A Study of Failure Development in Thick Thermal
Barrier Coatings
Karin Carlsson
LITH-IEI-TEK--07/00236--SE
Examensarbete
Examensarbete LITH-IEI-TEK--07/00236--SE
A Study of Failure Development in Thick Thermal
Barrier Coatings
Karin Carlsson
Handledare: Håkan Brodin
SIEMENS Industrial Turbomachinery AB
Examinator: Sten Johansson
IEI, Linköping University
Avdelning, Institution
Division, Department
Division of Engineering Materials
Department of Management and Engineering Linköpings universitet
SE-581 83 Linköping, Sweden
Datum Date 2007-12-06 Språk Language Svenska/Swedish Engelska/English Rapporttyp Report category Licentiatavhandling Examensarbete C-uppsats D-uppsats Övrig rapport
URL för elektronisk version
http://www.ikp.liu.se/kmt/ http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-10397 ISBN — ISRN LITH-IEI-TEK--07/00236--SE
Serietitel och serienummer
Title of series, numbering
ISSN
—
Titel
Title
A Study of Failure Development in Thick Thermal Barrier Coatings Studie av skadeutvecklingen hos tjockt termiskt barriärskikt
Författare
Author
Karin Carlsson
Sammanfattning
Abstract
Thermal barrier coatings (TBC) are used for reduction of component temperatures in gas turbines. The service temperature for turbines can be as high as 1100oC and the components are exposed to thermal cycling and gases that will cause the component to oxidize and corrode. The coatings are designed to protect the substrate material from this, but eventually it will lead to failure of the TBC. It is important to have knowledge about when this failure is expected, since it is detrimental for the gas turbine.
The scope of this thesis has been to see if an existing life model for thin TBC also is valid for thick TBC. In order to do so, a thermal cycling fatigue test, a tensile test and finite element calculation have been performed. The thermal cycling fatigue test and finite element calculation were done to find correlations between the damage due to thermal cycling, the number of thermal cycles and the energy release rate. The tensile test was preformed to find the amount accumulated strain until damage.
The thermal cycling lead to failure of the TBC at the bond coat/top coat interface. The measurment of damage, porosity and thickness of thermally grown oxide were unsatisfying due to problems with the specimen preparation. However, a tendency for the damage development were seen. The finite element calculations gave values for the energy release rate the stress intensity factors in mode I and mode II that can be used in the life model. The tensile test showed that the failure mechanism is dependent of the coating thickness and it gave a rough value of the maximum strain acceptable.
Nyckelord
Keywords Thick Thermal Barrier Coatings, Thermal Cycling Fatigue, Tenile Test with Acoustic Emission, Failure Development
Abstract
Thermal barrier coatings (TBC) are used for reduction of component tempera-tures in gas turbines. The service temperature for turbines can be as high as 1100oC and the components are exposed to thermal cycling and gases that will
cause the component to oxidize and corrode. The coatings are designed to protect the substrate material from this, but eventually it will lead to failure of the TBC. It is important to have knowledge about when this failure is expected, since it is detrimental for the gas turbine.
The scope of this thesis has been to see if an existing life model for thin TBC also is valid for thick TBC. In order to do so, a thermal cycling fatigue test, a ten-sile test and finite element calculation have been performed. The thermal cycling fatigue test and finite element calculation were done to find correlations between the damage due to thermal cycling, the number of thermal cycles and the energy release rate. The tensile test was preformed to find the amount accumulated strain until damage.
The thermal cycling lead to failure of the TBC at the bond coat/top coat in-terface. The measurment of damage, porosity and thickness of thermally grown oxide were unsatisfying due to problems with the specimen preparation. However, a tendency for the damage development were seen. The finite element calculations gave values for the energy release rate the stress intensity factors in mode I and mode II that can be used in the life model. The tensile test showed that the failure mechanism is dependent of the coating thickness and it gave a rough value of the maximum strain acceptable.
Preface
This thesis has been done for SIEMENS Industrial Turbomachinery AB in Fin-spång and is the final part of my education to achieve a Master’s degree in Science in Mechanical Engineering at Linköping University. The project was carried out during the summer and fall of 2007 at Linköping University and at SIEMENS.
I would like to thank my supervisors, professor Sten Johansson at Linköping Uni-versity and Håkan Brodin at SIEMENS. I would also like thank Annethe Billenius and Bo Skoog for assistance in the laboratory, Xin-Hai Li and Johan Moverare at SIEMENS for assistance with test equipment and all the others at the department of engineering materials for good advices and Friday-fika. I would also like to give a special thanks to my Jonas for giving me support at all times.
Linköping, 2007-11-12
Karin Carlsson
Contents
1 Introduction 3 1.1 SIEMENS . . . 3 1.2 Gas Turbines . . . 3 1.3 Problem Description . . . 6 2 Theory 9 2.1 Superalloys . . . 9 2.2 Coatings . . . 9 2.2.1 TBC . . . 10 2.3 Coating Materials . . . 12 2.3.1 Zirconia . . . 12 2.3.2 MCrAlY . . . 13 2.4 Coating Deposition . . . 13 2.4.1 Plasma Spraying . . . 142.5 Oxidation and Corrosion . . . 14
2.5.1 Pilling-Bedworth Ratio . . . 15
2.6 Failure Mechanism of TBC . . . 16
3 Experiment 19 3.1 Testing . . . 19
3.1.1 Thermal Cycling Fatigue . . . 19
3.1.2 Tensile Test with Acoustic Emission . . . 21
3.2 Specimen Preparation . . . 21
3.2.1 Traditional Specimen Preparation . . . 22
3.2.2 Ion Etching . . . 23
3.3 Evaluation . . . 23
3.3.1 SEM and LOM . . . 23
3.3.2 Graphics Processing . . . 24
3.3.3 Measurements . . . 25
3.4 Finite Element Calculations . . . 26
4 Results and Discussion 29 4.1 Thermal Cycling Fatigue . . . 29
4.1.1 Specimen Preparation . . . 30
4.1.2 Evaluation of Failure Mechanism . . . 31
x Contents
4.1.3 Evaluation of Porosity . . . 31
4.1.4 Crack Evaluation . . . 33
4.1.5 Evaluation of Thermally Grown Oxides . . . 33
4.1.6 Diffusion . . . 35
4.2 Finite Element Calculations . . . 37
4.3 Tensile Test with Acoustic Emission . . . 37
4.3.1 Failure Mechanisms . . . 38
5 Conclusions and Future Work 41 5.1 Conclusions . . . 41
5.2 Future Work . . . 42
Bibliography 43 A Appendix 45 A.1 Crack Evaluation . . . 45
A.2 Finite Element Calculations . . . 48
A.3 Tensile Test with Acoustic Emission . . . 51
List of Figures
1.1 Industrial gas turbine, SGT800. Copyright SIEMENS Industrial Turbomachinery. . . 4 1.2 A Brayton cycle which consist of a gas compressor, a combustion
chamber and a turbine. It is run in an open cycle. . . 5 1.3 Temperature/entropy and pressure/volume diagrams for an ideal
Brayton cycle. . . 5
2.1 The thermal gradient in a TBC system. . . 10 2.2 A duplex TBC system. In thick TBC the top coat is between 1-1,5
mm. The thermally grown oxide is less than 10 µm and the bond coat is approximatly 0,15 mm. . . 11
3.1 Thermal cycling fatigue. The specimen-table shifts from the furnace to air-cooling. . . 19 3.2 The cooling part of a recorded thermal cycle. . . 20 3.3 The specimen used at thermal cycling fatigue test. . . 20 3.4 The tensile test. The arrows points at the microphones which is
taped to the specimen holder. Attached to the specimen is a device measuring the strain. . . 21 3.5 The tensile test specimen. . . 22 3.6 Ion etching. An ion beam makes a cut in the specimen. . . 23 3.7 A picture used for evaluation of the porosity. Measurements were
made in three different areas, high, middle and low. . . 25 3.8 The crack length was measured as the horizontal ”image” of the
real crack. . . 26 3.9 The displacement of the nodes in the circle is used for calculation
of a relation of KI and KII. . . 27
3.10 The right picture shows the modeled undulation and the left picture shows the same part during loading. . . 27
4.1 A cut done in the ceramic layer by ion etching. . . 31 4.2 The as-coated specimen after ion etching. The ceramic layer is at
the bottom and the substrate at the top. The bond coat appears to have two shades; the darker one is cut once more than the brighter one. The cut was done from the ceramic down to the substrate. . . 32 4.3 The specimen has been cycled 300 times. The crack is an example
of a mixed failure. At number 1 the crack is propagating in the ceramic and at 2 in the thermally grown oxide. . . 32 4.4 A crack or a pull-out in a specimen cycled 300 times. . . 34 4.5 The results from all three crack measurment. The measurments are
represented by the dots and the line is a trendline. . . 34 4.6 There is no oxide between the bond coat and top coat. . . 35 4.7 The results from the measurements of the thermally grown oxide.
