- Email: [email protected]
SiC composites subjected to constant load and thermal cycling in oxidizing atmosphere
Scripta Materialia 54 (2006) 163–168 www.actamat-journals.com
Damage mechanisms of C/SiC composites subjected to constant load and thermal cycling in oxidizing atmosphere Hui Mei *, Laifei Cheng, Litong Zhang National Key Laboratory of Thermostructure Composite Materials, Northwestern Polytechnical University, 547 Mailbox, XiÕan Shaanxi 710072, PeopleÕs Republic of China Received 4 April 2005; received in revised form 9 September 2005; accepted 26 September 2005 Available online 20 October 2005
Abstract Properties of a carbon fiber reinforced silicon carbide matrix composite were investigated in controlled environments including constant load, thermal cycling and wet oxygen atmosphere. Damage was assessed by residual mechanical properties and scanning electron microscopy characterization. Thermal strain was shown to change with cyclic temperatures over the same period (120 s). Strain varies approximately from the initial linear elastic strain of 0.63% to the final nonreversible damage strain of 1.6% during the short time of the test. The experimental strain difference between two selected temperatures is about 0.16% and the theoretical calculation value is 0.1566%. After 50 thermal cycles, the YoungÕs modulus of the composites is reduced by a factor of 0.5 while the residual strength still retains 82% of the initial strength. It is observed that matrix cracks transversely and wave-shaped cracks are arranged on the coating surface at relatively regular spacing. A typical superficial oxidation can be found along the opening and propagating cracks beneath the coating. 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Fiber; Ceramic matrix composites; Thermal cycling; Creep; Residual properties
1. Introduction Carbon fiber reinforced SiC-matrix composites (C/SiC) fabricated by the chemical vapor infiltration process (CVI) have been proposed as advanced materials suitable for aerospace and gas turbine engine parts [1,2]. In particular, in recent years many efforts have been devoted to the high-temperature applications of C/SiC composites. These composites show some attractive properties and advantages over traditional ceramics: higher tensile and flexural strength, enhanced fracture toughness and impact resistance, lower density and no cooling requirement. In particular, the mechanical properties of C/SiC composites can be retained at high-temperatures and under severe service environments. In many of the instances under consider*
Corresponding author. Tel.: +86 29 88494616; fax: +86 29 88494620. E-mail addresses: [email protected], phdhuimei@yahoo. com (H. Mei).
ation, the composites will be also subjected to both thermal cycling and some rigid constraint conditions in an oxidizing atmosphere during its service. Consequently, thermal cycling damage to the composites under such conditions must be well understood before actual use in these environments. The effects of temperature cycling on the structural integrity of polyester matrix composites have been investigated using a large scale model composite [3] and a computational model of delamination of a two-layer composite laminate subjected to the cyclic loads, both mechanical and thermal, was obtained [4]. On the other hand, experimental thermal shock studies have also been conducted on unidirectional, two-dimensional and three-dimensional (3D) woven-fiber composites [5–7] and newly-developed ascending thermal shock test equipment has also been applied to study thermal shock and thermal fatigue of ceramic materials [8]. However, the mechanical response and damage features of C/SiC composites subjected to thermal
1359-6462/$ - see front matter 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.scriptamat.2005.09.044
164
H. Mei et al. / Scripta Materialia 54 (2006) 163–168
cycling under a constant load in oxidizing atmosphere have not been reported in a detailed manner, despite recent advances in the architecture design and processing of these materials. When evaluating C/SiC composites for potential use in structural applications where periodically changing temperatures occur, the basic characterization of the materials obtained from mechanical and environmental testing is very important in understanding the fundamental properties of the materials. In this paper, thermal cycling testing results of 3D braided C/SiC composites under a constant load of 60 MPa in a wet oxygen atmosphere are presented. Corresponding thermal stress or thermal strain during testing will be measured, calculated and analyzed by theoretical and experimental methods. Effects of thermal cycling on mechanical properties of composites will be discussed and the morphologies of fracture sections and coating surfaces will be observed. 2. Experimental 2.1. Preparation of C/SiC composite T-300TM carbon fiber from Toray (Japan) was employed. The fiber preform was prepared using a 3D braid method. The volume fraction of fibers was about 40% and the braiding angle was about 20. Low pressure CVI was employed to deposit a pyrolytic carbon layer and the silicon carbide matrix. A thin pyrolytic carbon layer was deposited on the surface of the carbon fiber as the interfacial layer with C3H8 at 800 C. Methyltrichlorosilane (MTS, CH3 SiCl3) was used for the deposition of the SiC-matrix. MTS vapor was carried by bubbling hydrogen. Typical conditions for deposition were 1000 C, a hydrogen:MTS ratio of 10:1, and a pressure of 5 kPa. Argon was employed as the diluent gas to slow down the chemical reaction rate of deposition. Finally, the test specimens were machined from the fabricated composites and further coated with SiC by isothermal CVI under the same conditions. The morphology and dimensions of the as-received specimens are shown in
Figs. 1 and 2. The properties of the composite are listed in Table 1. 2.2. Thermal cycling test Thermal cycling experiments under load constraints were conducted with a newly-developed integrated system including an induction heating furnace (with a controlled atmosphere chamber providing various kinds and concentrations of oxidizing gas) monitored by a programmable microprocessor and a servo-hydraulic machine (Model Instron 8801, Instron Ltd., England). Many experimental conditions/parameters must be taken into consideration, especially regarding (1) the load alignment, (2) the configuration of the specimen, (3) the heater, (4) the cooling water, (5) the measurement for temperature, (6) the induction coil for cyclic temperature, (7) the grip holder and (8) the pressure and flow of the controlled atmosphere (as shown in Fig. 3). The temperature was measured by an infrared pyrometer through a small window in the wall of the furnace and the wall was internally cut out to enable the circulating cold water to reach all over the surfaces. Thermal cycling was carried out between two selected temperatures and the period was 120 s: holding for 30 s at 900 C, heating to 1200 C in 60 s and holding for 30 s, then cooling back to 900 C immediately (temperature difference DT 300 C). Only the middle parts of the specimens (40 mm long, 3 mm wide and 3 mm thick) were kept in the hot zone and oxidizing atmosphere. In testing, a constant load of 60 MPa was applied to both the longitu-
Fig. 2. Drawing of as-prepared C/SiC specimen (all dimensions in mm).
