on Overall Electrical Response of Laser-heated Specimens Due to the high test temperatures reached in the laser-based heating approach, the inner ER electrodes utilized for recording the temperature-dependent electrical response are placed outside of the laser-heated region (± mm) near the ends of the tensile bar (± 61.2 mm) as shown in Figure 14. While this setup prevents damage to the ER electrodes from high temperature exposure, it means that (unlike the isothermal 70

87 testing performed in the ZEM3) the resistance recorded during laser testing is a measurement of the heated and non-heated length of the specimen and hence depends on the thermal gradient throughout the specimen. Therefore, it becomes important to understand the role that this thermal gradient has on the overall electrical behavior of the composite during high temperature testing. The methodology used for determining the temperature distribution along the length of a sample is outlined in Appendix A. For a given hot-zone temperature, the longitudinal thermal gradient in a laser-heated sample is determined by numerical solution of the steady-state heat equation, Eq. 21. An investigation of the effect of the thermal gradient on electrical response was performed by examining a ZMI tensile specimen machined from the same panel as the ZMI ZEM3 tested sample was heated using the laser-based approach. This sample (denoted as ZMI-A) is a 6 inch tensile dogbone specimen with nominal gage cross-sectional dimensions (w x t) of mm x 4.09 mm (extending to ~12.7 mm width in tab region). Because the ZMI-A tensile specimen was machined from the same panel as the isothermal test specimen, it contains a similar microstructure and likewise similar physical properties. The calculated temperature profiles along the horizontal centerline of the length of ZMI-A from the edge of the heated-region to the end of the sample is shown in Figure 19 for various steady-state values of hot-zone temperature. 71

88 Figure 19: The calculated longitudinal thermal gradient for the ZMI-A laser-heated sample. Note that each curve represents the temperature profile for a given hot-zone temperature to the end of the tensile specimen. The determination of the temperature dependent parameters used in the calculation of the temperature profiles are shown in Figure 42 (Appendix A). Heat losses due to thermal radiation were calculated assuming the temperature dependency of emissivity for bulk SiC [112], while the temperature dependent values of thermal conductivity were deduced following a linear fit of literature data established for an MI- CVI SiCf/SiC laminate material [7]. Finally, the values used for the convection heat transfer coefficient were determined via empirical relationships established for the natural convection from the surface of a horizontal flat plate [111]. It is worth noting that each profile shown in Figure 19 assumes constant heat transfer properties along 72

89 the length, based on the given temperature of the heated region. This will of course lead to some inaccuracies in the modeled temperature distributions. Determination of the temperature profile along the length of the specimen can be used in order to model the total resistance of the sample. A simple approximation of the overall resistance of the specimen is done by considering the tensile sample to be a series of distinct elements each assigned a single resistance value, hence the entire sample is modeled as of a series of discrete resistors. By rewriting Eq. 8, each resistor in the series can be expressed as a function of the length, cross-sectional area, and temperature-dependent resistivity of the element. Note that by taking a variable crosssectional area the model is able to account for the change in width from gage to end section of the tensile dog-bone specimen. The total composite resistance can therefore be expressed as the some of these resistance elements as: R c tot = R i = ρ(t i)l i A i (15) The temperature-dependent resistivity is taken from the experimentally determined isothermal response from the ZEM3 unit. Due to the temperature limitations of the ZEM3 unit, the high temperature response (> 900 C) can be estimated by extrapolating the data following a typical intrinsic semiconductor (ln(ρ) α 1/T) relationship. It is worth noting that, due to the large variation in resistivity values between composite samples from different panels, it is necessary to use experimental isothermal data taken from a similar resistivity sample to accurately populate the series model. Therefore, in order to 73

90 verify the validity of the proposed model, the isothermal ZMI data shown in Figure 17 is extrapolate to 1300 C and used to model the experimental response of ZMI-A. The modeling results, along with the associated error is shown in Figure 20. Figure 20: Comparison of measured electrical resistance to hot-zone temperature of ZMI- A to proposed series resistance model. The dashed line representing the results of the model are in general agreement with the experimental data. However, there does appear to be the highest discrepancy across the intermediate hot-zone temperatures (500 C C). This is likely due to a combination of the relative noise of the signal across these temperatures, as well as the simplifying approximations used in the solution of the temperature profile of the sample. Notably, the use of a constant thermal conductivity and convection coefficient. 74

91 Furthermore, while the ZEM3 sample used in the model is from the same composite panel as the laser-heated sample, the inhomogeneous nature of MI matrices of these CMCs certainly attributes to some variation of material response. 4.4 Conclusions The present study explores the temperature dependent electrical response of MI- CVI SiCf/SiC CMCs using both isothermal and laser-based heating approaches. The study examined composite systems containing various reinforcing fiber types, including: Tyranno ZMI and SA, as well as Hi-Nicalon Type S. The novel laser-based setup developed for this study was capable of overcoming the sample size and temperature limitations of typical characterization techniques. Specifically, the laser system was used to heat the gage section of standard tensile coupons used in mechanical testing to temperatures that represent the desired usage temperatures of MI-CVI SiCf/SiC composites. Therefore, the data collected using this methodology not only provides a unique materials database, but also insight into the possible complexities of extending the use of ER measurements into high temperature thermomechanical tensile testing and damage characterization. Comparison of the room temperature resistivity of the various test samples represented the wide variation in electrical properties of these materials due to microstructural differences. While sample machined from the same CMC panel appear to have similar resistivity values, samples taken from different panels can vary greatly. For composites containing the same type and relative volume of reinforcing fiber this is 75

92 presumably caused by variations in matrix microstructure. Specifically, the volume fraction and dispersion of the free silicon remaining in the matrix post melt-infiltration. Clearly, composites containing different fiber types have a combined effect of differences in fiber resistivity and content (however small) as well as these differences in matrix composition. Furthermore, it was found that exposure to very high test temperatures resulted in a residual increase in the room temperature resistivity of the sample. This finding is significant if ER measurements are to be used to characterize materials that have been exposed to extreme temperatures. That is, knowing the effect of high temperature heat treatment a prior is necessary for the deconvolution of a complex ER response. In terms of temperature dependent behavior, all of the composites tested experienced an initial increase in resistivity with temperature with an observable transition at high temperature. Similar to typical moderately-doped semiconductor behavior, the resistivity increased with temperature up to some transition temperature at which point the resistivity began to decrease. The ability of the laser-based heating setup to achieve temperatures above this transition point allows for a more accurate representation of material behavior in extreme environments. Moving forward, quantification of variations in activation energy and transition temperature could possibly be used to characterize differences in microstructure or impurity concentration. By considering the composite material as a parallel circuit of continuous phases of SiC fibers and effective matrix material, the contribution of each phase to the overall 76

93 conductivity of the material was investigated. It was found that the silicon rich siliconized-sic matrix resulting from the melt-infiltration process dominated the ER of the composite at all temperatures. This finding is extremely important if the use of electrical resistance monitoring is to be extended to damage monitoring of MI-CVI SiCf/SiC composites under high temperature tensile loading. It has been shown in room temperature uniaxial tension testing of similar materials that because the matrix dominates electrical conduction, that increasing matrix cracking correlates to increasing electrical resistance. Therefore, by demonstrating that the matrix continues to dominate electrical conduction al all temperatures of interest, the same working principle should apply and this type of ER monitoring remains a feasible and cost effective method of damage quantification. Finally, the significance to overall specimen response of the longitudinal thermal gradient developed using the laser-based heating approach was investigated. One major advantage of this test method is that it can be used to demonstrate the use of ER monitoring in high temperature mechanical testing. However, due to extreme temperatures it becomes necessary to remotely sense the ER response of the gage section. By calculating the temperature profile and using the temperature dependent resistivity data collected using the ZEM3 unit, the contribution of the thermal gradient along the length of the sample to the overall ER response was shown. The results of the approximation using the series circuit model were in general agreement with experimentally obtained results. 77

94 CHAPTER V EFFECTS OF GEOMETRIC STRESS CONCENTRATIONS AND HIGH HEAT-FLUXES ON TENSILE DAMAGE OF CMCS When implementing CMCs for structural applications, it is important to consider the effect of localized areas of stress concentrations on damage morphology and load carrying capabilities. Several studies have been conducted in order to determine the effect of stress concentrations arising from geometric discontinuities (notches, holes, etc.) on the room temperature ultimate strength of CMCs. Recent studies have found the strength of melt-infiltrated SiC/SiC composites to be only mildly notch sensitive [98, 99]. However, it is important to characterize not only ultimate strength, but damage resulting from localized stress concentrations at stresses well below the ultimate strength. This is due to the fact that because these materials are expected to withstand long durations in high-temperature oxidizing environments, localized matrix cracking in the vicinity of geometric stress concentrations allow for regions of oxidation ingress and fiber degradation that could result in decreased toughness and early composite failure. A second source of stress concentrations in CMC structures would be caused by large thermal gradients. As discussed in Chapter 3 of this work, many of the proposed 78

95 turbine engine applications for CMCs (turbine engine vanes/blades, combustor liners, etc.) will be subjected to high heat-fluxes and complex stress states caused by asymmetrical thermal loading and the use of forced air cooling [100, 101]. Therefore, if CMCs are to be successfully implemented as structural components in these environments, it becomes critical to examine and characterize the damage initiation and accumulation under similar thermal, mechanical and environmental conditions. Current standard testing techniques used for material characterization are therefore incapable of adequately simulating the high temperature, high heat-flux environments proposed for CMC systems. The work presented in this chapter will characterize the inelastic damage surrounding stress concentrations under high heat-flux testing conditions. A major aspect of this work is the incorporation of a novel laser-based heating apparatus with a standard electromechanical tensile testing fixture to accommodate high temperature testing. As part of NASA GRC s high heat-flux testing program, this rig allows for unique thermomechanical testing under simulated engine environments. This work also focuses on the investigation of various non-destructive evaluation (NDE) techniques that could provide critical information useful in assessing damage development and material state during complex testing scenarios. Ideally, such techniques would be simple to implement on a material level as well as a structural testing scale. Recent studies have shown the viability of using in-situ modal acoustic emission (AE) and electrical resistance (ER) monitoring of SiC/SiC CMC specimens during room temperature monotonic tensile loading in order to detect damage onset and 79

96 accumulation in both CVI [56] and MI-CVI matrices [57]. These works were successful in determining a correlation between electrical resistance increase and stress-dependent transverse matrix cracking. However, because CMCs are intended for high temperature use, investigation of damage mechanisms at elevated temperatures becomes the critical focus. The results in Chapter 4 suggest that ER measurements could be a useful technique for monitoring high temperature tensile damage in the form of matrix cracking and corresponding fiber sliding in MI composites. The tests performed in this chapter utilize a similar high temperature ER monitoring setup for in-situ measurements and post-test, room-temperature inspection technique. One benefit of using a notched tensile specimen in this study is that it helps to overcome the disadvantage of using a far-field in-situ ER measurement. That is, a notched specimen helps to establish a highly localized strain field within the heated gage section, ensuring that the majority of the plastic damage (and therefore ER increase) occurs in the heated area of interests. The stress concentration will also generate a damage gradient allowing for a controlled investigation of the fidelity of the post-test ER inspection technique. Finally, a novel DIC setup and procedure was designed in order to map localized strain fields under high heat-flux testing environments. Due to the inherent difficulties of implementing DIC in high temperature mechanical testing, the ability to overcome many of those issues can provide an invaluable tool for future research in this field. In summary, the objective of this chapter is to investigate the damage morphology of notched tensile specimens in high heat-flux environments. To this end, the feasibility of using ER measurements for high temperature testing is investigated. Furthermore, 80

97 in-situ modal acoustic AE was implemented in order to correlate the change in ER due mechanical loading to stress-dependent transverse matrix cracking. Analysis of modal AE events was performed to determine damage locations which can then compared with ER inspection as well as microscopy of the notched area in order to investigate the damage morphology associated with stress concentrations. A novel digital image correlation DIC technique was also developed that allows for comparison of high temperature ER and AE results with the localized strain mapping of the specimen gage section prior to ultimate failure. 5.1 Experimental Materials The ceramic composite sample tested in this study consist of a Hi-Nicalon Type S (HNS) fiber-reinforced, slurry cast, melt-infiltrated SiC/BN/SiC in 8 plies of balanced 0 /90, 5 harness-satin weave manufactured by Hyper-Therm HTC, Inc. (currently Rolls- Royce, Huntington Beach, CA). It should be noted that this sample is of the same architecture and was produced by the same manufacturer as the HNS reinforced samples described previously in Chapter 4. A 152 mm (6 in) tensile coupon was machined from a composite panel to be used for testing. Some physical details of the specimen used in this study have been listed in Table 5. Prior to testing, further machining was done on the sample in order to create the stress concentrations with the notch geometry shown in Figure

98 Table 5: SiC/SiC specimen geometry, fiber content and room-temperature electrical properties pre-tensile testing. Specimen Width (mm) Net Width (mm) Thickness (mm) v f0 Electrical Resistivity (ohm-mm) Room temp. Test temp. DN Notch geometry DOUBLE NOTCH configuration 1.29mm 1.05mm Figure 21: Notch geometry for the high-temperature tested double-notch SiC/SiC tensile bar. 5.2 Monotonic tensile testing with high heat-flux capabilities Mechanical testing of the specimen is performed in uniaxial tension and loading is conducted at a rate of mm/min. For the application of the thermal loading a high power (3.5kW) high heat-flux CO2 laser is used to asymmetrically heat the material specimen, generating a multi-axial (through-thickness and longitudinal) thermal gradient. A spinning optical system is used to distribute the beam into an approximately 32 mm heated section on the front surface of the specimen. The through-thickness thermal gradient can be increased by the addition of active air-cooling delivered by a shower-jet head aim at the back side of the material. The front and back-side temperatures of the heated zone are monitored using a pair of infrared pyrometers. For this study, the sample was heated to a CMC surface temperature of 1208 C and the resulting backside temperature of 1000 C. The tensile specimen is held by ceramic grip 82

99 inserts (to eliminate electrical interference of the ER measurement) in the electromechanical tensile load frame. Heating of the specimen is carried out under a no load condition in which free thermal expansion in the longitudinal direction is allowed to prevent excessive thermal stresses. Elongation of the specimen gage-section is monitored via two separate methods. First, the nominal strain of the gage section is measured using a high-temperature extensometer with a 25.4 mm gage section with a ±0.5 mm travel. Secondly, the above mentioned DIC technique was developed to determine localized strain fields within the gage. In order to track the displacement field, a high temperature Y2O3 based aerosol paint (ZYP Coatings, Inc., Oak Ridge, TN) was sprayed onto the sample surface prior to testing, forming a random speckle pattern. This paint was chosen due to its high temperature stability during elevated temperature testing. In order to track the evolution of the strain field, post processing was performed on the DIC images using commercially available DIC software (ARAMIS, GOM Co. Ltd., Germany). The sample was illuminated using two high intensity LED lamps angled toward the area of interest. Also, a fan was used to reduce heat distortion of the image by mixing the air in front of the heated specimen. The details of the DIC setup and testing rig are shown in Figure

100 Load Frame White LED Lamps Extensometer Specimen Laser Optics DIC cameras Pyrometer Figure 22: Image of tensile loading frame with incorporated high heat-flux laser heating and digital image correlation (DIC) apparatus. During tensile testing, the electrical resistance of the specimen is monitored using an Agilent 34420A digital multimeter utilizing a four-point probe method to minimize the contribution of contact resistance. Due to the high temperature of the heated area during testing, the ER electrodes are attached to the material specimen inside the test frame grips. This configuration ensures good contact pressure between electrode and sample, and also prevents exceeding the melting temperature of the copper electrodes. The outer electrodes are attached at the specimen ends, with the inner electrodes attached at 15 mm from the specimen ends (resulting in an inner probe separation distance of approximately 122 mm) as shown in Figure 23. Note the intentional similarity to the laser-heating ER setup described in Chapter 4. Acoustic emission events 84

101 were monitored using a Digital Wave Fracture Detector equipped with two wide-band (50 khz 2 MHz) sensors attached to the top and bottom of the specimen (approximately ±40mm from coupon centerline), and data was acquired at the rate of 10MHz when triggered. The location of the AE events was determined by calculation of the difference in time of arrival between each sensor. CMC Tensile Specimen CO 2 laser beam Heated Region Cooling Air Inner ER Outer ER Modal AE sensor Ceramic Grip-insert Figure 23: Schematic of laser heat-flux tensile test setup; including configuration of electrical resistance (ER) electrodes and modal acoustic emission (AE) sensors. 5.3 Results and Discussion The results of this study will first be presented in terms of mechanical behavior of the composite specimen. This will then be followed by an explanation and in-depth discussion of the results of the non-destructive evaluation techniques utilized. 85

102 5.3.1 Mechanical Behavior An approximation of the longitudinal thermal stress state in the heated region upon reaching steady-state temperature, prior to mechanical loading, was estimated using Eq. 32 and the simplifying approximations described in Appendix B. The results of this analysis based on the measured front and back side temperatures of the sample are shown in Figure 24. The composite elastic modulus Ec = 169 GPa was determined from the mechanical data shown in Figure 25 and the average in-plane coefficient of thermal expansion α = 3.74*10-6 / C was estimated based on reported literature values from a similar slurry cast MI-CVI SiCf/SiC composite system [7]. (a) 86

103 (b) Figure 24: (a) Estimate of through thickness thermal gradient based on measured front and back side temperatures of specimen heated-region. (b) Longitudinal thermal stresses in heated region as a function of sample thickness. As noted in Appendix B this is likely an overestimate of the true thermal stress state in the heated region. Furthermore, while not insignificant, the tensile composites stresses calculated on the back surface are well below the proportional limit and therefore are unlikely to induced matrix crack initiation. This assumption is confirmed by the lack of any AE events recorded prior to mechanical loading. The mechanical stress-strain relationship for the high heat-flux, monotonic tensile test of the double notched specimen is shown in Figure 25. For this plot of mechanical behavior, the stress is defined in as the net-section stress (i.e. the mechanical stress based on the minimum cross-section at the notch tip plane) and the nominal strain (i.e. the gage-section mechanical strain measured via the high-temperature extensometer). The specimen saw a 0.53% nominal strain increase from thermal loading, with an additional 0.29% nominal mechanical strain to failure at an ultimate strength of 240 MPa (net-section stress). 87