The line is a trendline based on the measurements diplayed as the dots. . . 36
2 Contents
4.8 The results from the mesurements of the thermally grown oxide.
The two lines represent the average and median values. . . 36
4.9 The stress intensity factor in mode I and II for a model with 6 µm thick thermally grown oxide. . . 38
4.10 The failure of a tensile specimen with 1000 µm thick coating. There are both transverse and interface cracks. . . 39
4.11 Spallation of the coating. Both transverse and interface cracks are visible. . . 40
A.1 Measurement 1. . . 45
A.2 Measurement 2. . . 46
A.3 Measurement 2. . . 46
A.4 The three different measurments. . . 47
A.5 Energy release rate for different oxide thicknesses. . . 48
A.6 The stress intensity factor in mode I for different oxide thicknesses. 48 A.7 The stress intensity factor in mode II for different oxide thicknesses. 49 A.8 The energy release rate for different oxide thicknesses. . . 49
A.9 The stress intensity factor in mode I for different oxide thicknesses. 50 A.10 The stress intensity factor in mode II for different oxide thicknesses. 50 A.11 500 µm test 1. . . . 51 A.12 500 µm test 2. . . . 51 A.13 1000 µm test 1. . . . 52 A.14 1000 µm test 2. . . . 52 A.15 1500 µm test 1. . . . 53 A.16 1500 µm test 2. . . . 53
A.17 K0. Analysis started in the top coat. . . 54
A.18 The result from the line scan. Spectrum (x) represent the point where the measurement was done. . . 54
A.19 K300. Analysis started at the top of the bond coat. . . 55
A.20 The result from the line scan. Spectrum (x) represent the point where the measurement was done. . . 55
A.21 K450. Analysis started at the top of the bond coat. . . 56
A.22 The result from the line scan. Spectrum (x) represent the point where the measurement was done. . . 56
A.23 A internal bond coat oxide from K450. . . 57
A.24 The results from the line scan. The different graphs represent dif-ferent materials oxygen, silicon, chromium, cobalt, nickel, yttrium and aluminum. . . 57
List of Tables
3.1 The procedure used for specimen preparation. . . 224.1 The measured porosity. High, middle and bottom indicates where on the ceramic the porosity is measured. . . 33
Chapter 1
Introduction
This chapter gives a brief introduction to SIEMENS Industrial Turbomachinery and some basic theory of gas turbines. A problem description is also given.
1.1
SIEMENS
SIEMENS Industrial Turbomachinery AB in Finspång produces steam and gas turbines, and engines for compressors and pumps in the oil and gas industry. The history of the plant in Finspång goes back to 1631 when Louis De Geer bought Finspong Bruk. The company De Geer made canons until 1911 and in 1913 the company was sold to STAL who started to produce steam turbines. In the 1940th STAL developed three jet engines on commission for the Swedish Air Force, but the Air Force chose a foreign engine. One of the jet engines was then developed into a stationary gas turbine that generated 10 MW. It was launched 1955. During the years, the plant has had different names. In the 1950th it became Stal-Laval after merging with Gustav De Laval’s De Laval Ångturbin AB, who was the first to construct turbines in Sweden. Today it is a part of SIEMENS.
Four types of gas turbines are produced in Finspång; 500, 600, SGT-700 and SGT-800, SGT-500 being the smallest and SGT-800 the largest. They are in the range of 15 to 50 MW.
1.2
Gas Turbines
Gas turbines can be divided into two groups; airborne and land-based. The air-borne turbines are used for airplane propulsion and a common application for the land-based turbines are the electric power generation. In the base-load electric power generation, which has been dominated by coal and nuclear power plants, gas turbines are being installed more and more. This can be explained by the gas turbines higher efficiency, lower capital cost, shorter installation time, better emission characteristics and the abundance of natural gas supplies [1]. Gas tur-bines are also used for peaking power plants. An industrial gas turbine is depicted
4 Introduction
Figure 1.1. Industrial gas turbine, SGT800. Copyright SIEMENS Industrial
Turboma-chinery.
in figure 1.1. Thermodynamically a gas turbine can be explained by the Brayton cycle, also known as the Joule cycle. The cycle consists of three components:
• A gas compressor (the compressor)
• A mixing chamber (the combustion chamber) • An expander (the turbine)
Air is drawn in to the air inlet (1). When the air reaches the compressor (2) the temperature and the pressure are raised. The high-pressure air then proceeds to the combustor (3) where the air is mixed with fuel and ignited, resulting in high-temperature gases. The gases flow through the turbine (4) where it expands to atmospheric pressure, producing mechanical energy. This energy is used to give the generator shaft (5) a torque and to drive the compressor. The exhaust is discharged through the exhaust outlet (6). Either the exhaust gas goes through exhaust treatments (run in an open cycle) or it is used to heat water in a steam turbine (combined cycle process). Figure 1.2 shows a simplified model of the Brayton cycle. The thermodynamic states for the gas are shown in figure 1.3. The medium is considered to be an ideal gas. Numbers 1-4 in figure 1.2 corresponds to the same numbers in figure 1.3. Between 1-2 and 3-4 the gas is compressed with no change in entropy (isentropic compression). Between 2-3 and 4-1 the pressure is kept constant, but between 2-3 there is an addition of heat while between 4-1 there is a rejection of heat. The thermal efficiency can be written as follows
ηth,Brayton= wnet qin = 1 −qout qin = 1 −cp(T4− T1) cp(T3− T2) (1.1)
where w is work per unit mass, q is heat transfer per mass unit, cp is constant
pressure specific heat and T is temperature. To increase the thermal efficiency in a gas turbine, the pressure ratio and/or specific heat ratio of the working fluid can be increased. But an increase in pressure means an increase in energy spent
1.2 Gas Turbines 5
Figure 1.2. A Brayton cycle which consist of a gas compressor, a combustion chamber
and a turbine. It is run in an open cycle.