Fig. 1. (a) Substrate of the 3D-C/SiC composite and (b) its top surface.
H. Mei et al. / Scripta Materialia 54 (2006) 163–168
165
Table 1 Properties of the as-received 3D-C/SiC composites Property
Density (·103 kg/m3)
Modulus (GPa)
Strength (MPa)
PoissonÕs ratio
Porosity (%)
Value
2.0
50
160
0.52
11
CTE (·106/C) 900 C
1000 C
1100 C
1200 C
4.739
3.428
3.840
5.702
Fig. 3. Schematic drawing of atmosphere chamber with a detailed view of the grip holders, the furnace and the specimen (eight major critical points are indicated).
dinal ends of the specimen and the oxidizing atmosphere was wet oxygen including oxygen (7.90%, partial pressure: P O2 8000 Pa), water vapor (14.85%, partial pressure: P H2 O 15 kPa) and argon (77.25%). The flux of gases was accurately controlled by a mass flow controller (5850 i series from BROOKS, Japan) and its precision could reach 0.1 SCCM. Strains were measured directly from the gauge length of specimen by a contact extensometer (Instron model number: A1452-1001B) with a gauge length of 10 mm.
Fig. 4. (a) Strains vs. thermal cycle number N-curve for 3D-C/SiC composite subjected to 50 thermal cycles under a constant load of 60 MPa. (b) Magnified view of the initial cycles by using logarithmic horizontal scale.
2.3. Measurements and observations Monotonic tensile tests of the specimens after thermal cycling experiments were done on the servo-hydraulic machine (Instron 8801). The coating surfaces and fracture sections of the specimens were observed with a scanning electron microscope (SEM, JEOL JSM-6460 and HITACHI S-4700). 3. Results and discussion 3.1. Strain response curves of C/SiC composites subjected to thermal cycling and constant load Strain vs. cycle number curves of the C/SiC composite subjected to a constant load of 60 MPa and thermal cycling (N = 50) in the wet oxygen are shown in Fig. 4(a). Fig. 4(b) is the magnified view of the initial cycles using a logarithmic horizontal scale. It is apparent that the measured
strains should be the coupled results of cyclic strain due to thermal cycling and creep strain due to the constant loading (60 MPa). The contribution of the creep strain of the composite to the total strain is rather large at the initial stage. Subsequently, thermal cycling becomes a dominant factor of increasing strain. The most important information demonstrated in Fig. 4(b) is that the strain of the specimen retains 0.63% under the constant load of 60 MPa at room temperature, and increases, gradually reaching a peak as the temperature ascends to the selected upper limit of 1200 C; the strain then decreases with cooling back to 900 C. As thermal cycling proceeds, saw-toothed strain repeats periodically during testing and the complete period of each strain cycle is equal to 120 s. All the differences between the peaks and the valleys of the strain waves are approximately identical in magnitude and the average value is about 0.16%. Periodical thermal expansion and
166
H. Mei et al. / Scripta Materialia 54 (2006) 163–168
cooling shrinkage caused by cyclic temperatures should be responsible for the results. Furthermore, cyclic stresses resulted from thermal cycling in the specimen are similar to external fatigue loads. This cyclic unloading and reloading increases the debonding length by increasing the interface sliding stress [9–12]. The constant load applied to the specimen accelerates the interface debonding and sliding. The larger the debonding length, the larger the crack opening displacement and the larger the contribution of nonreversible deformation to the total strain of the specimen, because the same gauge lengths of the specimens are used to calculate the strain of the composite [13]. In addition, a steady increase in strain with increasing thermal cycles is also one of the indicators of the accumulation of thermal cycling damage during constant loading. The range of thermal strain varies approximately from the initial linear elastic strain of 0.63% to the final nonreversible damage strain of 1.6% in only 100 minutes (50 thermal cycles). The elapsed strain due to the accumulated damage is rather larger than single creep strain. Consequently, thermal cycling could enhance the strain rate of composites by forming nonreversible damage. The strain measured in testing should include the crack opening displacement, interfacial debonding and sliding, and the total thermal expansion of the whole composite body.