104 Figure 25: Tensile stress-strain response of double notched specimen under high heatflux conditions. Note that the stress is the net-section stress and strain is the nominal mechanical strain recorded by the extensometer. After testing, a section of the test coupon was mounted and polished for optical microscopy. The main goal of the microscopy was to compare the crack densities surrounding the notch and in the far-field. Crack densities along the gage section were calculated for regions ±5 mm from the center of the notch, and the remaining far-field gage sections ± 5 mm < x < ± 12.7 mm. The results of these crack density measurements are listed in Table 6. Table 6: Crack density measurements for each region (around the notch and far-field) of the gage section ± 5mm ± 5mm < x < ± 12.7 mm Crack Density, mm (far-field/notch) 1 (0.74) 88

105 The specimen failed along the notch plane and an increased crack density was observed in the notched region. Specifically, the results show that the region surrounding the notch had the higher crack density (1.1 mm -1 ), with the far-field section of the gage at 74% of that value AE Waveform Analysis It has been shown that acoustic energy is directly related to transverse matrix cracking in CMCs [47]. Therefore, the cumulative energy of AE is often plotted versus applied stress in order to represent the accumulation of stress-dependent transverse matrix cracking. Due to the damage gradient generated along the length of the specimen, the amount of AE energy released from each region of the gage section is compared separately. To compare the relative matrix cracking in each area, the cumulative energy was normalized in each region by the total energy recorded in the notched region (i.e. ± 5mm from mid-plane). In this case, because the regions being compared are in different stress states, the cumulative AE energy is considered in terms of applied load. The results of this comparison are shown in Figure 26a. It is worth noting that the regions considered are of different length scales. Therefore, by normalizing by their respective lengths it is possible to consider the effective energy density of each region. The energy density has be plotted in Figure 26b. 89

106 ± 5 mm < x < ± 12.7 mm ± 5 mm (a) ± 5 mm < x < ± 12.7 mm ± 5 mm (b) Figure 26: Cumulative AE Energy of the sample: (a) normalized by the total by the total energy recorded in the notch region, and (b) also normalized by length to compare energy density. There are several noteworthy observations that can be made from the AE data shown in Figure 26b. First, the area in the vicinity of the notch shows increasing AE 90

107 energy (i.e. increasing matrix cracking) beginning at a much lower load. This is clearly an effect of the stress concentration at the notch resulting in an area of increased stress at the applied load. Therefore, the matrix in the notched region reaches the stress to initiate microcracking much sooner. Secondly, the region outside the notch has a lower energy density, approximately 73% of that of the notched region. This is particularly interesting as it is in exact agreement with the ratios of crack densities shown in Table 6. This observation appears to once again confirm that the acoustic energy can be used as a valid measurement of matrix crack distribution. The location of a given AE event (x) is estimated by the difference in times of arrival (determined by a minimum threshold crossing) of the AE wave at the top (ttop) and bottom (tbottom) sensors: x = v 2 (t bottom t top ) (16) AE wave velocity (v) was determined prior to high-temperature tensile testing by performing pencil lead breaks at various locations along the specimen length in order to manually generate AE sources. Because material damage will decrease the speed at which sound travels through the material, the wave velocity as a function of test time is ascertained by manual determination of AE velocity from a number of discrete events generated during testing [45]. The distribution of AE events are shown in Figure 27a. The AE events are plotted versus the length along the specimen (shown as ± distance from centerline of the sample). In the figure, the size of each data point represents the relative energy of the event. The loud AE events have been highlighted to emphasize 91

108 the overall trend of event location. In this case the loud events refer to any event that is at least 10% of the loudest event recorded. Notches (a) (b) Figure 27: (a) Left: AE event location within the heated gage region (show as ± mm from the specimen centerline) versus net-section stress. Note that the size of the marker indicates the relative energy of the AE event. Right: The DIC image of the gage section taken just prior to failure. (b) Distribution of AE event energy along the gage length. The most significant thing about the acoustic activity is that at low stresses it is only visible in the notch region (± 5mm). It only becomes significant away from the notch at 92

109 an applied load of approximately 3.5 kn (corresponding to net-section and far-field stresses of 150 and 120 MPa respectively). To the right of the AE event plot is the DIC image of the gage section just prior to ultimate failure. Not only was the DIC able to capture the localized strain field in the vicinity of the notched region, the image appears to be in good agreement with the damage location determined by the distribution of AE events. As discussed in Chapter 3, CMCs will experience multiple matrix cracking planes prior to ultimate failure of the sample. Therefore, to further investigate the regions of highest damage accumulation, Figure 27b shows the distribution of AE energy along the gage section of the sample. While it is clear that a high concentration of AE energy can be seen along the centerline of the composite (corresponding to the fracture plane), it is also interesting that there appears to be another high energy region toward the top of the gage Electrical Resistance Measurements The in-situ ER monitoring system capture the electrical response of the specimen during the thermomechanical testing. Figure 28 shows the percent change in resistance due to mechanical loading. It is worth noting that because the ER is monitored over the entire specimen length this change in resistance is due to the contribution of the localized strain accumulation around the notch as well as damage in the far-field. 93

110 Stress ER Figure 28: Percent change in electrical resistance of the specimen versus nominal strain increase of the gage section during monotonic tensile loading to failure. A few attributes of the nature of ER increase are worth noting. First, there is little increase in ER before 0.1% strain. This appears to coincide with the elastic limit of the composite. However, upon loading beyond the matrix cracking stress (signaled by the decrease in modulus), the ER begins to increase rapidly. It is clear therefore, that the ER monitoring is sensitive to matrix crack initiation. Secondly, it appears that the ER monitoring is very sensitive to plastic strain accumulation. That is, the data shows that less than a 0.3% strain to failure of the gage section correlates to an increase of resistance of the specimen of nearly 100%. After the specimen failed, it was allowed to cool back to room temperature and removed from the load frame. Post-test, room temperature ER measurements were taken along the length similar to the technique described in Chapter 4. The 94

111 measurements taken along the entire length of the sample prior to any thermal or mechanical loading is shown in Figure 29a. Similar to the specimen shown in Figure 15, the resistivity of the pristine sample is relatively uniform along the length. Note that the arrows are used to indicate the general region of the sample (between the paint marks) corresponding to the data point. These measurement can then be compared to similar measurements taken of the sample post high temperature tensile fracture (Figure 29b). 95

112 (a) (b) Figure 29: Comparison of room temperature ER measurements taken along the length of the specimen (a) prior to any thermal or mechanical loading (i.e. pristine state), and (b) post high heat-flux tensile strength test. Note that the measurements refer to the distance in mm from the centerline of the sample. The data shows that the entire specimen resulted in an increase in room temperature resistivity caused by the high-temperature mechanical loading. This is indicative of damage in the form of matrix cracking that resulted in a permanent increase in the 96

113 resistance of the specimen. An interesting observation can be made if these measurements are grouped into two distinct regions: (1) the measurements containing the heated gage section (25 to 5mm and -5 to -25mm), and (2) area outside the heated region. It can be seen that the increase in electrical resistivity is highest in the area immediately outside heated region (50 to 25mm). This result appears counterintuitive because, while the entire specimen was subjected to stress-dependent damage, the region containing the stress concentration would presumably result in the largest ER increase due to the highly localized damage it experienced. This seemingly anomalous behavior can be explained by recalling the post laser-heating room temperature ER results shown in Chapter 4 (Figure 15). Those results showed that exposure to high temperatures (>1000 C) produce microstructural changes that result in a residual decrease in resistivity. Therefore, in the heated gage region there is a competition between the increase in ER due to damage accumulation and the decrease in ER due to high temperature exposure. However, it can be observed that this competition was dominated by the damage accumulation and ultimately resulted in a net increase in ER. Another interesting note is that upon close visual inspection of the post-test CMC surface, it is evident that the specimen was aligned in such a way that the -5 to -25mm section contained more of the laser-heated region than the opposing 25 to 5 section. This is the likely cause of the lower resistivity measured from -5 to -25mm, by the same reasoning as above. 97

114 5.4 Conclusions This chapter focused on the damage characterization of SiCf/SiC CMCs with stress concentrations under high heat-flux tensile loading. A specialized laser-based testing rig was utilized in order to simulate the high heat-flux environments these materials will be subjected to in turbine engine hot-sections. Several techniques were used to in order to investigate the nature of damage accumulation during testing including post-test microscopy, AE and ER monitoring and strain mapping via a novel DIC technique. The matrix crack morphology seen in the post-test microscopy was in good agreement with the AE (energy and event location) as well as the areas of localized strain observed using the DIC system. In-situ ER measurement shown to be an effective tool for detecting damage onset and accumulation under these high-temperature testing conditions. Posttest room temperature ER measurements were also shown to be a useful post-test inspection technique. The results showed a net increase in electrical resistivity along the entire length of the samples. However, these results also indicated the convoluted nature of these ER measurements due to the competition in material response to high temperature exposure and mechanical loading. 98

115 CHAPTER VI ELEVATED TEMPERATURE, HIGH HEAT-FLUX EBC/CMC DAMAGE AND STRENGTH: POST ENVIRONMENTAL EXPOSURE Environmental barrier coated ceramic matrix composite (EBC/CMC) systems are currently being investigated for use as turbine engine hot-section components in extreme environments. Under these conditions, it becomes critical to understand material response to environmental exposure and performance under thermomechanical loading. Furthermore, because Si-based composites will likely require protective coatings for thermal and oxidation resistance it is important to consider them from an EBC/CMC systems approach. In the previous chapter electrical resistance measurements were shown to be correlated to tensile damage accumulation in SiC/SiC CMCs, and the focus of this study is to extend the use of ER to evaluate hightemperature thermal gradient fracture of EBC/CMC systems. Tensile strength tests were performed at high temperature (1200 C) using a laser-based heat-flux technique. Specimens included an as-produced SiC/SiC CMC and coated SiC/SiC substrate that have been exposed to simulated combustion environments in a high-pressure burner rig. Localized stress-dependent damage was determined using acoustic emission (AE) monitoring and compared to full-field strain mapping using a high-temperature digital 99

116 image correlation (DIC) technique. The results are compared with in-situ ER monitoring, and post-test inspection of the samples in order to correlate ER response to damage evolution. The present work compares the ER response of coated and uncoated MI-SiC/SiC CMC substrates subjected to exposure to simulated engine environments followed by high-temperature monotonic tensile strength tests in order to assess retained material properties and strength. From the previous chapters it is clear that high temperature electrical response of these systems can be quite convoluted and requires further examination. Therefore, if ER is to be used as a health monitoring technique, a better understanding of the effects of thermal and mechanical loading, as well as environmental effects is necessary. In-situ modal AE monitoring will be used in order to correlate changes in electrical response with stress-dependent matrix cracking during tensile loading, and to provide insights into material damage state that ER measurements may not be sensitive to. Furthermore, a novel high-temperature digital image correlation (DIC) technique is described that allows for the comparison of hightemperature ER and AE results with localized strain mapping of the specimen gage section during testing. The primary objective of the present paper is to investigate the difference in damage mechanics between coated and uncoated CMC substrates after exposure to simulated engine environments. The feasibility of using ER measurements in high-temperature testing environments, specifically combined high heat-flux and mechanical loading conditions, is further investigated. In addition, the various NDE methods described 100

117 earlier (ER, AE and DIC) are used to determine and quantify damage onset and accumulation, and CMC degradation due to the simulated engine environment exposure. Finally, evidence of EBC damage onset and failure is presented by post-test AE event waveform analysis. 6.1 Experimental Materials The ceramic composite substrates tested in this study consist of 152 mm (6 in) tensile bars machined from a panel of a Hi-Nicalon Type S (HNS) fiber-reinforced, slurry cast, melt-infiltrated SiC/BN/SiC in 8 plies of balanced 0 /90, 5 harness-satin weave manufactured by Hyper-Therm HTC, Inc. (currently Rolls-Royce, Huntington Beach, CA). Details of the two CMC specimens used in this study are shown in Table 7. Table 7: SiC/SiC specimen geometry, fiber content (vf0), and room-temperature electrical properties pre-tensile testing. Specimen Width (mm) Thickness (mm) vf0 Electrical Resistance (ohm) Room temp. Test temp. coated uncoated Of the CMC tensile specimens, one was coated with a multilayer EBC coating. The coating consisted of a thin bond coat layer to ensure good adhesion to the CMC substrate, followed by a top coat layer for environmental protection. Both layers were deposited to the surface via electron beam-physical vapor deposition (EB-PVD). The bond coat consisted of a newly developed NASA HfO2-Si composite bond coat, and the top coat was deposited as NASA HfO2-doped ytterbium-gadolinium di-silicate (Yb,Gd)2Si2O7 environmental barrier coating system [102, 103]. 101

118 6.2 High Pressure Burner Rig Exposure In order to simulate exposure to harsh engine environments, an existing high pressure burner rig (HPBR) at NASA Glenn Research Center (Cleveland, OH) was used. The HPBR has been used previously for high-temperature environmental durability studies of advanced aircraft materials and coatings. While design, construction and operation can be found elsewhere [12, 15], the HPBR burns jet-fuel and air at user controlled ratios. The burner rig has been significantly upgraded, and the test conditions (i.e. gas pressures, temperatures and velocities) closely simulate aero-turbine engine environments [104]. The coated and uncoated SiC/SiC specimens were tested in the high pressure burner rig at 1315 C for 30 hours at 10 atm and combustion gas velocities of 200 m/sec in order to evaluate SiC/SiC degradation and the effect of the environmental barrier coating. 6.3 High-Temperature Monotonic Tensile Testing The same laser high heat-flux testing rig described in Chapter 5 is utilized in this study to assess retained high temperature tensile properties post-hpbr exposure. Thermal gradient testing is used in order to more closely simulate the thermal loading conditions expected for many EBC-CMC aero-turbine engine applications. The front and back-side temperatures of the heated zone are monitored using a pair of infrared pyrometers. It should be noted that the front-side temperature measurement refers to either the coating surface or the substrate surface (depending on whether coating is applied), while the back-side measurement is on the opposing, non-heated face. For this 102

119 study, the samples were both heated in order to produce CMC surface temperatures of approximately 1200 C. Once again, the nominal strain of the gage section is measured using a hightemperature extensometer and strain mapping of the gage section was performed using the previously described high-temperature DIC. Furthermore, the same NDE setup for ER and AE monitoring shown in Figure 23 of Chapter 5 was also implemented to characterize damage accumulation during thermomechanical testing. 6.4 Results and discussion The results of this study will first be presented in terms of the mechanical behavior of the composite specimens during post-hpbr exposure retained strength testing. This will then be followed by an explanation and in-depth discussion of the results of the non-destructive evaluation techniques utilized Retained properties The monotonic tensile results for the two composite systems appear in Figure 30. The coated sample was heated to an EBC surface temperature of 1230 C and SiC/SiC substrate back side temperature of 1070 C, and the uncoated sample was heated to SiC/SiC front and back side temperatures of 1200 C and 1010 C respectively. Note that as these materials possess a similar architecture and are subjected to approximately the same through thickness temperature difference as the sample described in Chapter 5, they will have a similar composite thermal stress state caused by thermal gradients. 103

120 Extensometer DIC coated uncoated coated uncoated Figure 30: Mechanical behavior of samples under high temperature monotonic tensile loading, post environmental exposure. Strain measurement taken from high temperature extensometer, as well as from contact points of extensometer probes as observed from DIC images. The stress-strain behavior for the two samples is plotted in terms of nominal strain measured by the high temperature extensometer, as well as the change in elongation of two discrete points taken from the DIC mapping at the points of contact of the upper and lower extensometer probes. It is clear that while there is a higher degree of noise in the DIC analysis, both systems are in general agreement in terms of mechanical behavior. Note that the DIC data is offset by the thermal strain induced by specimen heat up, whereas the extensometer data reflects only the mechanical strains. 104

121 The stress-strain response of the coated sample is typical of the desired mechanical behavior for continuous fiber-reinforced ceramic matrix composites. As discussed in Chapter 3, unlike monolithic ceramics that generally fail by rapid growth of a single crack, fiber-reinforced CMCs rely on global load sharing between the fibers and matrix upon crack initiation. This mechanisms results in the region of pseudo-toughness and increased strain accumulation seen in the coated sample that follows its initial elastic loading response. Assuming that the interphase is sufficiently weak and the fibers sufficiently strong, when the applied stress exceeds the stress necessary to initiate matrix cracking the local stress on the matrix is relieved and the load is shed to the crack bridging fibers via an interfacial friction mechanism along the debonded length [27, 29-30]. The differences in retained mechanical properties between the samples are listed in Table 8. Table 8: Conditions and mechanical properties of test samples during high-temperature tensile testing. Specimen Surface Temp. ( C) Back Temp. ( C) E (GPa) Extensometer DIC σ UTS (MPa) ε fail (%) coated uncoated By comparison, the uncoated sample exhibits a decreased composite elastic modulus and significant decrease in retained ultimate strength and strain to failure. The mechanical response of the uncoated sample clearly shows far less inelastic deformation, resulting in behavior that more closely resembles the brittle failure of monolithic ceramics. This indicates that unbridged (as opposed to fiber-bridged) matrix 105

122 cracking was the dominant fracture mechanism. Strongly bonded and/or weakened fibers are unable to deflect the oncoming matrix crack and therefore the crack simply propagates through the interphase and fibers. The distinction between bridged and unbridged matrix crack growth is illustrated in Figure 31. Matrix Fiber Interphase Debond (a) (b) Figure 31: Illustration of different matrix cracking extension. (a) Fiber-bridged matrix cracking, with associated debond at fiber/matrix interphase. (b) Unbridged matrix crack propagation. Note that the interphase thickness is exaggerated to show cracking within interfacial layer. This prominent difference in damage accumulation is significant as it suggests that the uncoated sample was far more degraded than the coated sample from the HPBR exposure. As described in detail in Chapter 3, the lack of an EBC left the surface of the uncoated sample susceptible to the rapid SiC recession that occurs in high-temperature water vapor containing environments, like the one generated in the HPBR. Furthermore, if the oxidation front reaches the fiber tows and/or there exists an open pathway for the highly oxidative species to reach the fiber tows, a number of reactions can occur that ultimately result in unbridged crack growth. For instance, oxidation of the BN 106