Figure 1.3. Temperature/entropy and pressure/volume diagrams for an ideal Brayton
6 Introduction
on compression and that is a loss unwanted in the power generation. So therefore an increase of the turbine inlet temperature, T3, is the better solution. But it
is limited by the material in the turbine and combustor. Development of new materials for the turbine has increased the turbine inlet temperature from 540◦C in the 1940s to 1425◦C today [1]. The materials that make this possible are superalloys and coatings.
1.3
Problem Description
An industrial gas turbine is designed to be in service for about 120 000 hours (∼13 years). It can run on maximum load for months without stop and the inspection intervals are long, at least 10 000 hours. This requires reliable life models.
There is a thermal insulating and corrosion/oxidation resistant coating, called thermal barrier coating (TBC), deposited onto hot parts of gas turbines. The life model for this coating is based upon Paris’ law, a relation between crack growth rate and data for the fracture mechanism
da
dN = C(∆K)
n (1.2)
where a is the crack length, N number of cycles, K the stress intensity factor (in mode I) and C and n are constants depending on temperature, material, environ-ment, frequency and stress ratio. The relation has to be modified to suit the life model. A parameter describing the amount of damage in the coating is used in-stead of the crack length since this allows for a better coupling between modelling results and experimental data. Due to the nature of the TBC, crack growth in the system does not only occur in mode I, but also in mode II, mostly a mixture of the two. The interface where failure often occur has an undulated shape. At the ridges mode I is dominating, at the valleys mode II and in between there is a mixture of the two. When the crack is growing in mode I the crack surfaces move directly apart from one another and in mode II the surfaces slides over each other in an in-plane shear mode. Mode II results in friction leading to more energy needed to open the crack than in mode I. Using this assumption, the parameter for the stress intensity factor can be changed to a parameter δ describing the mode and a parameter ∆G which is the energy release rate describing the energy released at crack opening. The model needs to be calibrated with input data obtained from experiments and finite element calculations. The existing life model for TBC is modified for a thin coating, ∼300 µm. Nowadays are also thicker coatings, 1-1,5 mm, used. SIEMENS are therefore interested to find out if their existing life model could work also for thick TBC. This means that new input data has to be acquired but also that a study of the failure mechanism has to be done. In a thin TBC failure often occur near the coating interface and therefore the model is based on this type of cracking. If the cracking in a thicker coating would behave differently, a new life model might be needed.
The goal of this thesis has been to study the failure mechanism of thick TBC and find the input data needed, which are:
1.3 Problem Description 7
• A correlation between total damage of the coating, due to thermal fatigue,
and the number of cycles.
• A correlation between total damage and energy release rate. • The amount accumulated strain until damage of the coating.
The correlation between damage and number of cycles, together with the study of failure mechanisms, was obtained by a thermal cycling fatigue test. The correlation between damage and energy release rate was modeled in a finite element program. The strain limit was obtained through a tensile test with acoustic emission.
Chapter 2
Theory
This chapter covers the theory of the thermal barrier coatings and also gives a background to why coatings are used.
2.1
Superalloys
Superalloys are materials that can be used at service temperatures over 540◦C. Some can be used at temperatures as high as 80% of their melting temperature [2]. They are nickel, iron-nickel or cobalt based alloys with large amounts of alloying elements added in order to create a combination of high strength and resistance to creep and corrosion at elevated temperatures. In general, the melting temperature for superalloys is in the same range as the melting temperature of steel. For gas turbines, nickel-based alloys are the most suitable due to its excellent mechanical properties at high temperatures. The high temperature strength in superalloys is based on a stable face-centered cubic matrix that is combined with either precipi-tation hardening or solid-solution hardening. Generally, the superalloys consist of an austenitic (γ-phase) matrix and several secondary phases. The most common of secondary phases are metallic carbides and γ’, which is the ordered face-centered cubic-phase in iron-nickel and nickel-based alloys. Solid-solution hardened superal-loys withstand higher temperature better than precipitation hardened superalsuperal-loys. Their mechanical properties are more stable while the properties changes for pre-cipitation hardened superalloys. This is because at high temperature precipitates will grow at dislocations and the carbides will coarsen. In this thesis a wrought solid-solution hardened nickel-base superalloy has been studied.
2.2
Coatings
The superalloys are optimized for load-carrying capability, with less concern for environmental resistance. Therefore coatings are used, to protect superalloys from environmental attacks such as oxidation and/or corrosion. Coatings are tailored for their specific application. For example, a coating with aluminum is desired
10 Theory
if the material is subjected to oxygen, and a coating with chromium if subjected to a corrosive environment. The coating can either be a diffusion coating where the surface of the superalloy component has been enriched with elements that are protective against corrosion and oxidation (chemical vapour deposition type processes), or an overlay coating where a layer is deposited onto the component (physical deposition processes). In this thesis an overlay coating produced by plasma spraying has been studied.
2.2.1
TBC
As mentioned earlier, an increase in gas turbine inlet temperature is desired in order to increase the thermal efficency. To be able to do so, a thermal insulating coating is needed since the superalloys has reached their upper thermal limit. This coating together with an oxidation/corrosion resistant coating is called a thermal barrier coating (TBC) system. It is deposited on the hot-path components of gas turbines, such as combustionliners turbines vanes and blades. TBC systems can also increase the efficiency without increasing the inlet temperature by decreasing the need of air cooling. If both the air cooling and the inlet temperatures are the same, it will increase the reliability of the component, while the decrease in temper-ature will increase the creep resistance and the resistance to thermal/mechanical fatigue of the load carrying component. The TBC cannot compensate the air cooling completely. In order to produce a high temperature gradient, to reduce the temperature of the metal substrate, rear-side cooling is required. Figure 2.1 shows a typical temperature gradient in a TBC system. A 0.6-0.7 mm ceramic layer can reduce the substrate temperature with more than 200oC [3].
Figure 2.1. The thermal gradient in a TBC system.
Composition
The most common TBC system is the duplex system. It consists of two layers; one outer ceramic layer that insulates and one intermediate metallic layer that protects
2.2 Coatings 11
the substrate from oxidation and corrosion. The metallic layer also provides an adhesion surface for the ceramic layer. The ceramic coat is called the top coat and the metallic is called the bond coat. There are also multilayered TBC systems, where there is an erosion resistant layer, a corrosion-oxidation resistant layer, a thermal stress control layer and a diffusion resistant layer [4]. A schematic view of a duplex TBC system is illustrated in figure 2.2 and is the kind used in this thesis. In the interface between top coat and bond coat one can find the thermally grown oxide.
Figure 2.2. A duplex TBC system. In thick TBC the top coat is between 1-1,5 mm.
The thermally grown oxide is less than 10 µm and the bond coat is approximatly 0,15 mm.