In the present paper, according to Table 1, the coefficient of thermal expansion a is about 4.739 · 106/C at 900 C and about 5.702 · 106/C at 1200 C. Thus, their mean value is 5.22 · 106/C and the temperature gradient DT 300 C. From Eq. (2), the thermal strain difference between 900 C and 1200 C should approximate to 0.1566%, which is almost identical to the experimental metrical thermal strain difference (about 0.16% in Fig. 4(b)). In addition, from Eq. (3), we can also obtain: rthermal 78.3 MPa (the original YoungÕs modulus E is about 50 GPa according to the Table 1) when the temperature is held at the lower 900 C. The thermal stress, however, would decrease gradually with decreasing modulus of the specimen in testing. The modulus of the specimen was about 26.35 GPa after 50 cycles (see Section 3.3). Therefore, thermal stress approximates to 41.26 MPa. Resultant stress applied to the specimen can be estimated as rResultant ¼ rthermal þ rconstant
ð4Þ
Thus, maximum resultant stress applied to the specimen varies from the initial 138.3 MPa to a final 101.26 MPa. The stresses produced more and more cracks on the surfaces of the brittle ceramic coating, through which fibers were oxidized or damaged by the wet oxygen. 3.3. Monotonic tensile behavior after 50 thermal cycles
3.2. Theoretical analysis of thermal stress and thermal strain in testing As we know, a sudden change in the surrounding temperature generates thermal expansion or cooling shrinkage of materials. Therefore, when the expansion or shrinkage of the ceramic body is constrained, thermal stress is produced and its damage to the composite is a very important issue which requires discussion. What magnitude was this thermal stress theoretically? It was supposed that the increment of specimen length was Dl due to thermal expansion, and that the thermal stress produced from constrained thermal expansion was approximately equal to the stress which was required to compress the specimen to the original length (l), by the same amount of deformation Dl in the opposite direction. The thermal stress is simply expressed as Dl rthermal ¼ E ð1Þ ¼ Eethermal l
After C/SiC composites were subjected to 50 thermal cycles (100 minutes) under a constant load of 60 MPa in the wet oxygen atmosphere, the monotonic tensile test with a loading rate of 0.001 mm/s at room temperature was conducted on the Instron machine. Tensile curves are shown in Fig. 5. A typical characteristic of the curves is the occurrence of inflexion at a stress of 60 MPa, which was the same stress applied to the specimen in the thermal cycling test mentioned above. It indicated that the thermal cycling
where rthermal and ethermal refer to the thermal stress and thermal strain, respectively. E is the YoungÕs modulus. Similarly, the expansion or shrinkage of materials can be caused by sudden changes in the ambient temperature. Consequently, thermal stress is also obtained as follows: ethermal ¼ aDT rthermal ¼ EaDT
ð2Þ ð3Þ
where a is the coefficient of longitudinal thermal expansion of specimen and DT is the temperature difference.
Fig. 5. Monotonic tensile curve with a loading rate of 0.001 mm/s at room temperature after thermal cycling test in wet oxygen atmosphere.
H. Mei et al. / Scripta Materialia 54 (2006) 163–168
under the constant load damaged the interface between the fibers and the matrix, resulted in matrix cracking and broken fibers, and all these damages were recorded in the tested composites. The moduli calculated from the two linear portions of the curve were 48.86 GPa and 26.35 GPa, respectively, and the value is reduced by a factor of 0.5. The initial portion of the curve indicates a linear elastic behavior up to the proportional limit of 60 MPa and the modulus is slightly lower than the virgin value of 50 GPa given in Table 1. The stresses (less than 60 MPa) are not high enough to produce extensive matrix cracking leading to reduction of modulus. However, after the stress of 60 MPa, the slope of the curve decreases rapidly due to the accumulated damage in the composite. The decrease in modulus could be ascribed to matrix cracking, fiber fracture and interfacial debonding during the thermal cycling tests. Finally, the ultimate tensile strength is 131.4 MPa and the value is still 82% of the initial strength. The failure strain approximates to 0.58% and is dramatically lower than the maximum strain during the thermal cycling tests (about 1.6%). The constant loading and thermal cycling in testing led to increasing strain by forming the nonrevers-
167
ible damage. Therefore, elongation of the surviving specimens in the next monotonic tension seems to be limited. Thermal cycling damage to the modulus is more severe than to the strength of the C/SiC composites. This should be partially ascribed to the brittleness of the SiC ceramic matrix and its sensitivities to crack propagation. 3.4. Microstructural observations After the monotonic tensile test, coating surfaces, matrix and fracture sections were observed on a SEM microscope. As shown in Fig. 6, cyclic unloading and reloading under cyclic temperatures can result in matrix cracking (Fig. 6(a)) transversely, and wave-shaped coating cracks at regular spacing (about 278 lm in Fig. 6(b)). Fig. 7(a) shows that cracks appear in the SiC coating normally before testing due to the mismatch of substrate and coating. A typical superficial oxidation can also be found beneath the coatings along the cracks in Fig. 7(b). Oxidation regions were found along the opening cracks perpendicular to the substrate surface because these cracks were enlarged during the thermal cycling test with a constant load.