123 fiber/matrix interphase can result in solid oxidation reaction products which can strongly bond fibers to the matrix or neighboring fibers to each other. These strongly bonded fibers generate localized stress concentrations leading to local fiber failures. Fibers that are exposed to the oxidizing environment may also experience rapid surface recession/volatilization. These degraded fibers will have a decreased cross-sectional area and therefore a decreased load carrying capability [105]. Post-test observations of the fracture surfaces of the two CMCs showed a decreased level of fiber pullout in the uncoated sample. This decreased fiber pullout is indicative of the interphase oxidation mechanism described above, and is comparable to results found in previous high-temperature testing of similar SiC/SiC CMC materials [81]. The resulting unbridged matrix crack growth resulted in severe degradation of composite strain to failure (as seen in Figure 30). It is clear therefore, that the application of the EBC to the SiC/SiC substrate greatly increased composite durability and retained mechanical properties by limiting the detrimental effects of exposure to the highly oxidizing HPBR environment ER and AE comparison The variations in electrical response and acoustic emission behavior with applied stress for the coated and uncoated samples during high temperature retained strength testing are shown in Figure 32a. It should be noted that the cumulative AE energy shown here considered only those AE events that originated from the heated gagesection of the composite, while the events occurring outside this region-of-interest have not been included. In contrast, the electrical resistance (as stated above) is measured 107

124 along nearly the entire length of the specimen. Therefore, this lack in spatial resolution in high-temperature in-situ ER should be noted as one of its current disadvantages. ER uncoated AE coated x ER coated AE uncoated (a) 108

125 ER uncoated x ER coated (b) Figure 32: High temperature retained tensile strength test: (a) comparison of in-situ electrical resistance (ER) and acoustic emission (AE), and (b) change of ER with nominal mechanical strain. As discussed in the previous chapter, AE energy accumulation is directly related to stress-dependent matrix crack development. Therefore, the clear difference in the AE energy between the two samples is once again indicative of their difference in damage accumulation and failure. The coated sample exhibits far more AE energy accumulation due to the increasing density of bridged matrix cracking with applied load that continues until composite failure. Conversely, the low level of AE energy recorded during tensile testing of the uncoated sample is suggestive of far less matrix crack accumulation to failure. The lack of the uncoated sample to generate a large number of matrix cracks is once again related the brittle failure of the uncoated sample by localized unbridged matrix crack growth. 109

126 While the ER response with applied stress follows an increasing trend for both samples, comparison of the ER measurements shows some significant differences in electrical behavior between the two, thus providing further insight into the possible differences in their respective damage morphologies. The low stress region of both tests sees very little increase in resistance due to the elastically dominated response of the material at low applied loads. In general, the coated specimen sees a gradual increase in ER that follows the nature of the cumulative AE energy quite closely. However, at elevated stresses when the slope of the AE energy curve begins to decrease (approaching matrix crack saturation), the change in ER continues increasing dramatically. This behavior is more clearly illustrated by directly plotting the ER versus cumulative AE energy, as shown in Figure 33a. The continued increase in ER with little increase in AE energy is due to the effects of matrix crack opening and associated fiber sliding that continues even after the rate of matrix crack accumulation begins decreasing. That is, although the rate of matrix cracking has decreased, the length of high resistivity fibers that are debonded from the matrix and/or bridging transverse matrix cracks continues to increase. Alternatively, the response of the uncoated sample shows very little increase in electrical resistance for the first approximately 50% of AE energy accumulation (as seen in Figure 33b). The localized damage formation of the sample, results in the electrical response at low stresses to be dominated by a small number of bridged matrix cracks separated by bonded regions of composite that accumulate unbridged matrix cracks opening at the surface (accounting for the accumulation of AE energy and the corresponding minor change in ER). At a stress just 110

127 prior to failure, the ER increased very rapidly as the inability to develop a large number of bridged matrix cracks resulted in brittle failure. This was observed as in increase in ER of over 170% during the final 20 MPa of loading. Figure 32b shows the electrical response of the samples with increasing nominal mechanical strain. The coated and uncoated samples show increases in electrical resistance of 220% and 180% over the entire specimen length, with measured increases in gage-section strain of 0.371% and 0.134% respectively. As is the case of room temperature tensile loading of SiC/SiC composite systems [56, 57], the ER increase in these tests appears to be strongly related to damage accumulation. In the case of room temperature tests, the high volume fraction of free silicon in the matrix results in a dominance in electrical conductivity of the Si-SiC matrix material over the more resistive CVI-SiC and BN coated SiC fibers. Therefore, increasing the number of matrix cracks results in smaller lengths of the more conductive undamaged composite, separated by longer lengths of highly resistive fiber-bridged transverse cracks. Such damage would, of course, act to impede the flow of electrical current (i.e. increase the overall electrical resistance) of the composite. The similar ER behavior seen in this study leads to the conclusion that the high temperature tests are controlled by the same mechanism. Recent modeling efforts have consider the CMC as an electrical circuit consisting of a series of resistors representing: bonded composites segments, debonded fiber/matrix regions and matrix cracks bridged solely by the reinforcing fibers [57]. As a result, one potential advantage of in-situ ER is that it is sensitive to both stress-dependent matrix cracking and corresponding increases in composite strain (in the form of relative 111

128 fiber/matrix displacement in the fiber/matrix debond and fiber-bridged matrix crack regions). Furthermore, the residual increase of ER that would be associated with permanent plastic deformation leads to the possibility that, unlike strain or AE monitoring, ER measurements could be used as a scheduled inspection technique. Therefore, rapid or excessive increase in the periodically measured electrical resistance of a SiC/SiC structure could be used for component lifing as it has been shown to correlate with large scale composite damage. ER coated (a) 112

129 ER uncoated (b) Figure 33: In-situ electrical resistance versus normalized cumulative AE energy (approximate normalized matrix cracking) during high temperature tensile tests of (a) coated and (b) uncoated samples AE waveform analysis The spatial distribution of recorded AE energy for the coated and uncoated specimens are shown in Figure 34a. The AE energy is plotted versus its length along the specimen (shown as ± distance from centerline of the sample), and includes all of the events recorded between the AE sensors. The ultimate fracture plane of each sample is noted by a dashed line. It is worth noting that due to the random distribution of flaws and possible sample degradation, the CMC will experience multiple matrix cracking planes prior to catastrophic failure of the sample. Therefore, it is anticipated that the AE event location analysis will show multiple areas of increased AE activity. Figure 34b shows the associated DIC strain mapping for the both samples at their respective failure stresses. The DIC image of the coated sample at peak load shows a relatively uniform 113

130 strain mapping with some small areas of elevated strain, which is in good agreement with the recorded AE data. In particular, the AE energy distribution shows an increased energy region at approximately 10 mm from the centerline correlating with the CMC fracture plane. Similarly, the high energy region near the bottom of the gage section corresponds to the localized high strain region shown in the DIC mapping. The AE energy recorded during the uncoated testing shows a similar variation in energy regions associated with damage distribution, which in turn corresponds to the wide range of strain values seen in the DIC image. The localized regions of increased strain are attributed to unbridged matrix cracking on the specimen surface. Uncoated Coated (a) 114

131 Uncoated Coated Failure Plane (b) Figure 34: Distribution of AE energy along the gage length and associated DIC strain mapping of gage (at failure stress) during high temperature tensile testing for the (a) coated and (b) uncoated sample respectively. As previously discussed, this type of damage morphology is caused by environmental degradation, leading to highly localized stress concentrations. Unlike the coated sample, matrix stresses are not being relieved by load sharing between fibers and matrix at the site of matrix crack formation. Without this load sharing mechanism, the composite exhibits a more elastic failure, similar to a monolithic ceramic material. The failure plan of the uncoated sample occurred outside the heated gage region at approximately -28 mm from the sample centerline and clearly corresponds to the largest peak in recorded AE energy. While the anticipated ultimate fracture plane is within the heated section of the sample, the fact that this specimen failed outside this region was likely caused by highly localized damage resulting from exposure to the burner-rig environment. 115

132 Further AE waveform analysis was performed to extract the frequency and energy content of the signals. Figure 35a-b shows the frequency centroid (FC) of each event recorded by both the top and bottom AE sensors. The wave frequency centroid is calculated because it gives a more accurate representation of the frequency content of the entire waveform than peak frequency alone. The FC of damage events in MI SiC/SiC laminates have been shown to be in the in the range of 600 khz 1200 khz [52]. This is clearly consistent with the FC bands seen for both of the samples tested here. However, Figure 35a shows a dense cluster of low frequency events (270 khz 375 khz), initiating at approximately 125 MPa, not seen in the uncoated sample. Similar frequency content was captured by both the top and bottom sensors throughout the test, including this low frequency cluster. Investigation of the event locations show that they are relatively well dispersed throughout the gage section (Figure 36a). Further examination of this low frequency cluster in Figure 36b, shows that in general these events are all low in relative energy content. The similarity between these events in terms of content (FC and energy) signify that they are likely related to the same type of damage mechanism. It has been shown that AE events related to surface damage generate waveforms dominated by low frequency flexural waves, with a minimal high frequency extensional component [52]. Therefore, this behavior suggests that the low frequency cluster seen in this test are possibly related to EBC cracking that initiated at the applied stress of 125MPa. The frequency content of the uncoated sample (Figure 35b) shows some slightly different behavior. The top and bottom AE sensors show very different FC values for corresponding events, especially at high stresses, where the FC calculated from the AE 116

133 waveform taken from the bottom sensor is lower than expected for transverse matrix cracking. The proximity of the bottom sensor to the fracture plane could scatter the high frequency content of events occurring closer to the gage section. (a) 117

134 (b) Figure 35: Stress-dependent AE events recorded by each sensor during high temperature monotonic tensile testing, showing calculated frequency centroid for the (a) coated sample (with low frequency cluster highlighted), and (b) uncoated sample, respectively. (a) 118

135 (b) Figure 36: AE events during high temperature testing of coated sample with (a) gagelength location of events identified in low frequency (270 khz 375 khz) cluster, (b) Average AE event energies with low frequency events highlighted. 6.5 Conclusions The capability of using ER measurements for high heat-flux tensile strength testing was demonstrated. The ER measurements used in this study proved to be a valuable tool in determining material damage initiation and accumulation. Furthermore, modal acoustic emission monitoring and digital image correlation were also used in order to investigate material damage state and localized behavior. First, the retained mechanical properties of an uncoated SiC/SiC CMC and an EBC coated SiC/SiC CMC substrate post- HPBR exposure was investigated. The decreased strength and toughness of the uncoated sample is indicative of oxidation of the composite constituents from exposure to the corrosive burner rig environment. The in-situ ER and AE data showed the differences in damage onset and accumulation between the coated and uncoated 119

136 samples. Comparison of the ER and AE data shows that while AE measurements are sensitive only to the release of elastic waves generated by cracking events, the change in ER is caused by both matrix cracking and fiber/matrix displacements. Finally, waveform analysis to extract AE energy distributions demonstrated good agreement with DIC strain mapping for determining increased damage regions. Also, comparison of AE event frequency content shows promise in determining EBC crack initiation. 120

137 CHAPTER VII COUPLED THERMOGRAPHY AND MODAL ACOUSTIC EMISSION CHARACTERIZATION OF THERMAL BARRIER COATING DAMAGE UNDER LASER HEAT-FLUX THERMAL CYCLING The desire to increase gas-turbine engine operating temperatures requires improvements in cooling design, materials and protective coatings. In the past, increases in operating temperature were achieved through improved alloy design and internal cooling-air delivered by channels cast into components. However, as the temperature capabilities of single-crystal superalloy materials reached their inherent limits, one solution for increasing maximum structural temperatures was the development of thermal barrier coatings (TBCs) that can be deposited onto the surface of existing metallic turbine structures. TBCs are designed as multilayer films of a refractory material used to protect underlying metallic substrate materials from extreme temperatures in engine hot-section gas paths. The introduction of TBCs along with internal cooling of the underlying components have resulted in a considerable reduction in superalloy surface temperature [8]. In many advanced engine applications (e.g. turbine blades and vanes) TBCs are exposed to gas temperatures above that of the melting point of the underlying metallic 121

138 component, making them a prime reliant structure. That is, damage to the TBC can result in possible component failure and serious damage to the engine itself [115]. This makes characterizing coating damage progression and predicting coating lifetime of critical importance to preventing premature engine failure. TBCs consist of several distinct layers with each layer possessing individual physical, thermal and mechanical properties. The variation in material properties of individual coating layers and the underlying substrate, along with the extreme environments to which they are exposed, can produce incredibly complex stress-states and corrosive conditions. This can result in numerous cracking sites and failure modes, presenting a significant challenge for characterization of damage mechanisms. Typical TBC systems consist of: (1) a refractory top coat for thermal protection, (2) a thermally grown oxide (TGO) reaction layer which is generated between the top coat and subsequent layer during processing and thermal exposure, (3) a bond coat layer that supplies the oxidation protection and adhesion to the underlying substrate, as well as supplying the elements necessary to form the TGO layer, and of course (4) the metallic substrate material. Currently, one of the most widely utilized TBC top coats consists of some form of yttria-stabilized zirconia (YSZ), in particular the metastable tetragonal-prime structure [10, 11]. YSZ was initially selected as a TBC candidate primarily due to its relatively low thermal conductivity and good coefficient of thermal expansion (CTE) match with nickel and cobalt-based superalloys. Typical YSZ bond coats are comprised of a MCrAlY alloy, where M = Ni, Co, Fe or their combination (depending on the composition of the substrate material). 122

139 Because TBCs are expected to withstand the extreme temperatures and thermal cycling associated with typical turbine engine use, a critical design concern for these materials is high temperature thermal fatigue resistance. During regular thermal cyclic engine operation, several factors lead to the cracking and ultimate spallation failure of ceramic coatings. Of major concern is the development of large thermal strains within the TBC and at the coating/substrate interface. As these thermal stresses are greatly influenced by thermal gradients and transient thermal loading (rapid heating and cooling), it is crucial that testing of TBCs to assess life or damage mechanisms attempt to simulate the high heat-flux environments generated in engine environments. Since such conditions are impossible to reproduce in typical furnace-based thermal cycling experiments, a specialized high heat-flux laser rig at NASA Glenn Research Center (Cleveland, OH) has been developed to evaluate fundamental coating properties such as sintering and creep rates, thermal conductivity change and kinetics, and evaluation of coating durability and lifetimes [69-73, ]. Testing has shown that the high heatfluxes (and thermal gradients) developed via this laser-based testing greatly increase the driving force of many coating damage mechanisms as compared to a similar number of furnace-based cycles [73, 117]. Along with high heat-fluxes, thermal stresses can be further increased at areas of geometric stress concentrations in the underlying metallic structure and along structural joints. These localized areas of increased stresses are highly susceptible to the initiation of delamination cracks at the TBC/substrate interface. Upon transient heating and cooling these delaminations can propagate and join, leading to large scale spallation of 123

140 the TBC. Figure 37a illustrates the typical wedge-shaped vertical through thickness coating cracks that can be initiated by sintering effects or via local overstressing, as well as horizontal delamination cracks that can develop at coating layer interfaces [70]. (a) (b) Figure 37: SEM micrographs of coating cross-section depicting typical EB-PVD TBC coating crack morphologies. (a) Wedge-shaped through thickness vertical cracking and horizontal delamination. (b) Densified columns and delamination cracking caused by CMAS degradation. 124

141 Another source of premature coating failure has been observed in TBC systems that have been exposed to calcium-magnesium aluminosilicate (CMAS) via ingestion of materials such as sand, volcanic ash, etc. into engine hot-sections [119]. This CMAS debris becomes molten at high temperatures and can infiltrate the TBC microstructure. It has been shown in a number of studies that CMAS infiltration can be deleterious to ceramic coatings by both thermomechanical and thermochemical processes [77-79]. Recent aero-turbine engine components removed from service have revealed excessive coating degradation by CMAS infiltration [119]. Figure 37b shows the typical densification of EB-PVB coating columnar structure due to CMAS infiltration and delamination cracks that can develop within the coating at the depth of the molten CMAS infiltration [126]. Due to the level of complexity described, discrimination of cracking modes and quantification of damage accumulation via traditional experimental means becomes extremely difficult. Therefore, the development of in-situ monitoring techniques becomes crucial in providing understanding of various failure mechanisms and the improvement of predictive models. Over the years a number of non-destructive evaluation and health monitoring methods have been used in TBC testing. Among these, the monitoring of thermal conductivity evolution via thermography measurements has proved very effective in understanding damage progression during high heat-flux thermal cyclic testing [70, 73, 116]. This is due to the fact that thermal transport properties (e.g. thermal conductivity, thermal diffusivity) are closely related to coating microstructure and adhesion. Unlike isothermal (furnace-based) cyclic testing, by 125

142 utilizing thermography equipment it is possible to monitoring changes in throughthickness thermal gradient during heat-flux testing caused by coating changes that will affect the heat transfer properties through the system. While real-time thermal conductivity measurements have been used to observe sintering and cracking behavior under thermal cyclic testing of TBCs, such measurements alone fall short in directly quantifying the total degree of cracking and delamination. In order to overcome these limitations, the present work includes the use in-situ acoustic emission monitoring. Acoustic emission (AE) monitoring has been used to successfully monitor damage in a variety of material systems and loading conditions. Transient acoustic signals are generated when the fracture energy of solids, associated with material damage, are rapidly released in the form elastic stress waves. AE monitoring techniques can therefore be used to detect the sound emitted during various forms of multiscale inelastic deformation and give information regarding the initiation and accumulation of material damage. The fracture behavior of different TBC systems under thermal and mechanical loads have been qualitatively investigated using tradition AE (TAE) monitoring techniques [ ]. However, the TAE method only captures certain aspects of the acoustic signal (often referred to as AE parameters or features) which include number of AE counts, peak amplitude, etc. [40]. However, the modal acoustic emission (MAE) technique used in this study captures the entire waveform of the AE event allowing not only for determination of when the event occurred, but also AE event location as well as post-test analysis of waveform energy 126

143 and frequency content. This analysis can be used to determine damage accumulation and possibly discriminate between damage types and/or modes. The purpose of this study is to investigate thermal cyclic damage of TBC systems under the temperature and stress gradients that simulate conditions of advanced turbine engines. In particular, testing included comparison of an as-deposited coating to systems experiencing thermal stress concentrations from geometric variations in the metallic substrate as well as pre-test infiltration of molten CMAS deposited on the coating surface. Coating damage accumulation and probable mechanisms are assessed via coupled thermography and modal AE response observed under laser heat-flux cyclic testing. 7.1 Experimental Materials Three separate test coupons were produced in order to compare potential differences in damage accumulation under thermal cyclic loading conditions. The substrate materials for all of the samples consisted of Haynes 188 alloy (a high temperature Co-Ni-Cr-W alloy used primarily in aero-engine applications), with nominal dimensions of 102 mm x 19mm x 1.68 mm (L x w x t). The coating system used for this testing consisted of a ZrO2-7 wt.% Y2O3 (7YSZ) top coat and NiCrAlY bond coat deposited via electron beam-physical vapor deposition (EB-PVD) onto the sample substrate surface. The coating was deposited across the center 50.8 mm span of the substrate surface. Sample 1 is considered as a baseline sample consisting of the as-deposited coating on the metallic substrate. The substrate used for Sample 2 has two partial through holes 127