The bond coat is sprayed onto the substrate, most commonly, with plasma spray-ing, explained more in section 2.4.1. It creates a surface with good adhesion against the top coat. Two other important purposes are to level out the differ-ences in thermal expansion between the substrate and top coat and to act as an aluminum-reservoir. Aluminum will react with oxygen and create the thermally grown oxide, which protects the substrate from high temperature corrosion and oxidation. The bond coat is usually made of MCrAlY, see section 2.3.2, and is approximately 50 to 125 µm [5]. Ideally the BC should be free from pores since the pores will transport oxygen, leading to formation of oxides which can be detri-mental.
The purpose of the top coat is to act as a thermal insulator. It is often made of partly yttrium stabilized zirconia (PSZ). Zirconia is suited for TBC because of its low thermal conductivity (∼1,5-2 W/m2K) and it has a coefficient of thermal
expansion (6-8·10−6/K) to be compared to the substrates (∼13-14·10−6/K). The top coat is allowed to have a certain porosity since it decreases the heat conduc-tivity but also results in a reduction of stiffness. The thickness of the top coat also affects the reduction of heat, for example, using a thick TBC instead of thin ones could reduce the amount of cooling air by 60% [3]. The thickness can vary from 100 to 2000 µm.
12 Theory
The thermally grown oxide is formed during high temperature exposure. It mainly consist of alumina, Al2O3. Oxygen diffuses through the permeable top coat and
reacts with the aluminum, which has the highest affinity to oxygen, in the bond coat and forms the alumina. Oxygen transport through aluminum is very slow and therefore provides an oxidation protection for the metal substrate. The thick-ness of the thermally grown oxide is connected to the life time of the TBC system and is about 0,5 to 10 µm, depending on thermal exposure (time and temperature).
The substrate will sustain the structural load.
2.3
Coating Materials
The materials in the TBC must have a high melting point, no phase transformation between room temperature and operations temperature, low thermal conductivity, chemical inertness, a thermal expansion that match the thermal expansion of the metallic substrate, good adherence to the metallic substrate and low sintering rate of the porous microstructure [4]. Rare earth elements, Ti, Zr, Hf, Al and Si are elements whose oxides suits TBC applications, but the most widely used and studied material is ZrO2. It performs very well in high temperature applications,
such as gas turbines. A typical material for the bond coat is MCrAlY.
2.3.1
Zirconia
Zirconia, ZrO2, is a ceramic. Ceramics are built-up by ionic bonding between
an-ions and catan-ions. Anan-ions are generally larger than catan-ions. Therefore the ceramic can be seen as a close-packed structure of anions with cations in the interstitial sites. But for zirconia, the cation (Zr+) is larger than the anion (O−), so here it
is the cation that forms the close-packed structure with anions in the interstitial sites. Zirconia has a high diffusion coefficient for oxygen in the cubic phase, while oxygen vacancies are introduced to compensate for the dopants lower valency [2]. Therefore, it is important with an oxidation resistant coating underneath the top coat. Zirconia is an allotropic material. Allotropy (pure materials) and polymor-phism (alloys) materials are materials that have more than one crystal structure. At room temperature the stable structure for zirconia is monoclinic. At higher temperatures, more symmetric structures are stable. At 1240◦C the monoclinic transform into the tetragonal structure that is stable up to 2370◦C, where it trans-forms to cubic. It remains cubic until it melts at 2680◦C [6]. Under high pressure zirconia can form an orthorhombic structure. The reason why the ceramic under-goes transformation is that temperature or pressure on the material can change the interatomic distance and the atom vibration, and the initial structure may no longer be the most stabile one, leading to transformation to a more stable structure.
2.4 Coating Deposition 13
Stabilized Zirconia
A change in structure means a change of the crystal structure’s size. This leads to a change in volume, which is detrimental for components of pure zirconia. They tend to fail when the temperature is dropped and the tetragonal structure transforms into monoclinic, which means an increase in volume. To prevent this, dopants, such as yttria (Y2O3), are added. They stabilize the cubic phase, even at room
temperature. Yttria stabilized zirconia (YSZ) contain 7wt% yttria, which is the optimum amount [5]. Other dopants are MgO and CaO.
Crack Shielding
When a load is applied and a crack is growing, the metastable tetragonal phase at the tip of the crack will transform into monoclinic. The increase in volume will lead to a compressive stress at the crack tip and thereby inhibit crack growth. Additional load is then required for further crack growth. This only applies for stabilized zirconia, not pure zirconia. However, if too much yttria is added the cubic structure becomes too stable to undergo a tranformation.
2.3.2
MCrAlY
M stands for nickel and cobolt (and iron if applied on steel) or a combination of the two. Chromium and aluminum are added for creating protective oxides, Cr2O3 and Al2O3. Yttrium is improving the adhesion of the thermally grown
oxide [7]. NiCrAl undergoes a phase transformation at 1000oC. If cooled, β + γ
will transform to α + γ’. Cobalt is added to stabilize this transformation. 20 to 26wt% cobalt will also improve the ductility [5]. In a NiCrAl alloy system is the
γ phase a face-centered cubic Ni structure with occasional Al and Cr atoms. γ’ is
a face-centered cubic structure with Al or Cr as corner atoms and Ni as cube faces atoms. The β phase, NiAl, is a body-centered cubic (bcc) structure with Ni as corner atoms and Al as center atom. The α phase is α -Cu which is body-centered cubic. The microstructure of the MCrAlY consists of a β phase in a γ matrix. The
β/γ ratio decides the resistance against oxidation and corrosion, phase stability
and the resistance against cracking [5].
2.4
Coating Deposition
There are two common ways to deposit the overlay coatings of the TBC sys-tem: plasma spraying and electron beam physical vapor deposition (EB-PVD). In plasma spraying a molten powder is sprayed upon the component. EB-PVD is done by vaporizing a raw material with a high-energy electron beam. The vapor is then deposited onto the component. The two processes results in two different structures. The sprayed coating will have a structure consisting of molten particles in the shape of pancakes, pores (the amount depends on the powder size), trans-verse microcracks and, depending on spray technique, a thermally grown oxide. EB-PVD coatings have a columnar microsturcture. It has a finer structure with
14 Theory
no porosity, but it has higher heat conductivity due to its columnar structure. It is also more costly than plasma spraying. In this thesis air plasma sprayed specimens are used.
2.4.1
Plasma Spraying
The top coat and bond coat are deposited on the substrate by plasma spraying. A prealloyed powder is injected into a high-temperature plasma gas stream in a plasma-spray gun. The molten powder is then deposited with high velocity onto the substrate. The powder particles solidify on the substrate surface, creating a coating. When the droplets hit the substrate surface they form ”splats”, leading to porosity parallel to the surface. Plasma spraying is performed with temperatures as high as 10 000 - 20 000◦C. It can be done in vacuum (vacuum-plasma spraying) or in an inert gas of low-pressure (low-pressure-plasma spraying) to avoid the powder to react with oxygen. There is also air-plasma spraying, which is the method used for both the top and bond coat of the TBC in larger parts of the industrial gas turbines. Spraying in vacuum or inert gas would seem to be a better choice but the components are too big to be placed in a vacuum chamber or similar. The size of the powder used is around 40 µm. A larger powder will have problem melting and a powder finer than 10 µm will not be able to penetrate the plasma and will therefore also have problem melting. A fine powder gives a more dense coating than a coarser powder, since the pores between the ”splats” are relatively small for the finer powder. On the other hand, a fine powder does not give the coarser surface that is desired for good adhesion between bond coat and top coat. So in order to get a dense coating with an adherent surface, a fine powder is used to build up the coating. Then a coarser powder is sprayed upon the finer powder to create the undulating surface. The size of the powder also decides if there is a chemical or a mechanical bonding between the top and bond coat. A finer powder means a chemical bonding while a coarser gives the undulating surface which gives the mechanical bonding.