Fig. 6. Typical SEM micrograph showing (a) transverse cracks in the matrix and (b) wave-shaped coating cracks at relatively regular spacing after 50 thermal cycles. Matrix cracks marked by arrows in (a) are partially closed once unloaded.
Fig. 7. Typical cross section micrograph of the C/SiC composites (a) before and (b) after 50 thermal cycles under a constant load of 60 MPa in wet oxygen atmosphere.
168
H. Mei et al. / Scripta Materialia 54 (2006) 163–168
Fig. 8. Schematic diagram showing the effect of longitudinal strain on the braiding angle between neighbouring fiber bundles under cyclic thermal stress.
The constant load of 60 MPa and cyclic thermal stress should be responsible for these results. A schematic diagram (Fig. 8) could help us to understand why the waveshaped cracks formed on the ceramic coatings. When a constant load and cyclic thermal stress were applied to the 3D braided composite substrate (as shown in Fig. 1) in the longitudinal direction, all neighbouring two fiber bundles would slip and/or rotate around their contacts and the braiding angle was reduced by transverse compression. SiC ceramic coating, covering the 3D braided C/SiC composite substrate, was opened up due to the constant load and then the opening cracks were corrugated with transverse shrinkage strain. The cyclic thermal stress made the cracks uniformly distributed in the coating and matrix, and the crack spacing was almost equal. No matter how high or low the temperature was during testing, cracks had never closed up completely due to the constant loading. The cracks perpendicular to the substrate surface were most open when temperature was held at 900 C (the resultant stress reached a maximum). Diffused oxygen along the open coating cracks reached the surfaces of the carbon fibers in two directions marked by arrows in Fig. 7(b) and was depleted rapidly by the external fibers. Subsequently the loading was transferred to the internal fibers. Thus, when the new or propagated cracks of the matrix were produced inwards, internal oxidation would take place, and the stress oxidation of carbon fibers in particular was ready to occur upon loading. 4. Conclusions 1. Under a constant load and thermal cycling, a gradually increasing creep strain coupled with cyclic strain could be measured in good order. Thermal strain difference approximated to 0.16 % when the composite was subjected to a constant load of 60 MPa and the temperature difference was about 300 C.
2. Theoretical calculation was in agreement with the observed experimental phenomena. The value of the calculated thermal strain difference was about 0.1566% (experimental value was 0.16%). Maximum resultant stress applied to the specimen obtained by calculation varied from the initial 138.3 MPa to the final 101.26 MPa. 3. Thermal cycling can enhance the strain rate of the composites by forming nonreversible damage and the damage recorded in the tested composites could be replayed upon reloading. The damage strain in testing included crack opening displacement, interfacial debonding and sliding. Elongation of the surviving specimen in the next monotonic tension seems to be limited due to the damage strain. 4. The thermal cycling damage to the modulus was more severe than to the strength for C/SiC composites. After 50 thermal cycles, the modulus of the composites was reduced by a factor of 0.5 while the residual strength still retained 82% of the original strength. 5. Wave-shaped cracks were arranged on the coatings at relatively regular spacing (about 278 lm) and the matrix cracked transversely. A typical superficial oxidation was found along the opening cracks beneath the coating. This should be ascribed to the cyclic thermal stress, constant load, wet oxygen atmosphere and the special braided structure of 3D-C/SiC composites.
Acknowledgements The authors acknowledge the financial support of the Natural Science Foundation of China (Contract No. 90405015) and the National Young Elitists Foundation (Contract No. 50425208). References [1] Lamouroux F, Bourrat X, Sevely J, Naslain R. Carbon 1993;31:1273. [2] Cavalier JC, Lacombe A, Rouges JM. In: Bunsell AR, Lamicq P, Massiah A, editors. Developments in the science and technology of composite materials. London: Elsevier; 1989. p. 99 (in French). [3] Biernacki K, Szyszkowski W, Yannacopoulos S. Compos Part A 1999;30A:1027. [4] Figiel Ł, Kamin´ski M. Comp Struct 2003;81:1865. [5] Webb JE, Singh RN. J Am Ceram Soc 1996;79:2857. [6] Singh RN, Wang H. Compos Eng 1995;5:1287. [7] Yin XW, Cheng LF. Carbon 2002;40:905. [8] Panda PK, Kannan TS, Dubois J, Olagnon C, Fantozzi G. Sci Technol Adv Mater 2002;3:327. [9] Reynaud P, Rouby D, Fantozzi G. Scripta Metall Mater 1994;31:1061. [10] Fantozzi G, Reynaud P, Rouby D. Silic Indus 2001;66:109. [11] Cherouali H, Fantozzi G, Reynaud P, Rouby D. Mater Sci Eng 1998;250A:169. [12] Miyashita Y, Kanda K, Zhu S, Mutoh Y, Mizuno M, McEvily AJ. Int J Fatigue 2002;24:241. [13] Zhu S, Mizuno M, Kagawa Y, Cao JW, Nagano Y, Kaya H. Mater Sci Eng 1997;225A:69.