144 (not fully penetrated) machined from the back face of the substrate to generate thermally induced stress concentrations during testing. The 1.59 mm diameter (1/16 ) hole-centers are offset from the vertical centerline by 12.7 mm and from the horizontal centerline by ± 4.76 mm (0.5 and ± 3/16 respectively). These stress concentrations will initiate controlled coating delamination cracking during laser thermal cycles. Finally, the coating on Sample 3 was subjected to pre-test molten CMAS infiltration in order to quantify any possible differences in damage accumulation with that of the pristine (asdeposited) coating. The CMAS was deposited as a powder on the coating surface, then the sample was heated for 12 hours at coating surface and substrate back side temperatures of 1360 C and 955 C respectively. 7.2 Experimental Procedure A specially designed laser rig is used to test the TBC specimens under high temperature, high heat-flux thermal cyclic loading. This technique utilizes the same modified thermal conductivity rig described in Chapter 4, consisting of a high power (3.5 kw) CO2 laser to heat the TBC top coat surface thereby generating a through-thickness thermal gradient that can be increased by the addition of active air-cooling delivered to the back side of the metallic substrate. A description of the laser test rig system and the general approaches to testing the TBCs have been described elsewhere [70, ]. A steel aperture plate containing a 44 mm diameter circular opening is used to assure a consistent laser-heated area. This setup is capable of delivering a very stable beam profile, leading to a near constant temperature along the surface of the heated-region [70]. In-situ thermography measurements were taken in order to monitor the TBC 128

145 surface (Tsur) and substrate back side (Tback) temperatures of the heated-region by means of an 8μm and two-color infrared pyrometer respectively. For each test, the laser power was set in order to generate initial TBC surface and substrate back side temperatures of approximately 1475 C and 1100 C respectively. The laser was programmed for cyclic loading of: (1) a 15 second heat-up, followed by (2) a 1 hour hightemperature hold period (constant laser power), and (3) a 60 second cooldown. The temperatures at first cycle hold and the number of cycles each sample was exposed to during testing in listed in Table 9. Table 9: Details of EB-PVD 7YSZ coated samples used in laser heat-flux thermal cyclic testing. Specimen Description Initial Temperature ( C) Tsur/Tback # of Thermal Cycles (1 hr. hold time/cycle) Sample 1 As-deposited 1477/ Sample 2 Through-thickness hole in substrate 1475/ Sample 3 CMAS infiltrated 1475/ Acoustic emission events were monitored using a Digital Wave Fracture Detector equipped with two high-temperature Physical Acoustics S9215 sensors (50 khz 650 khz) attached to the back side of the metallic substrate (approximately ±40mm from sample centerline), and data was acquired at the rate of 10MHz when triggered. Hightemperature AE sensors were used in order to remove the need for waveguides in order to contact the specimen. The use of waveguides is often a limitation for AE waveform analysis as they can further attenuate the generated AE waveform. The AE sensors were clamped to the back side of the substrate ends at ± 40 mm from the sample centerline. 129

146 A schematic of the laser heating test setup, along with AE sensor configuration is shown in Figure 38. CO 2 laser beam IR Pyrometer Laser Aperture 7 YSZ TBC System Metallic substrate Heated Region Modal AE Sensors (± 40mm) IR Pyrometer Figure 38: Schematic of laser heat-flux testing apparatus used for characterization of thermal cyclic behavior of EB-PVD 7YSZ coated specimens. Including details of high temperature modal AE measurement setup. 7.3 Results and Discussion Using the experimental setup shown in Figure 38, thermal cyclic testing was performed on each of the specimens described above. The font and back side temperatures of the specimens were continuously monitored along with in-situ monitoring of AE events by the attached sensors. From the collected thermography 130

147 data, the change in overall through thickness temperature across the specimen (TBC and substrate) can be calculated as: T = T sur T back (17) Changes in the apparent thermal gradient across the samples are indicative of microstructural changes within the coating systems and/or delamination and spallation of the coating itself [70-73, ]. Furthermore, because the amount of AE energy released with time is a measure of damage accumulation within the TBC system, it is meaningful to compare the cumulative AE energy over the same time scale. In this way it is possible to compare changes in thermal properties with the release of damage events in the form of recorded AE energy. The temperature data in Figure 39a shows the evolution of front and back side temperature with number of thermal cycles of the as-deposited TBC system (Sample 1). The coating demonstrated relatively good thermal stability over the entirety of the thermal cyclic test. While the TBC surface temperature showed a marginal net increase after 100 cycles, there are some clear fluctuations at various points throughout the test. Furthermore there was a noticeable net increase in the substrate back side temperature that appeared to be dominated by the increase over the initial cycles. Plotting the temperature data as the apparent through thickness thermal gradient T further facilitates the observation of trends in the change in thermal transport properties with successive thermal cycling. The thermal gradient versus time plotted in Figure 39b shows a number of distinct regions. It is clear that the largest change occurs over the first approximately 10 thermal cycles. This noticeable decrease in the overall 131

148 thermal gradient was due to fluctuations in both surface temperatures and the significant increase in the back side surface temperature. This behavior is followed by a lower, continuous rate of decreasing T from cycles 10 to 30. Finally, a small jump in T around cycle 30 is followed by a slightly decreasing monotonic behavior with another distinct fluctuation around cycle 90 (seen as another minor shift in the TBC surface temperature). The overall thermal conductivity increase of the system observed over the entirety of the laser thermal cyclic testing is attributed to the sintering effects of both the intracolumnar and inter-columnar micro-porosity of the EB-PVD coating [70]. The featherlike fingers of each column can become appreciably sintered at high temperatures, decreasing the overall porosity within the column and in turn increasing the overall thermal conductivity. Furthermore, adjacent columns can also become bonded together resulting in a similar increase in thermal conductivity of the TBC system. Coating sintering however not only leads to changes in thermal transport properties, but can also result in coating crack initiation. It has been shown that the through thickness thermal gradients produced under laser heat-flux testing can result in variable sintering strains that are higher near the coating surface than at the TBC/substrate interface. These strains can lead to the formation of vertical wedge-shaped cracks that initiate at the coating surface like those seen in Figure 37a. The increased thermal stresses produced during transient heat/cooling phases causes propagation of these vertical cracks and can eventually result in minor delamination crack initiation within the coating or at the coating/substrate interface [70]. However, delamination cracking leads to 132

149 substantial decreases in thermal conductivity which was not observed in the temperature data recorded during Sample 1 testing. Hence, no large scale delamination events occurred. This was confirmed by visual inspection of the TBC surface post-test. T sur Laser Power T back (a) 133

150 Total AE Heat-up AE Hold AE Cool-down AE ΔT (b) Figure 39: Data collected from the thermal cyclic testing of the as-deposited TBC sample (Sample 1). (a) Thermography data of TBC surface and substrate back side temperatures and applied laser power. (b) Evolution in overall thermal gradient (ΔT) with corresponding accumulation of AE energy (total AE energy and breakdown of contribution of each phase of the thermal cycle). The release of AE energy during testing was also monitored as the accumulation of AE events correlate to inelastic damage accumulation of the thermal barrier coating. In order to compare the observed changes in thermography measurements with quantifiable damage that occurs during laser heat-flux testing, the total cumulative AE energy is plotted with the thermal gradient data in Figure 39b. To further understand the specific driving force behind the TBC damage mechanisms, the total AE energy data is separated based on when during the thermal cycle the event occurred: (1) heat up, (2) hold period, (3) cool-down. The first 10 thermal cycles show a rapid increase in the AE energy accumulation during the high-temperature dwell periods and during cool-downs. 134

151 This initial AE data is therefore completely consistent with the observations made from the thermography data, where the increase in overall through thickness thermal conductivity can be attributed to high temperature coating sintering. Sintering occurs during the high temperature hold periods, which in turn initiates vertical microcracking in the coating surface. These microcracks propagate during periods of transient thermal loading, resulting in the corresponding increase in measured AE energy. After these initial cycles, the AE energy increases at a much slower rate with jumps that correlate to clusters of higher energy events that occur during heating and cooling periods. In particular the rapid increases in AE energy around cycles 40 and 90 that corresponds to the perturbations seen in the thermography measurements. In general, the accumulation in AE activity shown in Figure 39b appears to be in excellent agreement with the observations made of the couple thermography data of Sample 1. A second laser heat-flux thermal cyclic test was performed on a sample with the same TBC system as Sample 1, but the substrate contained a pair of partial through holes that act as a thermal stress concentration in order to drive coating delamination behavior at a known location (denoted as Sample 2). The thermography data and corresponding AE energy recorded during testing for Sample 2 are shown in Figures 40ab. The first few cycles of the temperature measurements in Figure 40a show a slight increase in TBC surface temperature with a considerably larger increase in substrate back side temperature similar to the effect of high-temperature sintering seen in Sample 1. This initial behavior is followed by a sudden and dramatic increase in TBC surface temperature and corresponding decrease in substrate temperature. This 135

152 behavior indicates a rapid decrease in overall thermal conductivity by the initiation of a delamination crack occurring at the interface of the coating layers and/or at the coating/substrate interface [70, 73]. It should be noted that this behavior is in contrast to the observations made during high temperature sintering, where the increased coating densification resulting in an overall increase in the apparent through thickness thermal conductivity. Furthermore, vertical sintering cracks at the coating surface are parallel to the heat-flux direction and therefore have a much smaller effect on through thickness thermal properties. T sur Laser Power T back (a) 136

153 Total AE Heat-up AE Hold AE Cool-down AE ΔT (b) Figure 40: Thermal cyclic testing of the sample with substrate containing partial through holes (Sample 2). (a) Thermography data of TBC surface and substrate back side temperatures and applied laser power. (b) Evolution in overall thermal gradient (ΔT) with corresponding accumulation of AE energy (total AE energy and breakdown of contribution of each phase of the thermal cycle). The initial behavior is followed by a region (cycles 10-70) of gradually decreasing thermal conductivity, observed mainly by a continuously decreasing substrate back side temperature. The decrease in measured back side temperature is due to propagation of the coating delamination with successive thermal cycling. This behavior continues until another large change in coating surface temperature occurs following the 70th thermal cycle. This rapid change was caused by a major coating spallation event, which post-test inspection found resulted in complete loss of a section of the top coat. The coating failure lead to plastic deformation of the substrate in the form of localized buckling in the region surrounding the partial through holes. 137

154 Comparison of the cumulative AE energy data to the evolution of the throughthickness thermal gradient in Figure 40b allows for further investigation of coating damage mechanisms. Once again it is apparent that the accumulation of AE energy is in excellent agreement with the change in through thickness thermal properties of the system. The increase in total AE energy associated with early delamination cracking is dominated by the rapid increase in AE energy released during cooling and heating phases. This is because coating delamination propagation is driven by shear forces that develop upon transient heating/cooling due to the CTE mismatch between constituents [74]. It is worth noting that the accumulation of AE energy generated by this initial delamination cracking resulted in a total energy increase several times that of the entire Sample 1 test. The proceeding region shows a gradual increase in AE energy as delamination cracks continue to propagate during heating and cooling periods. Finally, another large increase in energy is recorded that corresponds with the final spallation event that lead to coating failure. The total AE energy accumulated by the conclusion of Sample 2 testing was an order of magnitude higher than the total energy recorded during a similar number of thermal cycles in Sample 1 testing. Therefore, the amount of AE energy accumulation correlates to the severity of coating damage and type. One common disadvantage of many damage monitoring techniques, including insitu thermal conductivity measurements, is the lack of spatial resolution necessary to accurately determine the origin of damage activity. However, the event-based modal AE used in this study collects the waves from each of the AE sensors used simultaneously. This allows for synchronization between the testing channels for every damage event. 138

155 By comparing the times of arrival of the waveforms captured on each sensor it is possible to determine the location of each AE event. The use of the partial through holes in Sample 2 to promote coating delamination under thermal cycling presents the opportunity to correlate specific AE events with a known damage mechanisms as well as to investigate the use of AE event location analysis. Additionally, the a priori knowledge of where the major delamination events will occur allows for the verification of the accuracy of said location analysis. The location of a given AE event is using Eq. 16, where the AE wave velocity was determined prior to high-temperature tensile testing by performing pencil lead breaks at the ends of the specimen outside the sensor span in order to manually generate AE sources [123]. The distribution of recorded AE energy along the length of Sample 2 is shown in Figure 41a. Heat-up AE Hold AE Cool-down AE (a) 139

156 Large-scale spallation of top coat (b) Location from centerline (mm) Figure 41: (a) Histogram of Sample 2 testing showing recorded AE energy versus location (± mm from center of sample). (b) Image illustrating coating spallation area event that occurred during thermal cyclic testing of Sample 2. The histogram of energy distribution shows that the highest concentration of energy was released between the center of the sample and the far end of partial through holes. Furthermore, the vast majority of the AE energy recorded in this area occurred during the cooling phases of the thermal cycles. This suggests that the delamination cracking initiated in the vicinity of the stress concentration and was propagated along the length of the sample via the increased thermal stresses caused by CTE mismatch between coating and substrate that occur during transient phases of thermal cycling. This is consistent with typical EB-PVD TBC failure under thermal cyclic conditions which is generally controlled by successive crack nucleation and propagation events that occur at the top/bond coat or TGO interface [74]. As noted previously, this behavior was confirmed by post-test inspection which found that this delamination ultimately resulted in full spallation of the coating around the partial through holes. The large spallation area seen in Figure 41b extends from the center of the sample to approximately 11.5 mm from the centerline. 140

157 Recent observations of aero-turbine engine components removed from service have revealed the existence of protective coating damage by molten CMAS in engine hotsections [119]. The porous columnar structure of EB-PVD coatings, utilized in order to increase strain tolerance, is unfortunately quite susceptible to molten CMAS infiltration. At temperatures above 1200 C, CMAS debris becomes molten and can partially infiltrate the TBC microstructure via capillary forces. The molten CMAS will penetrate the high temperature TBC surface until it reaches a low enough temperature within the coating to raise the viscosity enough to cease penetration [74, 119]. CMAS infiltration results in an increase is densification and overall stiffness of the coating. This causes a decrease in the strain tolerance of EB-PVD coating which may eventually lead to delaminations within the top coat (corresponding to the depth of CMAS infiltration) as seen in Figure 37b [74, 126]. The decrease in porosity and changes in sintering mechanisms can also adversely affect the thermal conductivity of the TBC, reducing its insulating effectiveness [79]. Furthermore, CMAS infiltration has been shown to dramatically change the chemical properties of the YSZ top coat, including depletion of yttrium and destabilization of the t -zirconia phase [79, 124]. The final sample in this study was prepared in order to investigate the effect of CMAS on TBC damage tolerance and mechanisms. Sample 3 consisted of the same EB- PVD 7YSZ system as the previous tests, however, prior to laser heat-flux thermal cyclic testing the sample was exposed to high temperature CMAS infiltration. This was done by first depositing a layer of CMAS powder across the surface of the TBC. The sample was then heated to TBC surface and substrate back side temperatures of 1360 C and 141

158 950 C respectively for a total of 12 hours. After molten CMAS infiltration, the sample was loaded into the laser thermal cyclic rig and tested for approximately 100 cycles. The results of the thermography measurements and corresponding cumulative AE energy are shown in Figures 42a and 42b. The substrate back side temperature data shows a minor decrease over the initial cycling, followed by a steady increase over the remainder of testing. The coating surface measurement demonstrated a monotonic increase in temperature over the entirety of the test. These increasing temperatures are attributed to coating cracks within the CMAS densified region of the TBC top coat. At no point was a large variation in temperature observed that would indicate a large-scale spallation event or coating failure. A comparison between the overall thermal gradient and AE energy accumulation, shown in Figure 42b, provides further insight into the mechanisms driving the changes in thermal properties observed during cyclic testing of Sample 3. T sur T back (a) 142

159 Total AE Heat-up AE Hold AE Cool-down AE ΔT (b) Figure 42: Thermal cyclic testing of the sample exposed to pre-test CMAS infiltration (Sample 3). (a) Thermography data of TBC surface and substrate back side temperatures and applied laser power. (b) Evolution in overall thermal gradient (ΔT) with corresponding accumulation of AE energy (total AE energy and breakdown of contribution of each phase of the thermal cycle). The nature of the change in through thickness thermal gradient and AE energy accumulation shows many distinct differences in thermal cyclic behavior to the asdeposited coating testing of Sample 1. The early cycles of Sample 1 were dominated by high temperature sintering evident by the resulting increase in thermal conductivity caused by densification of the porous EB-PVD coating. However, it is clear from the lack of thermal conductivity increase and rapid AE energy accumulation in the early cycles of Figure 42b that Sample 3 did not experience the same initial sintering behavior as Sample 1. The difference in sintering behavior is therefore a result of the infiltration of the inherent columnar micro-porosity of the EB-PVD top coat to the depth of the CMAS 143

160 penetration. Furthermore, the remaining region unaffected by CMAS likely experienced some level of pre-test sintering during the high temperature molten CMAS infiltration step. The accumulation of AE energy throughout Sample 3 testing was clearly dominated by AE events occurring during cooling cycles. There is little observable increase in AE activity up to cycle 30, at which point the accumulation of AE energy begins to increase at a much faster rate. These AE events are associated with coating crack initiation and propagation within the CMAS densified region of the top coat. The reduction in coating strain tolerance leads to damage accumulation caused by the large thermal stresses generated at the coating surface during the specimen cooling phase. These is another observable increase in the rate of AE energy accumulation that begins after approximately 80 hours of testing. This could be associated with the initiation of delaminations within the CMAS infiltrated top coat, however the lack of any rapid jumps in the through thickness temperature gradients suggests a lack of any large scale spallation. A fact that was confirmed upon post-test visual inspection of the sample. Further AE analysis of all three samples was performed in order to investigate the possibility of using the current experimental setup for distinguishing between damage events on the basis of AE waveform frequency content. The frequency content of AE signals can be quite complex, therefore the wave frequency centroid (weighted frequency average) of each AE event is determined. Frequency centroid (FC) accounts for the entire frequency spectrum and is therefore less sensitive than peak frequency analysis (determined from a single point in the frequency range) to minor variations in the overall frequency spectrum. The FC values calculated for each event from the 144