2.5
Oxidation and Corrosion
During operation the gas turbine will be exposed to gases causing both corrosion and oxidation. These phenomena can be divided into three groups:
• Type II Hot Corrosion. • Type I Hot Corrosion. • Oxidation.
The hot corrosion is a result of accelerated oxidation at 680-1050◦C [2]. The alloy, or in this case the coating, will be covered with thin films of salts, typically alkali and alkaline earth sulfates, which will react with the material. Generally, Type I hot corrosion occurs above the melting point of these salts and type II hot corrosion below. Corrosion has two stages; the initiation stage where there is a
2.5 Oxidation and Corrosion 15
breakdown of the protective oxide and the propagation stage where the salts have access to unprotected metal and corrosion will occur at a high rate. As mentioned earlier, chromium is most effective in giving a resistance to hot corrosion since Cr2O3 has a faster growing rate than Al2O3, meaning that it will reform faster
in case of a breakdown of the oxide scale due to hot corrosion. In the case of corrosion of the top coat, molten sulfates will react with the stabilizer. MgO is the most reactive one and Y2O3the least one, hence Y2O3 being more commonly
used. Reaction between the sulfates and the stabilizer leads to a ceramic more prone to phase transformation leading to premature failure of the top coat when thermally cycled.
Above 900◦C, oxidation will occur. The top coat will not react with the oxygen, but it has a high diffusion coefficient for oxygen (Zr in cubic phase) leading to oxidation of the bond coat. Islands of oxide will grow and form an oxide film between the top coat and bond coat (the thermally grown oxide). Also internal oxidation of the bond coat will take place. The oxide film will continue to grow by oxygen and/or metal ion diffusion. In the MCrAlY coat, alumina will grow with oxygen diffusion through the oxide, while nickel, chromium and cobalt will diffuse through the alumina to react with oxygen. Some metals react more easily with oxygen than other. This depends on the free energy for oxide formation. The larger the negative free energy is, the easier it is for a metal to form an oxide. Aluminum reacts more easily than nickel, chromium and cobalt. The characteristics of the oxide growth can be explained by the Pilling-Bedworth Ratio.
2.5.1
Pilling-Bedworth Ratio
The Pilling-Bedworth ratio (PBR) describes how much of a metal that has been oxidized. (PBR = volume of oxide formed / volume of metal consumed.) If the ratio is less than one, PBR < 1, then the oxide takes up less volume than the metal, from which it is formed. This gives a coating with porosity and the oxidation will continue rapidly. When the ratio is between one and two, 1 6 PBR 6 2, the volume of the oxide and the metal is almost the same. This gives an adherent and non-porous oxide. The oxide will work as a protecting film, for example aluminum and titanium. And when the ratio is greater than two, PBR > 2, the oxide will take up a large share of the volume. This can lead to the oxide flaking off and non-oxidized material will be exposed. In the TBC a ratio between one and two is wanted in order to have a protective coating. The mathematical expression for oxide growth is also connected to the PBR. For a porous oxide (PBR<1) the equation for oxide growth can be written as:
y = kt (2.1) where y is the thickness of the oxide, t time, k a constants depending on the temperature, environment and the composition. A non-porous (PBR>2) oxide gives the equation
y =√kt (2.2)
And a thin-oxide film (16 PBR 6 2)
16 Theory
where c is a constant.
2.6
Failure Mechanism of TBC
The gas turbine hot components will be exposed to high temperatures, ther-mal/mechanical cyclic loading and an environment causing both corrosion and oxidation. This will eventually lead to failure of the TBC. In the case of air plasma sprayed TBC, that means spallation of the top coat [5]. These failures will occur if either the stress increases, the material strength decreases, or a combina-tion of the two. The stress will increase due to mismatch in thermal expansion, temperature gradient and growth of the thermally grown oxide. The material can loose strength due to sintering at high temperatures, propagation of cracks and depletion of aluminum which will lead to growth of more brittle oxides.
The metallic and ceramic layers have a mismatch between their coefficients of thermal expansion. This mismatch will cause compressive residual stresses upon cooling. According to A.G. Evans et al. [7] these stresses are as big as 3-6 MPa. The stresses are redistributed around imperfections leading to both compressive and tensile stresses. They will be detrimental when exceeding the yield strength. During isothermal service these stresses will decrease as a result of relaxation and creep. There is a temperature gradient trough the TBC system, which will lead to different thermal expansions, causing internal stresses. As in the case above, these stresses will decrease at high temperature exposure. Thick TBC systems have a steeper temperature gradient than thin ones, and are therefore more sus-ceptible for this phenomenon [8]. The growth of the thermally grown oxide induces compressive stresses, generally less than 1 GPa [7]. This oxide is, as mentioned earlier, mostly made up by alumina. But due to aluminum depletion in the bond coat, other oxides called spinels are formed. These oxides are larger than the alu-mina and will therefore contribute to even higher compressive stresses. Since they are more brittle than alumina, the strength of the coating decreases. The major reasons for aluminum depletion are the formation of thermally grown oxides and diffusion into the substrate material. At high temperatures (above 1200◦C) the ceramic may sinter. It means a decrease of microcracks and pores, changing the materials stiffness leading to a change of fracture properties. There is an amount of microcracks in the ceramic top coat in the as-coated material. These cracks are beneficial since they reduce both the stiffness and the thermal conductivity. During service the cracks will propagate and formation of new cracks will be the result. This will with time lead to failure and delamination of the coating. The new cracks will start from locations where the stresses are large enough to ex-ceed the yield strength, for example at imperfections or due to thermal expansion mismatch. There are three types of failures
• White failure. This fracture occurs in the ceramic. It is called white failure
while the ceramic fracture surface appears white.
• Black failure. Which is a failure occurring in the thermally grown oxide.
2.6 Failure Mechanism of TBC 17
• Mixed failure, which is a mixture of black and white fractures.
According to D. Renusch and M. Schütze, who have studied thin TBC failure with acoustic emission, cracking predominantly occurs during cooling. They did not detect any cracking during heating and they mean that this shows that it is the thermal expansion mismatch stress, developed between the ceramic and metallic layers, that is the driving force for cracking [9].
Chapter 3
Experiment
In this chapter the experiments done will be presented and explained. Some ad-ditional background theory will also be covered.
3.1
Testing
Two tests were performed, a thermal cycling fatigue test (TCF) and a tensile test with acoustic emission. The TCF test was done to find a relationship between the number of thermal cycles and the damage due to these cycles and also to see were fracture will occur. The tensile test was performed to find out how much strain that can be applied until crack initiation and failure.
3.1.1
Thermal Cycling Fatigue
The thermal cycling fatigue, TCF, test was done in a thermal fatigue furnace, see figure 3.1, at SIEMENS Industrial Turbomachinery in Finspång, Sweden. The
Figure 3.1. Thermal cycling fatigue. The specimen-table shifts from the furnace to
air-cooling.