Damage mechanisms of C/SiC composites subjected to constant load and thermal cycling in oxidizing atmosphere Hui Mei *, Laifei Cheng, Litong Zhang National Key Laboratory of Thermostructure Composite Materials, Northwestern Polytechnical University, 547 Mailbox, XiÕan Shaanxi 710072, PeopleÕs Republic of China Received 4 April 2005; received in revised form 9 September 2005; accepted 26 September 2005 Available online 20 October 2005
Abstract Properties of a carbon fiber reinforced silicon carbide matrix composite were investigated in controlled environments including constant load, thermal cycling and wet oxygen atmosphere. Damage was assessed by residual mechanical properties and scanning electron microscopy characterization. Thermal strain was shown to change with cyclic temperatures over the same period (120 s). Strain varies approximately from the initial linear elastic strain of 0.63% to the final nonreversible damage strain of 1.6% during the short time of the test. The experimental strain difference between two selected temperatures is about 0.16% and the theoretical calculation value is 0.1566%. After 50 thermal cycles, the YoungÕs modulus of the composites is reduced by a factor of 0.5 while the residual strength still retains 82% of the initial strength. It is observed that matrix cracks transversely and wave-shaped cracks are arranged on the coating surface at relatively regular spacing. A typical superficial oxidation can be found along the opening and propagating cracks beneath the coating. 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Fiber; Ceramic matrix composites; Thermal cycling; Creep; Residual properties
1. Introduction Carbon fiber reinforced SiC-matrix composites (C/SiC) fabricated by the chemical vapor infiltration process (CVI) have been proposed as advanced materials suitable for aerospace and gas turbine engine parts [1,2]. In particular, in recent years many efforts have been devoted to the high-temperature applications of C/SiC composites. These composites show some attractive properties and advantages over traditional ceramics: higher tensile and flexural strength, enhanced fracture toughness and impact resistance, lower density and no cooling requirement. In particular, the mechanical properties of C/SiC composites can be retained at high-temperatures and under severe service environments. In many of the instances under consider*
Corresponding author. Tel.: +86 29 88494616; fax: +86 29 88494620. E-mail addresses: [email protected], phdhuimei@yahoo. com (H. Mei).
ation, the composites will be also subjected to both thermal cycling and some rigid constraint conditions in an oxidizing atmosphere during its service. Consequently, thermal cycling damage to the composites under such conditions must be well understood before actual use in these environments. The effects of temperature cycling on the structural integrity of polyester matrix composites have been investigated using a large scale model composite [3] and a computational model of delamination of a two-layer composite laminate subjected to the cyclic loads, both mechanical and thermal, was obtained [4]. On the other hand, experimental thermal shock studies have also been conducted on unidirectional, two-dimensional and three-dimensional (3D) woven-fiber composites [5–7] and newly-developed ascending thermal shock test equipment has also been applied to study thermal shock and thermal fatigue of ceramic materials [8]. However, the mechanical response and damage features of C/SiC composites subjected to thermal
1359-6462/$ - see front matter 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.scriptamat.2005.09.044
164
H. Mei et al. / Scripta Materialia 54 (2006) 163–168
cycling under a constant load in oxidizing atmosphere have not been reported in a detailed manner, despite recent advances in the architecture design and processing of these materials. When evaluating C/SiC composites for potential use in structural applications where periodically changing temperatures occur, the basic characterization of the materials obtained from mechanical and environmental testing is very important in understanding the fundamental properties of the materials. In this paper, thermal cycling testing results of 3D braided C/SiC composites under a constant load of 60 MPa in a wet oxygen atmosphere are presented. Corresponding thermal stress or thermal strain during testing will be measured, calculated and analyzed by theoretical and experimental methods. Effects of thermal cycling on mechanical properties of composites will be discussed and the morphologies of fracture sections and coating surfaces will be observed. 2. Experimental 2.1. Preparation of C/SiC composite T-300TM carbon fiber from Toray (Japan) was employed. The fiber preform was prepared using a 3D braid method. The volume fraction of fibers was about 40% and the braiding angle was about 20. Low pressure CVI was employed to deposit a pyrolytic carbon layer and the silicon carbide matrix. A thin pyrolytic carbon layer was deposited on the surface of the carbon fiber as the interfacial layer with C3H8 at 800 C. Methyltrichlorosilane (MTS, CH3 SiCl3) was used for the deposition of the SiC-matrix. MTS vapor was carried by bubbling hydrogen. Typical conditions for deposition were 1000 C, a hydrogen:MTS ratio of 10:1, and a pressure of 5 kPa. Argon was employed as the diluent gas to slow down the chemical reaction rate of deposition. Finally, the test specimens were machined from the fabricated composites and further coated with SiC by isothermal CVI under the same conditions. The morphology and dimensions of the as-received specimens are shown in