161 respective samples is shown in Figures 43a-c. The events from each test have been separated according to which phase of the thermal cycle they were recorded and the size of the markers have been normalized in order to represent the measured energy of each AE event. Heat-up AE Hold AE Cool-down AE (a) 145

162 Heat-up AE Hold AE Cool-down AE (b) Heat-up AE Hold AE Cool-down AE (c) Figure 43: Calculated frequency centroid of AE events, separated by which phase they occurred. Note that the size of the bubble is representative of the relative energy of the event. (a) Sample 1, (b) Sample 2, (c) Sample

163 By comparing the frequency content of AE events between Sample 1 and Sample 2 (Figures 43a and 43b) it can be observed that both tests exhibited very similar values of frequency centroid. Furthermore, the events occurring upon the transient heating/cooling and high temperature hold phases all appear to fall within the same narrow frequency band. Not only does the frequency content between phases cover the same band, but the frequency content appears to be independent of AE energy content as well. This is unfortunate as it drastically limits the ability of using AE event waveform frequency content as a means of distinguishing between the vertical coating cracking characterized by Sample 1 and the TBC/substrate delamination cracking detected during Sample 2 testing. This is likely due to how the stress waves generated by the damage events are transmitted to the AE sensors themselves. That is, coating damage events occur along the substrate surface and propagate through the substrate as AE waveforms that are dominated by a low frequency flexural wave component. The frequency content of the AE events recorded during thermal cycling of the Sample 3 (which experienced pre-test molten CMAS infiltration) exhibit an appreciably higher frequency band than the other samples. This is likely due to the changes in coating microstructure and increased stiffness caused by CMAS exposure. 7.4 Conclusions Laser heat-flux thermal cyclic tests were performed on a series of EB-PVD 7YSZ coated, superalloy substrates. The objective of testing was to investigate the difference in coating behavior of an as-deposited TBC to a sample containing thermal stress concentrations and molten CMAS infiltration. Couple in-situ thermography 147

164 measurements and modal acoustic emission monitoring was utilized in order to illuminate the damage evolution of the individual samples. The as-deposited coating exhibited good thermal stability with thermography measurements indicated an initial increase in apparent through thickness thermal conductivity as a result of high temperature coating sintering. Sintering of the TBC coating generated vertical surface cracking that was detected using modal acoustic emission monitoring. The results of the AE analysis showed little AE energy accumulation that was driven by initial AE events measured at high temperature hold periods and propagation of cracking during transient heating and cooling periods. Large scale coating delamination was not observed by either the thermography of AE measurements. A summary of some of the AE analysis results discussed in this study are listed below in Table 10. Table 10: Summary of results of AE analysis from thermal cyclic testing. Total AE Events AE Energy Avg. FC Hold Total Cum. Hold Hold Heatup Cooldown Heatup Cooldown Heatup Cooldown (#) (%) (%) (%) (V 2 μs) (%) (%) (%) (khz) (khz) (khz) Sample Sample Sample Sample 2 contained stress concentrations in the form of partial through holes in the underlying metallic substrate. The holes were capable of initiating delamination cracking between the TBC and substrate. These delaminations were evident in the thermography data which showed a rapid decrease in thermal conductivity. The AE data was in excellent agreement with the thermography observations and large increasing in AE 148

165 energy were detected during delamination initiation and subsequent propagation. In fact, the total AE energy recorded over the course of Sample 2 testing was an order of magnitude higher than that of Sample 1 indicating the severity of coating damage. Further analysis of the AE event waveforms was performed to produce an AE energy histogram of event location. The AE location analysis confirmed that the majority of AE energy was recorded in the vicinity of the stress concentrations, demonstrating the effectiveness of using modal AE monitoring to accurately determine the location of coating damage events. The final sample used in this study was exposed to molten CMAS infiltration prior to thermal cyclic testing. The resulting coating behavior was in contrast to the as-deposited coating response observed during Sample 1 testing. Sample 3 exhibited obvious differences in initial high temperature behavior and AE response. It was determined that the pre-test infiltration of CMAS changed the sintering behavior of the as-deposited EB- PVD 7YSZ coating. Successive thermal cycles resulted in increasing coating damage accumulation during transient cooling phases, observed as increasing AE energy accumulation and TBC surface temperatures. Furthermore, while the total number of AE events recorded during Sample 3 testing was similar to Sample 1, the total energy of the Sample 3 AE events was nearly four times that of Sample 1. This is significant as it quantifies the level of increased coating degradation and life limiting effects of CMAS infiltration. 149

166 CHAPTER VIII CREEP AND STRESS-RUPTURE OF CMCS AND EBC/CMC SYSTEMS UNDER HIGH HEAT- FLUX, THERMAL GRADIENTS As mentioned, SiC/SiC composite systems are among the leading candidate materials for aero-turbine engine structures in which their high temperature capabilities and resistance to oxidizing environments becomes a critical concern for component health. High temperature stress-rupture of CMCs can be quite complex making lifing of SiC/SiC components a significant challenge. It is critical therefore to understand the damage in these materials due to applied stress and extreme environment in order to understand the mechanisms that lead to degradation and ultimate creep rupture life. In general the key concern for component life under high temperature stressed-oxidation conditions are the stability of the CMC constituents (fibers, matrix, interphase) and the durability of the environmental barrier coatings (EBCs) necessary to prevent premature composite failure. Creep-related failures of these materials can be dominated by several different mechanisms depending on temperature, stress and environment. Therefore the ability to design health monitoring and inspection techniques to determine remaining life would prove very beneficial for modeling and design. 150

167 Previous studies have demonstrated the creep resistance and stress-rupture life of MI-CVI SiCf/SiC composites up to 1315 C at moderately high stresses under isothermal heating conditions [106, 107]. As previously discussed, many EBC/CMC engine components will be subjected to high heat-flux thermal gradients and complex stresses. Therefore, tradition isothermal tensile creep testing of CMCs is incapable of adequately simulating these types of engine environments. Furthermore, because degradation mechanisms leading to stress-rupture are dependent on applied mechanical loading, environment, temperature conditions and time it is necessary to consider each condition and in depth in order to fully understand the degradation response of the each material. In this chapter, a case study is presented in which three separate high heat-flux tensile creep tests were performed using the novel laser-based thermomechanical testing apparatus described in Chapter 5. Three representative tests are presented consisting of two uncoated ZMI fiber reinforced CMCs (ZMI-1 and ZMI-2) and one environmental barrier coated CMC substrate (ZMI-3). The first specimen is used as a baseline in order to characterize tensile stress-rupture of these CMC systems under constant load and high heat-flux. The second uncoated specimen was preloaded beyond the matrix cracking strength of the composite and subsequently unloaded prior to testing. Not only does the matrix of this composite have reduced load carrying capability, but the matrix cracks allow for an ingress of oxidizing species to the interior of the composite which could lead to fiber/matrix degradation. Therefore, this precracked sample is used to investigate possible differences in mechanical response of a 151

168 pre-damaged sample to that of an as-produced. The third sample was coated with and EBC and heated in order to produce a similar substrate temperature as the other samples and is subjected to the same constant tensile load. The main objective of testing this sample will be to investigate the possible contributions of EBCs on the overall stress-rupture life of CMCs under high heat-flux conditions. One of the major focuses of this study will be to demonstrate the feasibility of using ER measurements to characterize damage and monitor strain accumulation under high heat-flux thermal gradient creep conditions. The previous chapters in this report have been structured in such a way as to illuminate the complex nature of using ER measurements for high temperature testing of CMCs. The unique set of tests presented in this chapter will investigate the combined contributions of thermal and mechanical loading as well as time-dependent and environmental effects on the on the overall electrical response of the specimen. 8.1 Experimental Materials The ceramic composite samples tested in this study consist of Tyranno ZMI fiberreinforced, slurry cast, melt-infiltrated SiC/BN/SiC in 8 plies of balanced 0 /90, 5 harness-satin weave manufactured by the Goodrich Corporation (Brecksville, OH, now part of United Technologies). It should be noted that these samples are of the same architecture and were produced by the same manufacturer as the ZMI reinforced samples described previously in Chapter 4. The composite panels were machined into 152 mm (6 in) tensile bars consisting of a dog-bone shaped geometry. One of the tensile coupons was then coating with a NASA EBC deposited via electron beam physical vapor 152

169 deposition (EB-PVD). State-of-the-art rare earth (RE) silicate coatings such as this are currently being investigated not only for their superior performance for environmental protection of Si-based components in high temperature oxidizing environments but their inherent low thermal conductivity can provide thermal protection as well, leading to increased operating temperatures of the component. The coating was deposited on a single face of the specimen, leaving the opposing face of the CMC substrate exposed/uncoated. Details of the composites and coating system can be found in Table 11. Table 11: Geometric properties of ZMI fiber reinforced SiCf/SiC test specimen gage sections, and EBC specifications ID Width (mm) Thickness (mm) vf0 Bond Coat EBC Configuration Top Coat Deposit ZMI NA NA NA ZMI NA NA NA ZMI HfO2+Si Yb2Si2O7 (127 µm) + Hf-RE silicate (254 µm) EB-PVD 8.2 Experimental Procedure The same laser high heat-flux testing rig described in Chapter 5 is utilized in this study to assess tensile creep and stress-rupture properties. Once again, the front side Tsur and back side Tback temperatures of the heated zone are monitored using a pair of infrared pyrometers, where the front side refers to either the coating surface or the substrate surface (depending on whether an EBC has been applied), while the back side is on the opposing, non-heated substrate face. Heating of the specimens is carried out under a no load condition in which free thermal expansion in the longitudinal direction is allowed in order to prevent excessive thermal stresses. For this study, the 153

170 samples were heated in order to achieve similar CMC back side temperatures of approximately 1000 C. This was done in an effort to produce comparable thermal gradients across both the coated and uncoated SiCf/SiC substrates. Mechanical loading of the specimen is conducted at a rate of 0.127mm/min. Elongation of the laser-heated section is monitored with the use of an Instron high-temperature extensometer with a 25.4 mm gage section and 2% travel. All of the creep tests in this study were performed under load control resulting in a constant stress of 69MPa. Figure 44 shows mechanical data from unload/reload tensile tests of ZMI fiber reinforced dog-bone samples from the same panels as the specimens in this test. ZMI-1 and ZMI-2 were from panel A and ZMI-3 was from panel B. Note that the composite creep stress chosen for this study of 69 MPa is below the room temperature proportional limit of both panels. Figure 44: Room temperature mechanical response of tensile specimens from the same composite panels as the ones used in this study. ZMI-1 and 2 are from panel A, and ZMI- 3 is from panel B. 154

171 If a sample is capable of surviving 500 hours under these conditions a monotonic strength test is performed in order to determine retained tensile properties. The same high-temperature ER measurement setup as shown in Chapter 5 has been implemented for monitoring the electrical response during testing, however due to the length of these tests, AE and DIC were not implemented. 8.3 Results and Discussion The results of this study will be presented in terms of the thermo-mechanical and electrical behavior of the composite specimens during high temperature tensile creep testing. This will be followed by post-test microscopy that will be used to further investigate difference in material response Tensile Creep and Electrical Resistance A series of high temperature thermal gradient creep tests were performed in order to investigate time-dependent effects of creep and stressed oxidation on electrical resistance. Three representative tests were performed consisting of two uncoated ZMI fiber reinforced CMCs (ZMI-1, ZMI-2) and one coated ZMI substrate (ZMI-3), the details of which can be found in Table 11. The heating, loading and electrical response for ZMI- 1 can be seen in Figure 45a. As previously mentioned, during specimen heating the load frame is programmed to a no load condition (i.e. the thermal strain is accommodated in the load direction in order to reduce large scale thermal stress development). However, since the specimen ends are clamped into the load frame preventing bending, the application of a through thickness thermal gradient will induce some thermal stresses. 155

172 The time history in Figure 45a shows the ER response to heating and the subsequent mechanical loading to the applied composite stress of 69 MPa. Note that the jump in Tback prior to loading is due to an adjustment of the infrared pyrometer used to measure the back side CMC temperature. A large increase in electrical resistance (~70%) from room temperature to the front and back side gage temperatures of 1138 C and 994 C was observed. As discussed in Chapter 4 this ER increase is dominated by the temperature dependent nature of electrical resistivity, with an additional minor increase associated with the 0.4% elongation due to thermal strain observed upon heat up. A closer look at the specimen strain and electrical response due to the application of mechanical stress is shown in Figure 45b. The composite strain behavior is plotted as a total strain (creep strain plus strain due to loading) versus time. The specimen experienced a 0.089% strain increase due to loading followed by an additional 0.1% strain to failure via stress-rupture over the approximately 83 hours of test time. While it is unlikely that the specimen ever attained a true steady-state creep rate, a noticeable decrease in strain rate was observed following the initial primary creep response. Because the applied stress is below the proportional limit for this composite material [47] it is possibly that some microcracking was initiated in the outermost plys, but very unlikely that any fully bridged transverse matrix cracking developed upon loading. The rapid increase of composite tensile strain over the first couple of hours is therefore likely a result of time-dependent stress relaxation of the matrix material. That is, the primary creep region of ZMI-1 is typical of MI-CVI SiCf/SiC composites at low creep stresses in that the behavior was dominated by the increasing load transfer from the 156

173 relaxing MI matrix to the more creep resistant reinforcing fibers and (to possibly a more limited extent) the CVI-SiC matrix material [106, 107]. These conclusions are supported by observations that can be made about the ER response shown in Figure 45b as well. Here the electrical behavior is plotted using Eq. 17, as the percent difference in measured resistance R with that of the resistance recorded at the initial test temperature and prior to application of the creep load R0@T. Note that in this way the initial value corresponds with the zero value of the total strain measurement. It is clear from the figure, that the change in electrical resistance is directly correlated to strain accumulation. Change in ER (%) = (R R 0@T ) R 0@T 100 (17) Recent studies using ER measurements during room and high-temperature uniaxial tensile strength testing of MI-CVI SiCf/SiC composites have shown increases in resistance of several hundreds of percent to failure. This large increase is attributed to the accumulation of damage in the form of transverse matrix cracking and associated fiber debonding that act to impede the electrical current flow [57, 108]. Therefore, the minor increase in ER during mechanical loading of ZMI-1 (~ 1%) suggests little to no damage occurred to the composite during loading, thereby confirming that the loading was mainly elastic. Also, the time-dependent nature of the ER change appears to be in good agreement with that of the creep strain. Both curves show a rapidly increasing primary region over the same time interval, followed by a quasi-steady-state increase to rupture. The similarity between the strain and ER curves demonstrates the ability of ER monitoring to capture time-dependent effects. Furthermore, the ER measurements 157

174 seem to be more sensitive, increasing 16% over the less that 0.1% increase in elongation. It is likely that the electrical resistance is sensitive to not only strain, but environmental effects as well. The sensitivity and versatility of ER monitoring make it a useful addition to this type of high temperature testing, where traditionally only strain measurements have been employed. T sur T back ER Loading begins (a) 158

175 Strain ER Loading Strain Loading ER (b) Figure 45: (a) ER response and thermography data for initial loading (thermal followed by mechanical) of uncoated ZMI-1 sample. (b) Time dependent strain and ER change due to constant stress (69 MPa) under high temperature thermal gradient conditions in air of uncoated CMC sample ZMI-1. A second test was conducted on an uncoated, ZMI fiber reinforced CMC (ZMI-2) from the same panel as the previous test. The thermal and mechanical loading history for this test is shown in Figure 46a. A series of laser faults during specimen heating resulted in two thermal cycles before steady-state temperatures could be reached. While the load frame is programmed for a zero load-hold condition, there is evidence that a zero longitudinal stress state could not be maintained during these thermal cycles. It is believed that the laser faults resulted in a rapid cool down of the test specimen, leading to a high thermal stress increase and momentary load spike until the frame could compensate. A permanent change in electrical resistance at room 159

176 temperature from Ohm to Ohm was observed. The residual increase in ER at room temperature of nearly 100% indicates a significant degree of material damage due to stress induced matrix crack formation [81]. Not only did the room temperature electrical resistance increase, but Figure 46b shows how the temperature-dependent electrical resistance behavior differs from the original heat-up, to the post-thermalstress damage heat-up. It is interesting to note that the ER behavior is essentially the same, however the entire curve has been shifted, reflecting the residual increase caused by matrix crack formation. Because loading data is not typically monitored during specimen heating, without this change in ER it would have been difficult to quantify or even detect this unexpected thermal stress increase. As mentioned in Chapter 5, such a large permanent increase in ER due to matrix cracking leads the possibility of using ER measurements not only as a monitoring technique, but for inspection of components as well. It should be noted that the surface of ZMI-2 was subjected to the same laser power as ZMI-1, however the result was the slightly lower surface and backside temperatures of 1092 C and 1028 C at the beginning of ZMI-2 loading. 160

177 T sur T back ER X X O Loading begins O (a) Initial Post-damage (b) 161

178 Strain ER Loading Strain Loading ER (c) Figure 46: (a) ER response and thermography data for thermal and mechanical loading of ZMI-2. Initial heating cycles represent laser faults resulting in pre-test, thermal stress induced damage. Damage events evident in ER spikes during cool down (X), and residual increase in room temperature ER (O). (b) ER response of ZMI-2 during initial and post-thermal-stress-damage heating. The average CMC temperature assumes linear temperature distribution between measured surfaces. (c) Time dependent strain and ER change due to constant stress (69 MPa) under high temperature thermal gradient conditions in air of uncoated CMC sample ZMI-2. Following laser heating, the specimen was loaded to a composite stress of 69 MPa for stress-rupture testing. The total strain and change in electrical resistance at temperature are plotted in Figure 46c. The pre-cracked sample ZMI-2 exhibited a similar strain increase during loading to the previous ZMI-1 sample. While a decrease in tensile modulus is expected due to the presence of matrix cracking, the applied load of only 69 MPa does not appear to dramatically increase the strain accumulation upon reloading. Also worth noting is that the large increase in primary creep strain seen in the first couple of hours ZMI-1 testing is not observed in ZMI-2. This is likely because the 162