20 Experiment
specimens were placed on the specimen-table together with three thermocouples. The thermocouples logged the temperature inside a dummy specimen during the whole cycle. Figure 3.2 gives an example from the temperature log. The
specimen-Figure 3.2. The cooling part of a recorded thermal cycle.
table was kept inside the furnace for 60 minutes. The furnace was adjusted to a maximum temperature of 1100oC. After that the specimen-table was set under air-cooling, to a minimum temperature of 100oC for 11 minutes. During the air cooling, a dummy-table was positioned in the furnace to minimize the heat-loss. Consequently one thermal cycle is 60 minutes of heating and 11 minutes of cooling, giving a cooling rate of approximately 1.5oC/s. The test done was done for 50, 100, 150, 200, 300 and 450 (total damage) cycles. It was carried out by placing all the specimens in the furnace at the same time and then removing them one by one as they reached their predetermined number of cycles. The specimen used had the dimensions 30 x 50 x 5 mm, where the TBC system was 1,5 mm, see figure 3.3. They were coated by using air plasma spraying, at the same time as a real gas turbine.
3.2 Specimen Preparation 21
3.1.2
Tensile Test with Acoustic Emission
A tensile test with acoustic emission was preformed at Linköping University in order to find the maximum applied strain until fracture. The failure mechanism was also studied by visual inspection of the specimen during and after the test. All specimens were loaded until fracture. The test was performed with acoustic emission. Two microphones attached to the specimen holder, see figure 3.4, were used to ”hear” when damage was introduced in the material when loaded in ten-sion. When a crack is growing or is initiated, there is a release of energy which will generate an elastic wave. The wave will propagate through the material to the surface. There the microphones will detect this wave and transform it to an electrical signal that can be processed in a computer. The program used, Mistras, collected data on how many times there where energy releases (hits) and how big this release was (energy). It also recorded the strain, time and load. The tensile
Figure 3.4. The tensile test. The arrows points at the microphones which is taped to
the specimen holder. Attached to the specimen is a device measuring the strain.
specimens were coated with air plasma spraying. Three different top coat thick-nesses; 500, 1000 and 1500 µm, were used in order to see the effect of thickness on the result. 500-1500 µm represent the upper range of TBC thicknesses used in SIEMENS gas turbines. See figure 3.5 for the dimensions of the specimens.
3.2
Specimen Preparation
The preparation of the test specimens for microscopic investigation was carried out in two different ways; traditional specimen preparation and ion etching.
22 Experiment
Figure 3.5. The tensile test specimen.
3.2.1
Traditional Specimen Preparation
The specimens were first impregnated with a low-viscosity epoxy resin in vacuum. The epoxy cured under ambient conditions. This was done in order to distinguish original pores and cracks due to manufactoring from the ones produced later dur-ing specimen preparation. The test specimen was subsequently sectioned in a diamond-precision cutter and then cold mounted in epoxy. The reason for cold mounting, instead of using hot mounting, is that the specimen is sensitive to the additional stresses that are associated with the heating. The specimens were then ground and polished according to the procedure described in table 3.1, in order to reach the true structure. The idea is first to grind the specimen enough so that all material affected by the cutting is removed. Next step is to grind or polish with a finer grain size than the last step, to remove the affected material from previous step. This continues until the true structure is revealed.
Step Force Speed Lubricant Plate Medium Time
Grinding 20 N 300 rpm Water MD-Pian #200 - until plane
Diamond polishing 20 N 150 rpm Green MD-Plan 45 µm 15 min
Diamond polishing 20 N 150 rpm Green MD-Plan 15 µm 15 min
Diamond polishing 20 N 150 rpm Green MD-Plan 9 µm 15 min
Diamond polishing 20 N 150 rpm Green MD-Plan 6 µm 15 min
Diamond polishing 15 N 150 rpm Green MD-Dur 3 µm 15 min
Oxide polishing 5 N 300 rpm Water MD-Chem OP-A 20 s
3.3 Evaluation 23
3.2.2
Ion Etching
During ion etching, the specimen is bombarded with ions in order to etch or make a cut. In this case a cut with an ion beam was desired. When making a cut, a shield with a polished sharp edge is placed over the specimen, see figure 3.6, in order to make an accurate sharp cut. An advantage with ion etching over for example diamond cutting is that the cut area is free from mechanical deformations, meaning that no further specimen preparation is needed. The specimen for the ion etching machine used may not be bigger than 12 x 7 x 1,4 mm (for a 90o-stub).
It was cut in a diamond-precision cutter and then ground to the right thickness. Since the specimen only could be 1,4 mm thick and the top coat alone is 1,5 mm, grinding was done on both the substrate and the top coat.
Figure 3.6. Ion etching. An ion beam makes a cut in the specimen.
3.3
Evaluation
In order to measure and characterize porosity, cracks and oxides, studies were done on a microscopic level. Light optic microscope (LOM) and scanning electron microscope (SEM) was used together with graphics processing programs.
3.3.1
SEM and LOM
The scanning electron microscope (FEGSEM Hitachi SU70) can provide pictures with a magnification as high as x1.000.000, but only the range of x50 to x1000 were used here. The lower magnification was used when studying the whole ceramic top coat, to see differences within the coat, and the higher magnification was used when studying the thermally grown oxides and the crack path. In a SEM electrons are accelerated by an electron gun and brought together into a beam. The beam is then focused by a system of electromagnetic lenses. When the beam hit the spec-imen the electron interacts with the near-surface region and new electrons from the material are emitted, so called secondary electrons. A detector detects the secondary electrons and generates a signal to form an image. A scanning system moves the electron beam over the specimen surface building up an image point by
24 Experiment
point. Characteristic X-ray spectrums are also emitted from the specimen, which is used for chemical analyses. There are also detectors for backscattered electrons (BSE). This is all done in vacuum. The specimens used in SEM have to be con-ductive, and since the ceramic top coat and the epoxy mounting are not, they had to be coated with a thin layer of carbon. Two thin carbon rods are connected to each other by their sharp edges in a low pressure atmosphere of argon. A volt-age is applied to the carbon sticks to produce an arc emitting carbon atoms that will be deposited onto the specimen in a thin layer. The SEM was used to take micrographs to be used in the evaluation (cracks, porosity and oxides) but also to examine the material composition, by so called EDS-line scan and mapping, in order to study the oxide composition and diffusion phenomena.
The light optic microscope was also used to take micrographs of the specimens to be used in the evaluation of cracks and porosity. The range of magnification in the LOM was x5 to x100, where x20 and x40 were the magnifications used for investigating the coating. A LOM uses lenses and ocular to enlarge a picture of an object. The object can either be lit from below or from above to improve the picture. In this case the object is lit from above. A camera, connected to a computer, is attached to the microscope to ”photograph” the object.
One might find it unnecessary to use both SEM and LOM, but there are reasons for using both. One is that SEM produces pictures that appear to be more three dimensional, making it easier to see cracks. Another is that the LOM pictures was easier to export to an image processing program called MicroGOP (Contex Vision MicroGOP2000/S), which can be used for crack investigation.