Figs. 1 and 2. The properties of the composite are listed in Table 1. 2.2. Thermal cycling test Thermal cycling experiments under load constraints were conducted with a newly-developed integrated system including an induction heating furnace (with a controlled atmosphere chamber providing various kinds and concentrations of oxidizing gas) monitored by a programmable microprocessor and a servo-hydraulic machine (Model Instron 8801, Instron Ltd., England). Many experimental conditions/parameters must be taken into consideration, especially regarding (1) the load alignment, (2) the configuration of the specimen, (3) the heater, (4) the cooling water, (5) the measurement for temperature, (6) the induction coil for cyclic temperature, (7) the grip holder and (8) the pressure and flow of the controlled atmosphere (as shown in Fig. 3). The temperature was measured by an infrared pyrometer through a small window in the wall of the furnace and the wall was internally cut out to enable the circulating cold water to reach all over the surfaces. Thermal cycling was carried out between two selected temperatures and the period was 120 s: holding for 30 s at 900 C, heating to 1200 C in 60 s and holding for 30 s, then cooling back to 900 C immediately (temperature difference DT 300 C). Only the middle parts of the specimens (40 mm long, 3 mm wide and 3 mm thick) were kept in the hot zone and oxidizing atmosphere. In testing, a constant load of 60 MPa was applied to both the longitu-
Fig. 2. Drawing of as-prepared C/SiC specimen (all dimensions in mm).
Fig. 1. (a) Substrate of the 3D-C/SiC composite and (b) its top surface.
H. Mei et al. / Scripta Materialia 54 (2006) 163–168
165
Table 1 Properties of the as-received 3D-C/SiC composites Property
Density (·103 kg/m3)
Modulus (GPa)
Strength (MPa)
PoissonÕs ratio
Porosity (%)
Value
2.0
50
160
0.52
11
CTE (·106/C) 900 C
1000 C
1100 C
1200 C
4.739
3.428
3.840
5.702
Fig. 3. Schematic drawing of atmosphere chamber with a detailed view of the grip holders, the furnace and the specimen (eight major critical points are indicated).
dinal ends of the specimen and the oxidizing atmosphere was wet oxygen including oxygen (7.90%, partial pressure: P O2 8000 Pa), water vapor (14.85%, partial pressure: P H2 O 15 kPa) and argon (77.25%). The flux of gases was accurately controlled by a mass flow controller (5850 i series from BROOKS, Japan) and its precision could reach 0.1 SCCM. Strains were measured directly from the gauge length of specimen by a contact extensometer (Instron model number: A1452-1001B) with a gauge length of 10 mm.
Fig. 4. (a) Strains vs. thermal cycle number N-curve for 3D-C/SiC composite subjected to 50 thermal cycles under a constant load of 60 MPa. (b) Magnified view of the initial cycles by using logarithmic horizontal scale.
2.3. Measurements and observations Monotonic tensile tests of the specimens after thermal cycling experiments were done on the servo-hydraulic machine (Instron 8801). The coating surfaces and fracture sections of the specimens were observed with a scanning electron microscope (SEM, JEOL JSM-6460 and HITACHI S-4700). 3. Results and discussion 3.1. Strain response curves of C/SiC composites subjected to thermal cycling and constant load Strain vs. cycle number curves of the C/SiC composite subjected to a constant load of 60 MPa and thermal cycling (N = 50) in the wet oxygen are shown in Fig. 4(a). Fig. 4(b) is the magnified view of the initial cycles using a logarithmic horizontal scale. It is apparent that the measured
strains should be the coupled results of cyclic strain due to thermal cycling and creep strain due to the constant loading (60 MPa). The contribution of the creep strain of the composite to the total strain is rather large at the initial stage. Subsequently, thermal cycling becomes a dominant factor of increasing strain. The most important information demonstrated in Fig. 4(b) is that the strain of the specimen retains 0.63% under the constant load of 60 MPa at room temperature, and increases, gradually reaching a peak as the temperature ascends to the selected upper limit of 1200 C; the strain then decreases with cooling back to 900 C. As thermal cycling proceeds, saw-toothed strain repeats periodically during testing and the complete period of each strain cycle is equal to 120 s. All the differences between the peaks and the valleys of the strain waves are approximately identical in magnitude and the average value is about 0.16%. Periodical thermal expansion and
166
H. Mei et al. / Scripta Materialia 54 (2006) 163–168
cooling shrinkage caused by cyclic temperatures should be responsible for the results. Furthermore, cyclic stresses resulted from thermal cycling in the specimen are similar to external fatigue loads. This cyclic unloading and reloading increases the debonding length by increasing the interface sliding stress [9–12]. The constant load applied to the specimen accelerates the interface debonding and sliding. The larger the debonding length, the larger the crack opening displacement and the larger the contribution of nonreversible deformation to the total strain of the specimen, because the same gauge lengths of the specimens are used to calculate the strain of the composite [13]. In addition, a steady increase in strain with increasing thermal cycles is also one of the indicators of the accumulation of thermal cycling damage during constant loading. The range of thermal strain varies approximately from the initial linear elastic strain of 0.63% to the final nonreversible damage strain of 1.6% in only 100 minutes (50 thermal cycles). The elapsed strain due to the accumulated damage is rather larger than single creep strain. Consequently, thermal cycling could enhance the strain rate of composites by forming nonreversible damage. The strain measured in testing should include the crack opening displacement, interfacial debonding and sliding, and the total thermal expansion of the whole composite body.