179 high temperature matrix relaxation mechanism that causes such an effect is not present here due to the fact that stress in the MI matrix of ZMI-2 was already partially relieved via the transverse matrix cracking experienced prior to testing. Upon loading, ZMI-2 experiences a gradually decreasing creep rate over the first approximately 40 hours, followed by a quasi-steady creep rate until rupture. The electrical response due to loading Figure 46c shows that the pre-cracked sample demonstrated a much higher increase in ER during mechanical loading (~44%) than the previous ZMI-1 test (~1%). This large disparity in electrical response can be explained by examining how matrix cracking effects the electrical current flow through the composite. As previously discussed, the silicon rich MI matrix that is produced during processing is far more conductive that the other composite constituents. The essentially intact matrix of ZMI-1 that was present after loading, therefore allows for a continuous electrical current path through the bulk of the composite. The fiber-bridged matrix cracks that have been generated in ZMI-2 however act to impede current flow through the specimen. Thus resulting in the previously mentioned residual increase in electrical resistance even under zero applied load. Upon the application of load, the pressure forces between opposing matrix crack faces begin to decrease until a high enough tensile load is applied to separate them completely. This causes an increasing electrical contact resistance between crack faces that eventually reaches infinity when the crack has been completely opened. The obvious result of this is electrical current transfer from the matrix to the considerably more resistive crack bridging fibers, thereby drastically increasing the overall resistance of the composite. 163

180 After the specimen is loaded, some interesting observations can be made regarding the time dependent electrical response. Figure 46c shows the time-dependent composite behavior, in which the limited creep strain to rupture (~ 0.048%) increases the ER an additional 59% over the 127 hour rupture time. While an undamaged composite has essentially an iso-strain condition between constituents during increasing creep strains, the pre-cracked composite experiences crack opening and relative fiber/matrix displacements in the associated debonded regions. This results in large increases in electrical resistance as the current is forced into longer fiber lengths in order to bridge matrix cracks. It is also likely that the high density of open matrix cracks allowed for oxygen ingress deep into the composite, which was able to react with the matrix, fibers, and fiber/matrix interphase in the localized debonded regions. Over time the reaction products in these debonded regions could degrade the electrical contact between fiber and matrix in each crack wake, as well as degrade and reduce the crosssectional area of the current carrying crack bridging fibers, adding to the total increase in electrical resistance. Once again the electrical response appears to be exceptionally sensitive to material state; including the addition of thermal, mechanical, and environmental effects. Also, it is clear that the thermal stress damage recorded by the ER monitoring system during heating played a significant role in the increased strain accumulation and ER change during loading. The third in this series of heat-flux creep tests was performed on a similar ZMI reinforced substrate, that was subsequently coated on the face and sides with an EB- PVD EBC consisting of a HfO2+Si bond coat, Yb2Si2O7 + Hf-RE silicate top coat. Like ZMI-1, 164

181 this sample (ZMI-3) did not experience composite damage caused by thermal stresses during heating, and therefore can also be considered as a pristine specimen. This EBC- CMC system was heated to an EBC surface temperature TEBCsur = 1271 C. At this EBC surface temperature the resulting back side substrate temperature Tback of 989 C was recorded. If it is reasonably assumed that this substrate has the same (or very similar) thermal conductivity as the previously tested ZMI reinforced composites, then this back side temperature indicates that a similar thermal gradient is produced through the thickness of the substrate. After being heated to the surface temperature mentioned, ZMI-3 was loaded to a composite stress of 69 MPa to investigate the time-dependent creep and electrical properties. The total strain and percent increase in measured ER are plotted in Figure 47a. Due to an initial extensometer malfunction, the specimen was allowed to cool (maintaining the applied 69 MPa load) and the extensometer was reset and strain data acquisition was resumed. This unfortunately resulted in the lack of accurate total strain data being presented from time zero to approximately 40 hours. Upon resetting the extensometer the creep strain response of the sample shows similar behavior to the previous tests in which a primary region of strain accumulation was observed, followed by a secondary region of decreased creep rate. Although being subjected to the same constant load and similar substrate temperature conditions as the other tests, the coated ZMI-3 sample did not fail and was able to survive +500 hours of testing. It is therefore reasonable to assume that the addition of the EB-PVD EBC was responsible for the increased durability of the CMC substrate. It is likely that a well bonded EBC could 165

182 prevent oxygen ingress to any matrix micro-cracks that could have been generated on the CMC surface during loading, thereby increasing creep life by reducing fiber, matrix and or interphase oxidation and degradation. After the 526 hour constant tensile stress condition, the sample was maintained at the same temperature as the creep testing and unloaded and monotonically reloaded (at a rate of 0.127mm/min) to failure. Figure 47b shows results of the post-creep retained high temperature strength test. The figure shows a distinct change in material response beginning at 0.14% strain, at which point the applied stress exceeds that of the creep loading and the data becomes increasingly nonlinear to failure at 0.2% strain and a peak stress of 114 MPa. This nonlinear behavior is indicative of some level of matrix crack initiation/propagation resulting in load sharing between matrix and intact bridging fibers as described in previous chapters. Strain ER Loading ER (a) 166

183 Stress ER (b) Figure 47: (a) Time dependent strain and ER change due to constant stress (69 MPa) under high temperature thermal gradient conditions in air of coated ZMI-3 sample. (b) Mechanical and electrical response of post-creep, retained high temperature strength testing of ZMI-3. The associated electrical behavior during ZMI-3 creep testing is shown in Figure 47a. Similar to the ZMI-1 sample, only a ~4% increase in ER was observed upon specimen loading. This indicates that loading was mainly elastic, resulting in only micro-level matrix cracking. Furthermore, the following ER response appears to once again correlate with the rate of creep strain of the sample. That is the change in electrical resistance demonstrates a similar rapid primary increase followed by a slower, seemingly steady rate of increase for the remainder of the test. Once again the ER measurements appear to be quite sensitive to the increase in total strain, with an increase of over 40% in resistance at temperature to a corresponding 0.44% increase in total strain. It is interesting that the ratio of percent ER increase to total strain for ZMI-3 is quite similar 167

184 to that of ZMI-1 (110 and 84 respectively). Also, it is worth noting that the approximately forty percent increase in ER over the 526 hours of creep testing at 69 MPa observed in ZMI-3 is far less than the ER increase of the pre-cracked ZMI-2 sample that lasted only 127 hours. Figure 47b shows the ER response during the post-creep retained tensile strength testing of ZMI-3. Note that here the change in electrical resistance uses the value of electrical resistance at temperature post-creep and at zero applied load The increase in ER with strain shows a few distinct changes that appear to correlate with changes in mechanical behavior. The most striking being the rapid increase in change in ER that begins at the same time that the stress-strain relationship becomes increasing nonlinear. At this point, the ZMI-3 sample exceeds the constant creep load and matrix crack propagation occurs. This results in a rapid increase in ER of approximately 14% until fracture of the specimen. Once again it is worth noting that the ER is quite sensitive to strain accumulation Post-test Damage Assessment of Creep Specimens In order to investigate the possible differences in failure modes between the laser heat-flux tensile creep specimens, post-test analysis via microscopy was performed. A representative portion near the one side of the fracture plane of each post-tested sample was mounted in epoxy and polished along the edge in order to observe matrix crack density and morphology with the use of an optical microscope. For illustrated purposes, low magnification images of representative sections of each sample are shown in Figure 48a-c (with observed matrix cracks highlighted). Matrix cracks are 168

185 considered to be of two types: unbridged microcracks (did not penetrate through fiber tows), or fiber-bridged cracks which propagated through several fiber tows. Sample (ZMI-2) showed a high density of bridged transverse matrix cracks (2.5 cracks/mm), while ZMI- 1 and ZMI-3 contained a considerably lower density of un-bridged microcracks of 0.6 cracks/mm and 0.45 cracks/mm respectively. In general, the majority of the unbridged matrix cracking seen in the non-pre-cracked specimens appeared to initiate along the back face of the CMC and propagate only through the first few fiber plys. Secondly, the results of the microscopy seem to verify that the large change in ER for ZMI-2 was in fact produced by thermal-stress induced matrix cracking. Heated face Heated face Loading direction 1 mm 1 mm (a) (b) 169

186 Vertical coating cracking Crack coalescence and intra-coating delamination 1 mm (c) (d) Figure 48: Optical microscopy images showing matrix crack morphologies of (a) ZMI-1, (b) ZMI-2, (c) ZMI-3 as well as (d) typical cracking observed in the EBC deposited on ZMI- 3 surface. The multilayer EBC deposited on the surface of ZMI-3 exhibited fairly good thermal stability over the course of the high temperature testing. The coating showed excellent bond coat adhesion, with no apparent areas of macro-scale bond coat rumpling or coating spallation. However, the EBC did exhibit vertical cracking originating at the coating surface, generally associated with strains induced via volumetric changes during coating sintering and creep at elevated temperatures [70, 71]. Furthermore, several horizontal cracks were observed along the interface of the two top coat materials as well as a limited number of small delamination cracks at the top coat/tgo interface. Scanning Electron Microscopy (SEM) was also performed on the fracture surfaces of ZMI reinforced specimens to assess any differences in damage mechanisms between the uncoated CMCs and the EBC-CMC system under the laser heat-flux testing. Composite 170

187 micrographs of the fracture surfaces of ZMI-2 and ZMI-3 are shown in Figure 49 and Figure 50 respectively. The most obvious feature of the ZMI-2 fracture surface is the complete lack of any apparent fiber pullout. This lack of appreciable fiber pullout lengths is indicative of degradation of the composite fibers and/or fiber/matrix interphase resulting in the appearance of brittle failure of the sample [34, 109]. It is clear that the high density of bridged transverse matrix cracks that were generated pretest, resulted in an increased path for the ingress of oxidizing species into the interior of the composite. Close-up images of the fracture surface from different locations through the thickness show the uniformity of the fracture surface and lack of fiber pullout throughout. 171

188 Laser Heated Face 2 1 Back Face 1 2 Figure 49: At top: Composite SEM micrograph of ZMI-2 fracture surface. At bottom: higher magnification SEM micrographs of specific areas of the fracture surface to show lack of fiber pullout through thickness. 172

189 3 Back Face (Uncoated) 2 1 Laser Heated Face (Coated) Figure 50: At top: Composite SEM micrograph of EBC-CMC sample ZMI-3. At bottom: higher magnification images of specific areas of the fracture surface to highlight difference in apparent fiber pullout, progressing from near the coated surface to the back (uncoated) surface. The fracture surface of ZMI-3 shown in Figure 50 shows many distinct differences to that of ZMI-2. Most apparent from the composite micrograph is the presence of pulled out fibers. As discussed the existence of fiber pullout along the fracture surface indicates to some extent the lack of oxidation ingress within the interior of the 173

190 composite. The evidence for decreased environmental degradation of this sample is further supported by the fact that ZMI-3 was able to successfully survive the entire +500 hour high temperature sustained loading. That is, the abundance of pulled out fibers indicates that enough of the reinforcing fibers were sufficiently intact by the end of creep testing to result in some level of fiber bridging and sliding (global load transfer between constituents) prior to failure during post-creep retained strength testing. A closer look at the fracture surface as you move through the thickness from front (coated) surface to the back surface is shown below the ZMI-3 composite image in Figure 50. From these micrographs, it appears that the magnitude of fiber pullout decreases as you move from the front to back of the sample. This observation is in agreement with the observations from the optical micrographs of the longitudinal crosssection of ZMI-3 shown in Figure 48c. That is, the optical microscopy showed that the majority of matrix cracks appeared as unbridged matrix cracks that initiated along the back surface of the sample and propagated into the composite within only the first few fiber plys. Therefore, the results of the microscopy suggest that the back surface of the sample was the primary source for oxidation ingress, however as the surface cracks did not penetrate through the thickness the interior of the composite remained less exposed to environmental degradation. Furthermore, it is also possible that the well adhered EBC prevented oxygen ingress into the CMC front surface cracks further increasing the creep resistance of ZMI

191 8.4 Conclusions The purpose of this study was to investigate the damage mechanisms and timedependent thermo-mechanical and electrical response of two uncoated MI-CVI SiCf/SiC CMCs (ZMI-1 and ZMI-2) and an EBC-CMC system (ZMI-3) under laser heat-flux testing. A series of tests were performed to simulate the high-temperature thermal gradient and stressed oxidation environments these materials will experience if implemented in turbine engine hot-sections. Furthermore, in-situ electrical resistance (ER) measurements were included during testing in order to assess the applicability and possible advantages and/or difficulties in correlating ER with composite behavior. The specimens were tested at a constant applied composite stress of 69 MPa (below the proportional limit of these systems). However, prior to testing the ZMI-2 sample was exposed to loading via thermal stresses above the matrix cracking limit. It is worth noting once again that the thermal stress induced matrix cracking of ZMI-2 would not have been detected without the use of ER monitoring as there was load was not recorded during heat-up. The reason for the difference in mechanical loading and creep behavior of ZMI-2 and ZMI-1 would therefore otherwise be unrealized. Hence, a significant conclusion from this study is the usefulness of ER monitoring in complex high temperature mechanical testing. Also, the inclusion of a pre-cracked sample allows for the comparison between the stressed oxidation behavior between a previously damage (ZMI-2) and undamaged (ZMI-1) CMC. While both samples showed similar (quasisteady) secondary behavior, it was observed that the pre-cracked sample did not demonstrate the same large initial increase in composite strain during the first few 175

192 hours of testing. The behavior of a fully bonded composite is associated with matrix stress relaxation and load shedding to the reinforcing fibers under and iso-strain condition. This is likely not seen in ZMI-2 because the stress in the matrix was already relaxed by the creation of a large number of transverse matrix cracks prior to testing. One of the most striking observations that can be made by comparing the creep response between the different samples is the dramatic increase in stress-rupture life of ZMI-3 provided by the application of an environmental barrier coating (EBC). This specimen was able to survive over 500 hours of testing (far beyond that of the uncoated samples), demonstrating the increase in lifetime when a protective coating is used to decrease surface recession of the CMC and reduced infiltration of the interior of the composite with oxidizing species. Electrical resistance measurements were also taking in order to determine any possible correlation between ER increase and time-dependent strain accumulation as well as to investigate the convoluted change in electrical properties of the materials to thermal and mechanical loading, and oxidizing environments. A summary of the change in ER from the measured room temperature as-produced value for each sample is shown in Figure

193 Heating Loading Creep Post-test FF UNCOATED Tsur =1092 C Tback = 1028 C σ = 69 MPa Rupture: 127 hr COATED T EBCsur =1271 C Tback = 989 C σ = 69 MPa Run-out: 526 hr UNCOATED Tsur =1138 C Tback = 994 C σ = 69 MPa Rupture: 83 hr Figure 51: Summary plot of in-situ ER measurements collected during testing of each sample to demonstrate the increase in electrical resistance due to thermal and mechanical loading and time-dependent effects. The temperature dependent response of MI-CVI SiCf/SiC CMCs has already been discussed in great detail in Chapter 4. However, the disparity in ER increase due to thermal loading is quite evident in the ZMI-2 sample as compared to the rest. The ER measurements taken during laser heating of ZMI-2 therefore proved very useful in identifying damage induced from high temperature thermal stress that would have otherwise not been detected. The residual increase in ER at room temperature caused by this damage once again shows the possibility of using ER as an inspection technique as well as for in-situ monitoring during laboratory testing. Upon loading at high temperature the ER measurements also indicate the existence of a large accumulation of matrix cracks prior to testing as the increase is several times that of the other 177

194 samples in the study. The extent of transverse matrix cracking was confirmed via posttest optical microscopy, showing a relatively high number of bridged cracks in sharp contrast to the significantly lower number of mostly surface cracks observed in the ZMI- 1 and ZMI-3 samples. ER measurements demonstrated a higher level of sensitivity to elongation during high temperature loading than typical extensometer measurements. In particular an increase of only 0.089% strain during loading of ZMI-1 resulted in a corresponding increase of ER at temperature of 1%. Furthermore, an increase in resistance at temperature of 44% was observed in ZMI-2 which accounted for the crack opening and debonded fiber sliding during loading. During the creep, the ER increase of ZMI-1 and ZMI-3 appear to be correlate well with their respective strain under the same constant stress condition. While ZMI-2 experience a similar failure strain as ZMI-1, the large ER increase seen in ER during stress-rupture testing appears to be caused by environmental attack of the interior of the composite which further inhibited the flow of electrical current through time-dependent oxidation mechanisms. This conclusion was bolstered by evidence of material degradation observed through SEM microscopy of the ZMI-2 fracture surface. The creep response of ceramic composites is temperature-stress-timeenvironment dependent, and one of the most interesting this about the collection of sample in this study is that is shows how electrical resistance of these systems is as well. While further study of the induvial physical mechanisms affecting the electrical behavior of MI-CVI SiCf/SiC composites under high heat-flux creep conditions is required, this study demonstrated the potential for using ER as a diagnostic tool that can be used to 178

195 characterize material behavior under complex high temperature time-dependent deformation. 179

196 CHAPTER IX CONCLUSIONS This worked attempted to go beyond typical room temperature characterization of mechanical properties and instead focused on several critical high temperature lifelimiting damage issues of woven melt-infiltrated SiC/SiC ceramic matrix composites and various protective ceramic coatings systems. The experimental work presented here has been constructed to simulate actual engine environments by introducing high temperature thermal gradients, environmental exposure, thermal cyclic and time dependent damage mechanisms otherwise unaddressed by current research. Furthermore, several novel non-destructive evaluation and monitoring techniques were introduced to characterize material damage behavior and provide a basis for further modeling and design efforts. The first case study included the use of a commercially available thermoelectric testing device as well as the development of a laser-based approach to determine the temperature-dependent electrical response of several MI-CVI SiCf/SiC CMCs. Testing was performed in order to create a useful data base for modeling and to determine the capability of using ER measurements as a monitoring technique in high-temperature mechanical testing. Once the high temperature ER properties were determined, the 180