3.3.2
Graphics Processing
Three programs were used for graphic processing; Adobe Photoshop CS2, Kappa Image Base and MicroGOP. Adobe Photoshop CS2 was used to stitch pictures together. The crack evaluation was done from six SEM-pictures, with the mag-nification x200, stitched together to make one ”panoramic” picture. One single picture might not display the whole crack, while a ”panoramic” picture more likely will. Pictures of the top coat were also stitched together to evaluate possible differ-ences through out the ceramic, for example differdiffer-ences in porosity, but also where cracks appear. Kappa Image Base was used when taking picture in the LOM but also to make measurements of the thermally grown oxide and the cracks. Kappa Image Base has a function that allows one to measure for example distances. The measurements are gathered in a text file that can be exported to Microsoft Office Excel for compilation. MicroGOP is a program used both for image processing and analysis. It has a program module allowing the user to write scripts that auto-matically or semi autoauto-matically analyzes pictures. Marek Jan Chalupnik wrote a program described in his thesis, for evaluation of defects within thin TBC [10]. The idea in this thesis was to use this program for the evaluation of thermally grown oxide thickness, porosity, and crack length. Using a program like that would give a measuring technique that is repeatable, in other words would give the same result
3.3 Evaluation 25
for the same picture at different measurings. But unfortunately it did not work due to problems with the coding of the program. The porosity level was however determined in the MicroGOP by using a function in the program which measures the percentage area of a certain shade, in this case the pores.
3.3.3
Measurements
The porosity, crack length and thickness of the thermally grown oxide were mea-sured. Percentage porosity was as explained above measured in the MicroGOP-program. In figure 3.7 there is an example of a image used for measurement. The image is in fact three, some times four, images stitched together in Adobe Pho-toshop CS2, hence the narrow and tall picture. Crack lengths for the thermally
Figure 3.7. A picture used for evaluation of the porosity. Measurements were made in
three different areas, high, middle and low.
cycled specimens were measured in Kappa Image Base. Two assumptions were made:
• The non-cycled (as-coated) specimen has no damage.
• Cracks are interface and/or interface near cracks (see section 4.1.2).
leading to no measurement of the as-coated specimen and only consideration of the interface/interface near cracks. The length of the cracks was measured as the horizontal ”image” of the real crack, which may follow the undulating interface, see figure 3.8. It is not really the crack length that is sought but the amount of damage, which can be calculated as
D = total crack length
26 Experiment
Figure 3.8. The crack length was measured as the horizontal ”image” of the real crack.
The damage can be defined as zero for the as-coated specimen and one when spalling is started [9]. In this thesis, the damage is defined as zero for the as-coated specimen and one when total spallation has occurred. The measured crack lengths were compiled in Microsoft Office Excel for calculation of damage and to find the relation between number of cycles and amount of damage.
The thermally grown oxide thickness was also measured in Kappa Image Base. Every specimen was represented by six pictures taken in the SEM with the mag-nification x1000. In every picture was, when possible, five measurements made.
3.4
Finite Element Calculations
The program Trinitas, a finite element program developed at Linköping University, was used to calculate the energy release rate, G, and stress intensity factors for mode I and II, KI and KII, for different crack lengths. A preexisting model, made
be M. Jinnestrand for his PhD-thesis, for a thin TBC system was modified to correspond to a thick TBC. The model uses the displacement at the crack surface, see figure 3.9, to calculate a relation between KI and KII. This relation together
with G is used to calculate KI and KII. The top coat/bond coat rough interface is
modeled as a sinusoidal curve, but with only two wavelengths to simplify the model, see figure 3.10. A preexisting crack is placed on top of the wave, propagating along the interface until it reaches the valley of the undulation/wave. The wavelength is 140 µm, giving a maximum crack length of 70 µm. The amplitude of the sine curve is 10 respectively 15 µm giving a total height of 20 respective 30 µm. The calculations were done for 20 different crack lengths, ranging from 4 µm to 66 µm, giving a damage ranging from 6% to 94%. The damage is calculated as the crack length divided by the maximum crack length. For every crack length, three different thermally grown oxide thicknesses (4, 6 and 8 µm) and two different average bond coat surface roughness (Ra) values were used in order to see how
they influence the life of the TBC system. The TBC system is assumed to be stress free at high temperatures. The model represents a TBC system at 1000◦C
3.4 Finite Element Calculations 27
which is cooled down to 0◦C. The cooling causes stresses in the material, due to the thermal expansion mismatch and growth of oxides. Since the cool-down time to low temperature is short, it is assumed that an elastic analysis is sufficient for retrieval of the fracture mechanical data.
Figure 3.9. The displacement of the nodes in the circle is used for calculation of a
relation of KI and KII.
Figure 3.10. The right picture shows the modeled undulation and the left picture shows
Chapter 4
Results and Discussion
The results from the experiments will be presented in this chapter together with a discussion around them.
4.1
Thermal Cycling Fatigue
The thermal cycling of the specimen was carried out successfully. The test should ideally be preformed continuously, all cycles are the same, but to be able to re-move the specimens after the right number of cycles the TCF-machine stopped at a preset number. That lead to a longer cooling part of some cycles with room temperature as minimum temperature. This should however not have any effect on the results.
Six specimens were cycled; 50, 100, 150, 200, 300 cycles and one that was cy-cled until total spallation, approximately 450 cycles. The different specimen will hereafter be referred to as
• K0 or as-coated - the uncycled specimen • K50 - specimen cycled 50 times
• K100 - specimen cycled 100 times • K150 - specimen cycled 150 times • K200 - specimen cycled 200 times • K300 - specimen cycled 300 times
• K450 - specimen cycled until total spallation
The TCF test was done in order to simulate the thermal cycles in a gas turbine, but the test differs from real gas turbines in some ways. For example, the furnace in a TCF test does not contain the same gas composition as the turbine, which will give a lower number of cycles in reality if, for example, corrosion influences the top coat adherence. Another difference is that in the TCF test the substrate is
30 Results and Discussion
heated to the same temperature as the TBC system, while in a gas turbine there is rear-side cooling of the substrate leading to a thermal gradient. The stresses due to different thermal expansions caused by the thermal gradient will in other words not occur in the TCF test. In TBC systems, the thermal gradient is one of the causes of failure, especially in thick TBC systems which have a larger gradient. However, during air cooling in the TCF test a thermal gradient will occur, but this one will be reversed to the one in the gas turbine and it will only occur during a limited amount of time. This gradient will probably not influence the results and therefore will the fractures due to the thermal cycling not derive from a thermal gradient, only from the thermal expansion mismatch between ceramic and metallic layers, and the thermally grown oxide.
4.1.1
Specimen Preparation
There were some problems with the specimen preparation leading to uncertainties in the microstructure analyses. Firstly, the low-viscosity epoxy did not penetrate all the way through the top coat when impregnated. A reason for this can be that the viscosity of the epoxy was not low enough. The result may also been better if a higher pressure had been applied onto the specimen when the epoxy impregnated it. Not having epoxy all the way through the ceramic coat led to problems in distinguishing the real cracks caused by the thermal cycling from the one caused by specimen preparation. There were also problems in reaching the true structure. The grinding and polishing created a lot of pull-outs in the ceramic. The pull-outs led to problems with getting a scratch-free specimen (the metallic part) but fore-most it meant a too high measured porosity level in the ceramic. Different variants of the specimen preparation used, table 3.1, were tested. For example, the time was varied and an additional step with 1 µm diamond paste was tested. Another, completely different method could have been tested to possible get better results. For example using SiC grit papers as done in [11]. This option was however not chosen since the epoxy impregnation did not work anyway. Instead more time was given to the ion etching method. The specimen prepared with the traditional specimen preparation was however used in the present analysis.