In the present paper, according to Table 1, the coefficient of thermal expansion a is about 4.739 · 106/C at 900 C and about 5.702 · 106/C at 1200 C. Thus, their mean value is 5.22 · 106/C and the temperature gradient DT 300 C. From Eq. (2), the thermal strain difference between 900 C and 1200 C should approximate to 0.1566%, which is almost identical to the experimental metrical thermal strain difference (about 0.16% in Fig. 4(b)). In addition, from Eq. (3), we can also obtain: rthermal 78.3 MPa (the original YoungÕs modulus E is about 50 GPa according to the Table 1) when the temperature is held at the lower 900 C. The thermal stress, however, would decrease gradually with decreasing modulus of the specimen in testing. The modulus of the specimen was about 26.35 GPa after 50 cycles (see Section 3.3). Therefore, thermal stress approximates to 41.26 MPa. Resultant stress applied to the specimen can be estimated as rResultant ¼ rthermal þ rconstant
ð4Þ
Thus, maximum resultant stress applied to the specimen varies from the initial 138.3 MPa to a final 101.26 MPa. The stresses produced more and more cracks on the surfaces of the brittle ceramic coating, through which fibers were oxidized or damaged by the wet oxygen. 3.3. Monotonic tensile behavior after 50 thermal cycles
3.2. Theoretical analysis of thermal stress and thermal strain in testing As we know, a sudden change in the surrounding temperature generates thermal expansion or cooling shrinkage of materials. Therefore, when the expansion or shrinkage of the ceramic body is constrained, thermal stress is produced and its damage to the composite is a very important issue which requires discussion. What magnitude was this thermal stress theoretically? It was supposed that the increment of specimen length was Dl due to thermal expansion, and that the thermal stress produced from constrained thermal expansion was approximately equal to the stress which was required to compress the specimen to the original length (l), by the same amount of deformation Dl in the opposite direction. The thermal stress is simply expressed as Dl rthermal ¼ E ð1Þ ¼ Eethermal l
After C/SiC composites were subjected to 50 thermal cycles (100 minutes) under a constant load of 60 MPa in the wet oxygen atmosphere, the monotonic tensile test with a loading rate of 0.001 mm/s at room temperature was conducted on the Instron machine. Tensile curves are shown in Fig. 5. A typical characteristic of the curves is the occurrence of inflexion at a stress of 60 MPa, which was the same stress applied to the specimen in the thermal cycling test mentioned above. It indicated that the thermal cycling
where rthermal and ethermal refer to the thermal stress and thermal strain, respectively. E is the YoungÕs modulus. Similarly, the expansion or shrinkage of materials can be caused by sudden changes in the ambient temperature. Consequently, thermal stress is also obtained as follows: ethermal ¼ aDT rthermal ¼ EaDT
ð2Þ ð3Þ
where a is the coefficient of longitudinal thermal expansion of specimen and DT is the temperature difference.
Fig. 5. Monotonic tensile curve with a loading rate of 0.001 mm/s at room temperature after thermal cycling test in wet oxygen atmosphere.