197 respective contribution of the longitudinal fibers and effective matrix material to the overall electrical response was estimated using a parallel circuit model. Furthermore, the temperature profile along the length of the specimen developed by the laser-based heating approach was determined and a series circuit model was used to show the contribution of the longitudinal thermal gradient on the measured ER value. It was concluded that the temperature-dependent electrical behavior of all of the composite systems tested demonstrated similar behavior in the form of increasing ER with temperature followed by an observable transition to decreasing ER at moderate to high temperatures. This behavior is analogous to moderately doped semiconductor materials that display a transition from extrinsic to intrinsic behavior at a transition temperature based on their dopant type and concentration. It was further concluded that the silicon rich MI matrix dominated the ER response at all temperatures of interest, thereby confirming the capability of using ER monitoring in high temperature mechanical testing to detect stress dependent matrix crack accumulation. The following section focuses on damage characterization of MI-CVI SiCf/SiC CMCs with stress concentrations under high heat-flux tensile loading. A modified tensile frame that incorporates a laser-heating apparatus was used to simulate high thermal gradients and mechanical stresses. Various monitoring techniques (AE, ER and DIC) were utilized to determine localized damage accumulation resulting from the applied mechanical loading. The matrix crack morphology observed in post-test microscopy measurements was in good agreement with the localized matrix cracking determined from AE analysis and strain mapping produced using DIC. It was concluded that the energy density of AE 181

198 measurements taken near the stress concentration and in the far-filed are directly related to matrix crack density. Also, in-situ ER monitoring was shown to be sensitive to damage accumulation in the form of matrix cracking and sliding of debonded bridging fibers in the matrix crack wake. Finally, the use of post-test room temperature ER inspection was shown to be sensitive to the competing factors of material damage and microstructural changes arising from high temperature exposure. The next section investigated the difference in high temperature retained mechanical properties of an uncoated SiCf/SiC sample and a coated EBC/CMC system after 30 hours of exposure in a specialized high pressure burner rig used to simulate oxidizing engine environments. A decrease in both strength and toughness was observed in the severally degraded uncoated sample, while the substrate protected by the EBC showed the inelastic behavior and graceful failure desired from engineered ceramics. It was found that differences in ER and AE monitoring data between the samples correlate to the differences in damage accumulation between the highly localized unbridged matrix cracking of the uncoated sample and the large scale fiberbridged matrix cracking behavior of the EBC/CMC. Through AE location analysis and DIC strain mapping it was further concluded that AE is a successful tool in determining the highly localized damage associated with the failure plane of the uncoated sample as well as the more dispersed damage of the EBC coated sample. Finally, it was found that AE waveform frequency centroid shows promise as a method of distinguishing EBC cracking due to high mechanical stresses, however this requires further development. 182

199 The following chapter addresses the ceramic coating life-limiting case of thermal cyclic loading. Laser heat-flux tests thermal cyclic tests were performed on EB-PVD 7YSZ coated superalloys substrates. Samples were produced in order to show difference in damage mode and accumulation of EB-PVD ceramic coatings due to thermal stress concentrations and molten CMAS infiltration. Couple thermography and AE monitoring was used to show when different damage mechanisms occurred and the severity to which they affected coating performance and durability. The as-deposited coating exhibit vertical cracking produced by high temperature sintering effects that lead to an increase in thermal conductivity and a small initial increase in measured AE energy. The sample containing a stress concentration initiated large scale coating debonding cracks that was observed in the thermography data as a decrease in through thickness thermal conductivity and a large increase in AE energy accumulation during transient heating/cooling cycles. Finally, the sample that was exposed to CMAS infiltration demonstrated little to no sintering due to prior densification from the CMAS, followed by increased damage accumulation due to coating cracking occurring due heating/cooling. It was observed that the overall AE energy accumulation in the CMAS infiltrated sample was several times that of the as-deposited coating. These finding show the level of increased degradation to coating systems caused by the presence of stress concentrations and CMAS. It can be concluded from this study that AE analysis can be used to detect various forms of coating damage, quantify the amount of damage accrued, and has the capability of determining damage location. Furthermore, the use of 183

200 coupled thermography measurements is can be used to qualitatively discriminate between specific damage modes. The final chapter investigated the time-dependent thermo-mechanical and electrical responses of uncoated and EBC coated MI-CVI SiCf/SiC composite systems under laser heat-flux testing. The first two tests demonstrated the difference in creep and ER response between CMCs with and without damage prior to testing. While the ER response of the previous uncracked sample is indicated of time-dependent elongation of the sample due to creep, the sample containing a high density of matrix cracks showed a much larger increase in ER due to crack opening and environmental degradation of the composite due to large scale oxidation of the of the interior of the composite. The effectiveness of the EBC to increase stress-rupture life was demonstrated as the coated sample was capable of surviving +500 hours under similar thermal and mechanical conditions. Post-test microscopy of the fracture surfaces showed the increase in fiber pullout and minimal oxidation seen in the coated sample. The ability of ER monitoring to identify matrix cracking due to rapid application of thermal stresses that would have otherwise gone undetected demonstrates the usefulness of ER for use in complex thermo-mechanical testing. It was concluded that ER measurements correlate to all aspects of this type of testing, including temperature, stress and environmental effects. However, further investigation is required to fully understand and quantify each. 184

201 CHAPTER X FUTURE WORK The demonstrated use of electrical resistance measurements for high temperature and thermo-mechanical testing leads to some possible advancements in electromechanical modelling to help characterize these complex testing scenarios. Current ER models under development to explain room temperature tensile damage response could be expanded to include the temperature dependent composite and constituent electrical properties determined in this work. It would be useful to perform tests on single fibers and/or fiber tows in order to establish more accurate temperature dependent ER behavior of current vintage SiC fibers and fiber coatings. Further robustness of the model could be achieved by including the effects of non-uniform damage accumulation by thermal and geometrical stress concentration. From the study on using coupled thermography and AE monitoring to characterize coating cracking and failure further AE waveform analysis should be investigate. It may be possible to correlate specific characteristics of modal AE waveforms with specific damage modes and/or cracking types. In terms of investigating the creep and stress-rupture of ceramic matrix composites and environmental barrier coatings under high heat-flux thermal gradient conditions, there are several areas that should be investigated further. The usefulness of ER 185

202 measurements has been shown here, but the cumulative effect of temperature, stressstate and environmentally assisted degradation on ER response requires further investigation in order to be effectively used in general analysis and future modeling efforts. This world likely require further testing to establish specific relationships between the various factors affecting ER response. The ability to model the electrical behavior of such convoluted tests would prove extremely useful for damage characterization and future material design. 186

203 BIBLIOGRAPHY [1] DiCarlo, J.A., Yun, H-M, Morscher, G.N., Bhatt, R.T. SiC/SiC composites for 1200 C and above. In: Bansal NP, editor. Handbook of Ceramics Composites. NY, NY: Kluwer Academic; p [2] Dicarlo, J.A., Advances in SiC/SiC Composites for Aero-Propulsion. In: Narottam, B.P., Lamon, J, editors. Ceramic Matrix Composites: Materials, Modeling and Technology. Wiley; p [3] D. Brewer, HSR / EPM combustor materials development program, Mater. Sci. Eng. A. 261 (1999) [4] Gardiner, Ginger, Aeroengine Composites, Part 1: The CMC invasion, Composites World, July 31, [5] Motz, G., Schmidt, S., Beyer, S., The PIP-process: Precursor Properties and Applications. In: Krenkel, W., editor. Ceramic Matrix Composites: Fiber Reinforced Ceramics and their Applications. Wiley; p [6] Leuch, M., Chemical Vapor infiltration for Ceramic Matrix Composites: Manufacturing, Properties, Applications. In: Krenkel, W., editor. Ceramic Matrix Composites: Fiber Reinforced Ceramics and their Applications. Wiley; p [7] Corman, G.S., Luthra, K.L., Silicon melt-infiltrated ceramic composites (HiPerComp). In: Bansal NP, editor. Handbook of Ceramics Composites. NY, NY: Kluwer Academic; p [8] D.R. Clarke, S.R. Phillpot, Thermal barrier coating materials, Materialstoday. (2005) [9] Padture, Nitin P., Maurice Gell, and Eric H. Jordan, Thermal barrier coatings for gasturbine engine applications, Science (2002) [10] Darolia, R, Thermal barrier coatings technology: critical review, progress update, remaining challenges and prospects, International Materials Reviews. Volume 58, Issue 6 (August 2013), pp [11] J.A. Dever, M. V Nathal, J.A. Dicarlo, Research on High-Temperature Aerospace Materials at NASA Glenn Research Center, J. Aerosp. Eng. 26 (2013)

204 [12] Robinson, R.C., and Smialek, J.L., SiC recession caused by SiO2 scale volatility under combustion conditions: I, Experimental Results and Empirical Model, J. Am. Ceram. Soc., 82(7) (1999), [13] Opila, E.J., Smialek, J.L., Robinson, R.C., Fox, D.S., and Jacobson, N.S., SiC recession caused by SiO2 scale volatility under combustion conditions: II, Thermodynamics and gaseous-diffusion model. J. Am. Ceram. Soc., 82(7) (1999), [14] K.L. More, P.F. Tortorelli, L.R. Walker, N. Miriyala, J.R. Price, M. Van Roode, High- Temperature Stability of SiC-Based Composites in High-Water-Vapor-Pressure Environments, J. Am. Ceram. Soc. 81 (2003) [15] J.L. Smialek, R.C. Robinson, D.S. Fox, N.S. Jacobson, SiC and Si3N4 recession due to SiO2 scale volatility under combustion conditions, Adv. Compos. Mater. 8 (1999) [16] K.N. Lee, Current status of environmental barrier coatings for Si-Based ceramics, Surf. Coatings Technol (2000) 1 7. [17] K.N. Lee, R.A. Miller, N.S. Jacobson, New generation of plasma-sprayed mullite coatings on silicon-carbide. J. Am. Ceram. Soc., 78(3), (1995) [18] K.N. Lee, Key durability issues with mullite-based environmental barrier coatings for Si-based ceramics. J. Eng. Gas Turbines Power, 122(10), (2000). [19] Eaton, H.E., Linsey, G.D., Sun, E.Y., More, K.L., Kimmel, J.B., Price, J.R. et al., EBC protection of SiC/SiC composites in the gas turbine combustion environment continuing evaluation and refurbishment considerations. ASME paper 2001-GT-0513 ASME TURBOEXPO June 4 7, New Orleans, Louisiana. [20] K.N. Lee, D.S. Fox, J.I. Eldridge, D. Zhu, R.C. Robinson, N.P. Bansal, et al., Upper temperature limit of environmental barrier coatings based on mullite and BSAS, J. Am. Ceram. Soc. 86 (2003) [21] K.N. Lee, D.S. Fox, N.P. Bansal, Rare earth silicate environmental barrier coatings for SiC/SiC composites and Si3N4 ceramics, J. Eur. Ceram. Soc. 25 (2005) [22] D. Zhu, D.S. Fox, N.P. Bansal, R.A. Miller, Advanced Oxide Material Systems for 1650 C Thermal/Environmental Barrier Coating Applications, (2004) NASA TM [23] D. Zhu, R.A. Miller, Thermal and environmental barrier coatings for advanced propulsion engine systems, (2004) NASA TM [24] D. Zhu, R.A. Miller, Fox, D.S., Thermal and environmental barrier coating development for advanced propulsion engine systems, (2007) NASA TM

205 [25] D. Zhu, Advanced Environmental Barrier Coatings for Sic/Sic Ceramic Matrix Composite: Performance and Directions, presented at 7th International Conference on High Temperature Ceramic Matrix Composite (HT-CMC-7), Bayreuth, Germany, September 20-22, 2010 [26] D. Zhu, Advanced Environmental Barrier Coating Development for SiC/SiC Ceramic Matrix Composites, presented at Workshop on the Design of Ceramic-Fiber Based Composites for Service Above 1400 C, Boulder, CO, June 11, [27] J. Aveston, G.A. Cooper, A. Kelly, Single and Multiple Fracture, Prop. Fiber Compos. Proc. Conf., Nov., [28] Pryce, A.W., Smith, P.A., Matrix cracking in unidirectional ceramic matrix composites under quasi-static and cyclic loading, Acta Metall. Mater. 41 (1993) [29] J.W. Hutchinson, H.M. Jensen, Models of Fiber debonding and Pullout in Brittle Composites with Friction, Mech. Mater. 9 (1990) [30] D.B. Marshall, Analysis of Fiber Debonding and Sliding Experiments in Brittle Composites, Acta Met. Mater. 40 (1992) [31] Parthasarathy, T.A., Marshall, D.B., Analysis of the Effect of Interfacial Roughness on Fiber Debonding and Sliding in Brittle Matrix Composites, Acta Met. Mater. 42 (1994) [32] D.B. Marshall, W.C. Oliver, An indentation method for measuring residual stresses in fiber-reinforced ceramics, Mater. Sci. Eng. A. 126 (1990) [33] W.A. Curtin, Exact theory of fibre fragmentation in a single-filament composite, J. Mater. Sci. 26 (1991) [34] W.A. Curtin, Theory of Mechanical Properties of Ceramic-Matrix Composites, J. Am. Ceram. Soc. 74 (1991) [35] W.A. Curtin, Multiple Matrix Cracking in Brittle Matrix Composites, Acta Met. Mater. 41 (1993) [36] N. Lissart, J. Lamon, Damage and Failure in Ceramic Matrix Minicomposites: Experimental Study and Model, 45 (1997) [37] W. A. Curtin, B.K. Ahn, N. Takeda, Modeling brittle and tough stress strain behavior in unidirectional ceramic matrix composites, Acta Mater. 46 (1998)

206 [38] E. Vagaggini, J.M. Domergue, A.G. Evans, Relationships between Hysteresis Measurements and the Constituent Properties of Ceramic Matrix Composites: I, Theory, J. Am. Ceram. Soc. 78 (1995) [39] J.M. Domergue, E. Vagaggini, A.G. Evans, Relationships between Hysteresis Measurements and the Constituent Properties of Ceramic Matrix Composites II, Experimental Studies on Unidirectional Materials, J. Am. Ceram. Soc. 78 (1995) [40] Huang, M., Jiang, L., Liaw, P. K., Brooks, C. R., Seeley, R., & Klarstrom, D. L., Using acoustic emission in fatigue and fracture materials research, JOM, 50 (11), (1998) [41] Gorman, M. R., New technology for wave based acoustic emission and acoustoultrasonics. In International Instrumentation Symposium, 41st, Denver, CO, (1995) [42] Gorman, M. R., Modal AE: A new understanding of acoustic emission. J. Acoust. Emiss., Technical Note (1997). [43] M.R. Gorman, Plate waves produced by transverse matrix cracking, Ultrasonics. 29 (1991) [44] G.N. Morscher, N. Godin, Use of Acoustic Emission for Ceramic Matrix Composites, in: Ceram. Matrix Compos. Mater. Model. Technol., 2015: pp [45] G.N. Morscher, Modal acoustic emission of damage accumulation in a woven SiC / SiC composite, Compos. Sci. Technol. 59 (1999) [46] G.N. Morscher, A.L. Gyekenyesi, The velocity and attenuation of acoustic emission waves in SiC/SiC composites loaded in tension, Compos. Sci. Technol. 62 (2002) [47] G.N. Morscher, Stress-dependent matrix cracking in 2D woven SiC-fiber reinforced melt-infiltrated SiC matrix composites, Compos. Sci. Technol. 64 (2004) [48] E. Maillet, N. Godin, M. R Mili, P. Reynaud, G. Fantozzi, J. Lamon, Damage monitoring and identification in SiC/SiC minicomposites using combined acoustoultrasonics and acoustic emission, Compos. Part A Appl. Sci. Manuf. 57 (2014) [49] M. Moevus, D. Rouby, N. Godin, M. R Mili, P. Reynaud, G. Fantozzi, G. Farizy, Analysis of damage mechanisms and associated acoustic emission in two SiC/[Si-B-C] composites exhibiting different tensile curves. Part I: Damage patterns and acoustic emission activity, Comp. Sci. Technol. 68 (2008)

207 [50] M. Moevus, N. Godin, M. R Mili, D. Rouby, P. Reynaud, G. Fantozzi, G. Farizy, Analysis of damage mechanisms and associated acoustic emission in two SiC/[Si-B-C] composites exhibiting different tensile curves. Part II: unsupervised acoustic emission data clustering, Comp. Sci. Technol. 68 (2008) [51] E. Maillet, N. Godin, M. R Mili, P. Reynaud, G. Fantozzi, J. Lamon, Real-time evaluation of energy attenuation: A novel approach to acoustic emission analysis for damage monitoring of ceramic matrix composites, J. Eur. Ceram. Soc. (2014). [52] E. Maillet, G.N. Morscher, Waveform-based selection of acoustic emission events generated by damage in composite materials, Mech. Syst. Signal Process (2015) [53] E. Maillet, C. Baker, G.N. Morscher, V. V. Pujar, J.R. Lemanski, Feasibility and limitations of damage identification in composite materials using acoustic emission, Compos. Part A Appl. Sci. Manuf. 75 (2015) [54] X. Wang, D.D.L. Chung, Fiber Breakage in Polymer-Matrix Composite During Static and Fatigue Loading, Observed By Electrical Resistance Measurement, J. Mater. Res. 14 (1999) [55] D.D.L. Chung, S. Wang, Self-sensing of Damage and Strain in Carbon Fiber Polymer- Matrix Structural Composites by Electrical Resistance Measurement, Polymers and Polymer Compos. 11 (2003) [56] C. Smith, G. Morscher, Z. Xia, Monitoring damage accumulation in ceramic matrix composites using electrical resistivity, Scr. Mater. 59 (2008) [57] G.N. Morscher, C. Baker, C. Smith, Electrical Resistance of SiC Fiber Reinforced SiC/Si Matrix Composites at Room Temperature during Tensile Testing, Int. J. Appl. Ceram. Technol. 11 (2014) [58] Baker, Christopher R. Assessing Damage in Composite Materials. Diss. The University of Akron, [59] B. Pan, K. Qian, H. Xie, A. Asundi, Two-dimensional digital image correlation for inplane displacement and strain measurement: a review, Meas. Sci. Technol. 20 (2009) [60] V.P. Rajan, M.N. Rossol, F.W. Zok, Optimization of Digital Image Correlation for High-Resolution Strain Mapping of Ceramic Composites, Exp. Mech. 52 (2012) [61] M.D. Novak, F.W. Zok, High-temperature materials testing with full-field strain measurement: experimental design and practice, Rev. Sci. Instrum. 82 (2011)