The ion etching method was supposed to be used for all the seven specimens but was only used for the as-coated specimen since it was too difficult to get good results. It was used to produce a specimen that could be used as a reference spec-imen for the traditional prepared specspec-imen for how the true structure should look like. A cut was done some millimeters into the specimen. However, the etching machine only managed to cut to a depth of approximately 15 µm, see figure 4.1, which here was not enough to study the structure. Therefore new tests were pre-formed to cut just at the edge of the specimen. With this method the ion etching machine managed to cut all the way through the specimen. The idea was to cut some microns off at a time until the true structure was reached. Every cut took about four hours to do. Unfortunately, the ion gun in the etching machine was short-circuited every now and then. Material that is evaporated during etching is deposited onto other surfaces in the machine, for example on the ion gun, leading
4.1 Thermal Cycling Fatigue 31
Figure 4.1. A cut done in the ceramic layer by ion etching.
to the short-circuit. A conclusion was drawn that the machine was not suitable for cutting the material in question. To be able to make a cut, an ion gun with a more focused beam is needed, in other words a beam with higher intensity. The result obtain can be seen in figure 4.2. The ion etching machine has managed to cut through the whole way, but the cuts have been done on a surface still affected by the previous sectioning.
4.1.2
Evaluation of Failure Mechanism
By studying K300 and K450 in SEM and LOM a conclusion was drawn that the damage due to the thermal cycling had occurred at or near the top coat/bond coat interface. K450 fracture surface showed a mixture between black and white failures, in other words a mixed failure. K300 had larger cracks along the interface and in the interface near area. The overall impression was that the cracks started from the ridge of the undulation and either followed the thermally grown oxide (black failure) or propagated into the ceramic (white failure), see figure 4.3. As mentioned earlier, the stresses induced by the thermal expansion mismatch and the thermally grown oxide are likely the driving forces for failure. This together with the failure mechanism agrees with literature that says that these two causes will lead to damage at or near the interface [5, 7, 8, 12, 13].
4.1.3
Evaluation of Porosity
According to the manufacturer of the ceramic coating, the porosity of the as-coated specimen should be about 20%. The mean value of the measured porosity for K0 was 34%. That means 14 percentage points higher than expected. It can be explained by the unsatisfying specimen preparation. Measurements for all specimens are presented in table 4.1. The measurements were done on three
32 Results and Discussion
Figure 4.2. The as-coated specimen after ion etching. The ceramic layer is at the
bottom and the substrate at the top. The bond coat appears to have two shades; the darker one is cut once more than the brighter one. The cut was done from the ceramic down to the substrate.
Figure 4.3. The specimen has been cycled 300 times. The crack is an example of a
mixed failure. At number 1 the crack is propagating in the ceramic and at 2 in the thermally grown oxide.
4.1 Thermal Cycling Fatigue 33
positions of each picture; at the top, the middle and the bottom of the ceramic coat. There were differences in the measurement of the porosity for the different places, but not large enough to make any conclusions from it. Furthermore, there was no tendency that for example the bottom always had higher porosity than the top. Neither could any conclusions be made whether sintering had occurred or not as a result of the uncertain measurements.
Specimen Porosity Top Middle Bottom
K0 34% 26% 31% 42% K50 27% 23% 26% 31% K100 28% 25% 27% 32% K150 28% 30% 27% 28% K200 35% 39% 30% 35% K300 30% 31% 28% 32%
Table 4.1. The measured porosity. High, middle and bottom indicates where on the
ceramic the porosity is measured.
4.1.4
Crack Evaluation
Due to the large amount of pull-outs it was often difficult to distinguish a crack from a pull-out in the crack evaluation, for example see figure 4.4. One can see that the area marked could have been a crack as well as a pull-out. The measuring was done manually, hence there was no repeatability which is desired. Because of these two reasons, there is a high uncertainty around the results acquired. Three measurements were performed and the results are displayed in figure 4.5 and appendix A. The large distribution of the results shows that an accurate measurent is not possible to acquire. However, a tendency for the failure development can be seen. It appears that the failure development is exponential. In a thin TBC there is first a rapid crack initiation and propagation followed by a plateau where the propagation is slowed down, thereafter is yet again a rapid crack growth leading to failure [12]. This indicates that thin and thick TBC have different damage developments.
4.1.5
Evaluation of Thermally Grown Oxides
The thickness of the thermally grown oxide has been measured for all the specimens except K0. K0 should have a thin layer of oxide due to the air plasma spraying, but no oxide could be detected. In figure 4.6 one can se that there is no contact between bond coat and top coat, as it should have been, with or without oxide. The gap between bond coat and top coat can not be used for measuring the possible thickness of the oxide since parts of the bond and top coat as well may have been pulled out. Instead, a thickness of 1 µm has been assumed after [12]. The measurements on the remaining specimens were preformed on parts of the specimen where the ceramic part was intact. If the ceramic layer is missing, there is a big possibility that parts of the oxide is missing too. The results are presented
34 Results and Discussion
Figure 4.4. A crack or a pull-out in a specimen cycled 300 times.
Figure 4.5. The results from all three crack measurment. The measurments are
4.1 Thermal Cycling Fatigue 35
in figure 4.7 and figure 4.8. The first part of the thermally grown oxide-curve should represent the time when the oxide mainly consist of alumina. Alumina has a Pilling-Bedworth ratio of 1.28 [5] which gives a parabolic oxide growth - time (here cycles) curve. As a result of aluminum depletion a mixed oxide (aluminium, chromium, cobalt and nickel) called spinels may form. This oxide have higher growth rate than alumina, leading to a curve that is logarithmic. This is however not how the obtained curves look like and the most likely reason for this is yet again the specimen preparation. Chemical analyses of K300 and K450, see appendix A.4 shows that the bond coats internal oxides consist of both aluminum and spinels and a line analysis of the bond coat confirms that the bond coat is depleted of aluminum in K300 and K450. However, the thermally grown oxide layer of K300 and K450 only consists of alumina. This together with the fact that the internal oxides are thicker than the thermally grown oxide, also indicates that parts of the thermally grown oxide has been pulled out during preparation meaning and that the measured thicknesses are not valid. The missing oxide results in a large marginal of errors especially for K300 and K450 since spinels are more brittle and will therefore more easily be pulled out. The true thickness of the thermally grown oxide when spallation occurred, could not be determined, but a rough estimation of the thickness, based on the measurements and the fact that the spinels are missing, is 6 µm. This estimated thickness is, compared to thin TBC [12], the same for the corresponding amount of cycles. When failing, thin TBC has a thermally grown oxide thickness of approximately 7 µm.
Figure 4.6. There is no oxide between the bond coat and top coat.
4.1.6
Diffusion
Chemical analyses have been done in form of mapping and line scans (see ap-pendix A.4 for some of the analyses). These analyses gave suggestions to how the
36 Results and Discussion
Figure 4.7. The results from the measurements of the thermally grown oxide. The line
is a trendline based on the measurements diplayed as the dots.
Figure 4.8. The results from the mesurements of the thermally grown oxide. The two