208 [62] A.G. Evans, D.R. Mumm, J.W. Hutchinson, Mechanisms controlling the durability of thermal barrier coatings, Prog. Mater. Sci. 46 (2001) [63] Sampath, S., Schulz, U., Jarligo, M. O., & Kuroda, S., Processing science of advanced thermal-barrier systems. MRS bulletin, 37 (2012), [64] A. Rabiei, A.G. Evans, Failure mechanisms associated with the thermally grown oxide in plasma-sprayed thermal barrier coatings, Acta Mater. 48 (2000) [65] K.W. Schlichting, N.P. Padture, E.H. Jordan, M. Gell, Failure modes in plasmasprayed thermal barrier coatings, Mater. Sci. Eng. A. 342 (2003) [66] A.G. Evans, M.Y. He, J.Wnaga. Hutchinson, Mechanics-based scaling laws for the durability of thermal barrier coatings, Prog. Mater. Sci. 46 (2001) [67] D.R. Mumm, A.G. Evans, On the role of imperfections in the failure of a thermal barrier coating made by electron beam deposition, Acta Mater. 48 (2000) [68] M.Y. He, A.G. Evans, J.W. Hutchinson, Ratcheting of compressed thermally grown thin films on ductile substrates, Acta Mater. 48 (2000) [69] D. Zhu, R. A. Miller, Sintering and creep behavior of plasma-sprayed zirconia- and hafnia-based thermal barrier coatings, Surf. Coatings Technol (1998) [70] D. Zhu, R.A. Miller, B.A. Nagaraj, R.W. Bruce, Thermal conductivity of EB-PVD thermal barrier coatings evaluated by a steady-state laser heat flux technique, Surf. Coatings Technol. 138 (2001) 1 8. [71] D. Zhu, R.A. Miller, Determination of creep behavior of thermal barrier coatings under laser imposed high thermal and stress gradient conditions, J. Mater. Res. 14 (1999) [72] D. Zhu, S.R. Choi, R. Ahal. Miller, Development and thermal fatigue testing of ceramic thermal barrier coatings, Surf. Coatings Technol (2004) [73] Eldridge, J. I., Zhu, D., & Wolfe, D. E., Monitoring Delamination of Thermal Barrier Coatings during Interrupted High-Heat-Flux Laser Testing Using Upconversion Luminescence Imaging, Ceramic Engineering and Science Proceedings, Vol. 32, No. 3, (2011) [74] A.G. Evans, J.W. Hutchinson, The mechanics of coating delamination in thermal gradients, Surf. Coatings Technol. 201 (2007) [75] B.D. Choules, K. Kokini, T.A. Taylor, Thermal fracture of ceramic thermal barrier coatings under high heat flux with time-dependent behavior. Part 1. Experimental results, Mater. Sci. Eng. A. 299 (2001)

209 [76] Y. Tan, J.P. Longtin, S. Sampath, D. Zhu, Temperature-Gradient Effects in Thermal Barrier Coatings: An Investigation Through Modeling, High Heat Flux Test, and Embedded Sensor, J. Am. Ceram. Soc. 93 (2010) [77] C. Mercer, S. Faulhaber, A.G. Evans, R. Darolia, A delamination mechanism for thermal barrier coatings subject to calcium-magnesium-alumino-silicate (CMAS) infiltration, Acta Mater. 53 (2005) [78] S. Krämer, S. Faulhaber, M. Chambers, D.R. dever, C.G. Levi, J.W. Hutchinson, et al., Mechanisms of cracking and delamination within thick thermal barrier systems in aeroengines subject to calcium-magnesium-alumino-silicate (CMAS) penetration, Mater. Sci. Eng. A. 490 (2008) [79] S. Krämer, J. Yang, C.G. Levi, C. a. Johnson, Thermochemical Interaction of Thermal Barrier Coatings with Molten CaO-MgO-Al2O3-SiO2 (CMAS) Deposits, J. Am. Ceram. Soc. 89 (2006) [80] N.L. Ahlborg, D. Zhu, Calcium-magnesium aluminosilicate (CMAS) reactions and degradation mechanisms of advanced environmental barrier coatings, Surf. Coatings Technol. 237 (2013) [81] Appleby, M., Morscher, G. N. and Zhu, D., Creep and Environmental Durability of Environmental Barrier Coatings and Ceramic Matrix Composites under Imposed Thermal Gradient Conditions, in Advanced Ceramic Coatings and Materials for Extreme Environments III (eds H.-T. Lin, T. Hwang, S. Kirihara and S. Widjaja), John Wiley & Sons (2013). [82] R. Scholz, F. Santos, B. Riccardi, Electrical conductivity of silicon carbide composites and fibers, J. Nucl. Mater. 307 (2002) [83] G.E. Youngblood, E. Thomsen, R. Shinavski, Electrical Conductivity of 2D-SiCf/CVI- SiC, Fusion Sci. Technol. Int. J. Am. Nucl. Soc. 60 (2011) [84] B. North, K.E. Gilchrist, Effect of Impurity Doping on a Reaction Bonded Silicon Carbide, Am. Ceram. Soc. Bull. 60 (1981) [85] ULVAC Website Available at [86] G.N. Morscher, V. V. Pujar, Design Guidelines for In-Plane Mechanical Properties of SiC Fiber-Reinforced Melt-Infiltrated SiC Composites, Int. J. Appl. Ceram. Technol. 6 (2009) [87] G.-D. Zhan, M. Mitomo, A.K. Mukherjee, Effects of heat treatment and sintering additives on thermal conductivity and electrical resistivity in fine-grained SiC ceramics, J. Mater. Res. 17 (2011)

210 [88] G. Roewer, U. Herzog, K. Trommer, E. Miiller, S. Friihauf, Silicon Carbide - A Survey of Synthetic Approaches, Properties and Applications, in: High Perform. Non-Oxide Ceram. I, 2002: pp [89] Abderrazak, Houyem, and Emna Selmane Bel Hadj Hmida. Silicon carbide: synthesis and properties. INTECH Open Access Publisher, [90] K. Inoue, S. Mori, A. Yamaguchi, Thermal Conductivity and Temperature Dependence of Electrical Resistivity of Al 4 SiC 4 SiC Sintered Bodies Prepared by Pulse Electronic Current Sintering, J. Ceram. Soc. Japan. 111 (2003) [91] C.E. Smith, G.N. Morscher, Z. Xia, Electrical Resistance as a Nondestructive Evaluation Technique for SiC/SiC Ceramic Matrix Composites Under Creep-Rupture Loading, Int. J. Appl. Ceram. Technol. 8 (2011) [92] J.H. Racette, Intrinsic Electrical Conductivity in Silicon Carbide, Phys. Rev. 107 (1957) [93] B. G. Yacobi, Semiconductor Materials: An Introduction to Basic Principles, Springer Science & Business Media, [94] P.Y. Yu, Fundamentals of Semiconductors: Physics and Material Properties, Springer, [95] R. Bhatt, Y.L. Chen, Heat Treatment Effects on Microstructure of SiC Fiber Preforms, NASA TM (1999). [96] T.S. Orlova, V.V. Popov, J. Quispe Cancapa, D. Hernández Maldonado, E. Enrique Magarino, F.M. Varela Feria, et al., Electrical properties of biomorphic SiC ceramics and SiC/Si composites fabricated from medium density fiberboard, J. Eur. Ceram. Soc. 31 (2011) [97] Robert F. Pierret, Semiconductor Device Fundamentals, Addison-Wesley, [98] A.L. Gyekenyesi, G.N. Morscher, Damage Progression and Stress Redistribution in Notched SiC/SiC Composites, J. Mater. Eng. Perform. 19 (2010) [99] Baker, C., Maillet, E., Appleby, M., Smith, R., Morscher, G. N. and Cook, T., Effects of Stress Concentrators on Damage Evolution in SiC/SiC Composites, in Mechanical Properties and Performance of Engineering Ceramics and Composites IX (eds D. Singh and J. Salem), John Wiley & Sons, Inc., Hoboken, NJ, USA. (2014) [100] Halbig, M.C., Jaskowiak, M.H., Kiser, J.D. and Zhu, D., Evaluation of Ceramic Matrix Composite Technology for Aircraft Turbine Engine Applications, 51st AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition, AIAA ,

211 [101] R. Naslain, Design, preparation and properties of non-oxide CMCs for application in engines and nuclear reactors: an overview, Compos. Sci. Technol. 64 (2004) [102] D. Zhu, Development and Performance Evaluations of HfO 2 -Si and Rare Earth-Si Based Environmental Barrier Bond Coat Systems for SiC / SiC Ceramic Matrix Composites NASA EBC and CMC System Development, 41st Int. Conf. Metall. Coatings Thin Film., San Diego, CA, [103] D. Zhu, J. Hurst, Advanced High Temperature and Fatigue Resistant Environmental Barrier Coating Bond Coat Systems for SiC/SiC Ceramic Matrix Composites, US Patent No , [104] D. Zhu, B.A. Sakowski, C. Fisher, Film Cooled Recession of SiC/SiC Ceramic Matrix Composites: Test Development, CFD Modeling and Experimental Observations, 41st Int. Conf. Metall. Coatings Thin Films. San Diego, CA, April 28-May 2, [105] G. N. Morscher. Tensile stress rupture of SiCf/SiCm minicomposites with carbon and boron nitride interphases at elevated temperatures in air. J. Am. Ceram. Soc., 80(8), (1997). [106] G.N. Morscher, Tensile creep and rupture of 2D-woven SiC/SiC composites for high temperature applications, J. Eur. Ceram. Soc. 30 (2010) [107] G.N. Morscher, V. V. Pujar, Creep and Stress-Strain Behavior after Creep for SiC Fiber Reinforced, Melt-Infiltrated SiC Matrix Composites, J. Am. Ceram. Soc. 89 (2006) [108] M.P. Appleby, D. Zhu, G.N. Morscher, Mechanical properties and real-time damage evaluations of environmental barrier coated SiC/SiC CMCs subjected to tensile loading under thermal gradients, Surf. Coatings Technol. 284 (2015) [109] A.G. Evans, C. Weber, Creep damage in SiC / SiC composites, Mater. Sci. Eng. A. 208 (1996). [110] Mackey J, Dynys F, Sehirlioglu A. Uncertainty analysis for common Seebeck and electrical resistivity measurement systems. Rev Sci Instrum; 85: [111] Incropera, Frank P., DeWitt, David P., Introduction to Heat Transfer 3 rd ed., Wiley & Sons, [112] Touloukian, Y.S., DeWitt, P.D., Thermal radiative properties: nonmetallic solids, Thermophys. Prop. Matter, Vol. 8 (1972). [113] 195

212 [114] Goldstein, R.J., Sparrow, E.M., Jones, D.C., Natural Convection Mass Transfer Adjacent to Horizontal Plates, Int. J. Heat Mass Transf. 16 (1973) [115] D.R. Clarke, M. Oechsner, N.P. Padture, Thermal-barrier coatings for more efficient gas-turbine engines, MRS Bull. 37 (2012) [116] S.R. Choi, D. Zhu, R. a. Miller, Fracture behavior under mixed-mode loading of ceramic plasma-sprayed thermal barrier coatings at ambient and elevated temperatures, Eng. Fract. Mech. 72 (2005) [117] D. Zhu, A. Miller, Thermal Conductivity and Elastic Modulus Evolution of Thermal Barrier Coatings under High Heat Flux Conditions, NASA TM (1999). [118] D. Zhu, A. Miller, Thermal Conductivity Change Kinetics of Ceramic Thermal Barrier Coatings Determined by the Steady-State Laser Heat Flux Technique, NASA TM (1999). [119] C.G. Levi, J.W. Hutchinson, M.-H. Vidal-Sétif, C. a. Johnson, Environmental degradation of thermal-barrier coatings by molten deposits, MRS Bull. 37 (2012) [120] X.Q. Ma, M. Takemoto, Quantitative acoustic emission analysis of plasma sprayed thermal barrier coatings subjected to thermal shock tests, Mater. Sci. Eng. A. 308 (2001) [121] X.Q. Ma, S. Cho, M. Takemoto, Acoustic emission source analysis of plasma sprayed thermal barrier coatings during four-point bend tests, Surf. Coatings Technol. 139 (2001) [122] D. Renusch, M. Schütze, Measuring and modeling the TBC damage kinetics by using acoustic emission analysis, Surf. Coatings Technol. 202 (2007) [123] M.R. Gorman, Plate Wave Acoustic Emission, J. Acoust. Soc. Am. 90 (1991) 358. [124] H. Peng, L. Wang, L. Guo, W. Miao, H. Guo, S. Gong, Degradation of EB-PVD thermal barrier coatings caused by CMAS deposits, Prog. Nat. Sci. Mater. Int. 22 (2012) [125] [126] M.H. Vidal-Setif, N. Chellah, C. Rio, C. Sanchez, O. Lavigne, Calcium-magnesiumalumino-silicate (CMAS) degradation of EB-PVD thermal barrier coatings: Characterization of CMAS damage on ex-service high pressure blade TBCs, Surf. Coatings Technol. 208 (2012)

213 APPENDICES 197

214 APPENDIX A DETERMINATION OF TEMPERATURE PROFILE USING SIMPLIFIED HEAT TRANSFER MODEL In order to calculate the longitudinal thermal gradient generated by the gage-section heating the tensile specimens used in the laser heating approach, various assumptions and boundary conditions need to be defined. First, it should be noted that the sample is relatively thin in comparison to the longitudinal dimension. Therefore, a major simplifying assumption is that the temperature can be assumed relatively constant in the thickness direction. Secondly, because the incident laser energy supplied to the surface of the specimen results in an extremely uniform heating area the temperature of the heated-section is considered constant as well. Since the hot-zone is assumed to be a constant value, a Dirichlet condition (boundary condition of the first kind) is defined along the edges of the heated region. Negligible heat is assumed to be lost through the other edges of the sample (along the sides and top/bottom) and are therefore assigned an insulating Neumann (second kind) boundary condition. Consequently, the resulting convection and radiation losses from the edge of the heated region to the end of the specimen are assumed to be between the faces of the specimen and the surrounding ambient environment. These assumptions simplify the 198

215 subsequent heat transfer analysis to a 2D approximation of heat losses from the faces of the sample outside the gage section. The heat transfer from each face of the specimen per unit area do to free convection Qconv can be written as: Q conv = h(t T amb ) (18) where h is a given uniform heat transfer coefficient, T is the temperature at a given location on the sample surface and Tamb is the ambient temperature of the surroundings. The heat loss due to radiation per unit area of the front and back specimen faces is defined as: Q rad = εσ(t 4 4 T amb ) (19) where ε is the composite surface emissivity and σ is the Stefan-Boltzmann constant. Therefore, the general for of the heat equation can be written as: Dc p t c T t kt c ( 2 T x T y 2) + 2Q conv + 2Q rad = 0 (20) where the first term represents the time rate of change of thermal energy of the medium in terms of material density D, specific heat cp and thickness tc (note that as written this assumes the material is isotropic). However, when there is no change in the amount of energy storage (i.e. steady-state condition) this term goes to zero. The second term represents net conduction along the length of the composite assuming a 199

216 constant thermal conductivity k, and the final two terms represent the heat losses due to convection and radiation respectively (the factor of two on these terms accounts for the losses of both the front and back faces of the sample). In terms of the steady-state temperature profile, the equation can be rewritten as follows: kt c 2 T + 2hT + 2εσT 4 4 = 2hT amb + 2εσT amb (21) Note that because the radiation losses are proportional to temperature to the fourth power, the governing partial differential equation (PDE) shown above is nonlinear. The resulting PDE could be quite difficult to solve analytically because of this nonlinearity and coupling with other modes of energy transfer. Therefore, a numerical approach is often chosen in order to solve for the temperature distribution. Specifically a nonlinear PDE solver based on a damped Newton iteration method was used in order to numerically solve for the temperature field under given material and environmental parameters [113]. A.1 Establishment of temperature dependent heat transfer parameters Because there is no forced air cooling on the sample, convection heat transfer is driven by the free (natural) convection of warm air ascending from the hot surface of the horizontal samples being replaced by descending cooler fluid from the ambient. The coefficient to be used in calculation of natural convection losses is approximated by considering the buoyancy-driven flow at the surface of the constant heated-region. An average heat transfer coefficient h can be expressed in terms of Nusselt number Nu L as: 200

217 h = k L Nu L (22) where k is the thermal conductivity of the fluid medium (air) and L is the characteristic length of the heated surface. Empirical correlations for external free convection have been determined in which Nu L is expressed in terms of the dimensionless Rayleigh number, Ra L. A correlation for convection flows from the surface of horizontal flat plates was determined for Ra L > 200 as [114]: Nu 1 4 L = 0.59Ra L (23) where the Rayleigh number is the product of the Grashof and Prandtl numbers and is defined as: Ra L = gβ(t sur T )L 3 να (24) where Tsur and T are the surface and ambient temperatures respectively, g is the acceleration due to gravity, and β, ν and α are thermal expansion coefficient, kinematic viscosity and thermal diffusivity of air evaluated at the film temperature, T f = (T sur + T ) 2. It has been shown that improved accuracy of these correlations can be obtained by considering the form of the characteristic length as the ratio [114]: L = A s P (25) where As and P are the plate surface area and perimeter, respectively. Therefore, by using the thermophysical properties of air at ambient pressure [111] and the characteristic length of a nominal tensile specimen with surface dimensions of mm x 12.7 mm, it is possible to estimate the free convection coefficient of a horizontal sample as a function of surface temperature by the solution of Eq. 22 in terms of Eqs. 201

218 Results of this analysis are plotted in Figure 52 for 200 C < Tsur < 1300 C. Note that the use of this correlation will lead to a slight overestimation of the convection losses. This is due to the fact that Eq. 23 was established from empirical data found for a constant temperature (isothermal) plate, whereas our sample experiences a temperature drop outside the constant temperature heated section. The solution of Eq. 21 also requires knowledge of the temperature dependent emissivity of the sample. It is known that semiconductive materials have relatively high emissivity values, and literature data on the temperature dependence of SiC is readily available [112]. The emissivity values used in analysis are shown in Figure 52. While emissivity depends on the nature of the surface and sample microstructure, this should be a reasonable approximation for the purposes of this study. Finally, it is necessary to establish the temperature dependence of the thermal conductivity to be used in the solution of the heat equation. In-plane thermal conductivity data for SiCf/SiC composites has been reported for both room and high temperature (1200 C) [7]. A linear fit to these engineering estimates has been produced and is shown in Figure 52. This trend should be a responsible estimate as the data refers to similar slurry melt-infiltrated SiCf/SiC composites. Furthermore, this data was established for in-plane data as opposed to the more commonly reported transverse conductivity. 202

219 Figure 52: Temperature dependencies of thermal conductivity (k), natural convection coefficient (h), and emissivity (ε) used in the numerical solution of longitudinal thermal gradient. 203