H. Mei et al. / Scripta Materialia 54 (2006) 163–168
under the constant load damaged the interface between the fibers and the matrix, resulted in matrix cracking and broken fibers, and all these damages were recorded in the tested composites. The moduli calculated from the two linear portions of the curve were 48.86 GPa and 26.35 GPa, respectively, and the value is reduced by a factor of 0.5. The initial portion of the curve indicates a linear elastic behavior up to the proportional limit of 60 MPa and the modulus is slightly lower than the virgin value of 50 GPa given in Table 1. The stresses (less than 60 MPa) are not high enough to produce extensive matrix cracking leading to reduction of modulus. However, after the stress of 60 MPa, the slope of the curve decreases rapidly due to the accumulated damage in the composite. The decrease in modulus could be ascribed to matrix cracking, fiber fracture and interfacial debonding during the thermal cycling tests. Finally, the ultimate tensile strength is 131.4 MPa and the value is still 82% of the initial strength. The failure strain approximates to 0.58% and is dramatically lower than the maximum strain during the thermal cycling tests (about 1.6%). The constant loading and thermal cycling in testing led to increasing strain by forming the nonrevers-
167
ible damage. Therefore, elongation of the surviving specimens in the next monotonic tension seems to be limited. Thermal cycling damage to the modulus is more severe than to the strength of the C/SiC composites. This should be partially ascribed to the brittleness of the SiC ceramic matrix and its sensitivities to crack propagation. 3.4. Microstructural observations After the monotonic tensile test, coating surfaces, matrix and fracture sections were observed on a SEM microscope. As shown in Fig. 6, cyclic unloading and reloading under cyclic temperatures can result in matrix cracking (Fig. 6(a)) transversely, and wave-shaped coating cracks at regular spacing (about 278 lm in Fig. 6(b)). Fig. 7(a) shows that cracks appear in the SiC coating normally before testing due to the mismatch of substrate and coating. A typical superficial oxidation can also be found beneath the coatings along the cracks in Fig. 7(b). Oxidation regions were found along the opening cracks perpendicular to the substrate surface because these cracks were enlarged during the thermal cycling test with a constant load.
Fig. 6. Typical SEM micrograph showing (a) transverse cracks in the matrix and (b) wave-shaped coating cracks at relatively regular spacing after 50 thermal cycles. Matrix cracks marked by arrows in (a) are partially closed once unloaded.
Fig. 7. Typical cross section micrograph of the C/SiC composites (a) before and (b) after 50 thermal cycles under a constant load of 60 MPa in wet oxygen atmosphere.
168
H. Mei et al. / Scripta Materialia 54 (2006) 163–168
Fig. 8. Schematic diagram showing the effect of longitudinal strain on the braiding angle between neighbouring fiber bundles under cyclic thermal stress.
The constant load of 60 MPa and cyclic thermal stress should be responsible for these results. A schematic diagram (Fig. 8) could help us to understand why the waveshaped cracks formed on the ceramic coatings. When a constant load and cyclic thermal stress were applied to the 3D braided composite substrate (as shown in Fig. 1) in the longitudinal direction, all neighbouring two fiber bundles would slip and/or rotate around their contacts and the braiding angle was reduced by transverse compression. SiC ceramic coating, covering the 3D braided C/SiC composite substrate, was opened up due to the constant load and then the opening cracks were corrugated with transverse shrinkage strain. The cyclic thermal stress made the cracks uniformly distributed in the coating and matrix, and the crack spacing was almost equal. No matter how high or low the temperature was during testing, cracks had never closed up completely due to the constant loading. The cracks perpendicular to the substrate surface were most open when temperature was held at 900 C (the resultant stress reached a maximum). Diffused oxygen along the open coating cracks reached the surfaces of the carbon fibers in two directions marked by arrows in Fig. 7(b) and was depleted rapidly by the external fibers. Subsequently the loading was transferred to the internal fibers. Thus, when the new or propagated cracks of the matrix were produced inwards, internal oxidation would take place, and the stress oxidation of carbon fibers in particular was ready to occur upon loading. 4. Conclusions 1. Under a constant load and thermal cycling, a gradually increasing creep strain coupled with cyclic strain could be measured in good order. Thermal strain difference approximated to 0.16 % when the composite was subjected to a constant load of 60 MPa and the temperature difference was about 300 C.
2. Theoretical calculation was in agreement with the observed experimental phenomena. The value of the calculated thermal strain difference was about 0.1566% (experimental value was 0.16%). Maximum resultant stress applied to the specimen obtained by calculation varied from the initial 138.3 MPa to the final 101.26 MPa. 3. Thermal cycling can enhance the strain rate of the composites by forming nonreversible damage and the damage recorded in the tested composites could be replayed upon reloading. The damage strain in testing included crack opening displacement, interfacial debonding and sliding. Elongation of the surviving specimen in the next monotonic tension seems to be limited due to the damage strain. 4. The thermal cycling damage to the modulus was more severe than to the strength for C/SiC composites. After 50 thermal cycles, the modulus of the composites was reduced by a factor of 0.5 while the residual strength still retained 82% of the original strength. 5. Wave-shaped cracks were arranged on the coatings at relatively regular spacing (about 278 lm) and the matrix cracked transversely. A typical superficial oxidation was found along the opening cracks beneath the coating. This should be ascribed to the cyclic thermal stress, constant load, wet oxygen atmosphere and the special braided structure of 3D-C/SiC composites.
Acknowledgements The authors acknowledge the financial support of the Natural Science Foundation of China (Contract No. 90405015) and the National Young Elitists Foundation (Contract No. 50425208). References [1] Lamouroux F, Bourrat X, Sevely J, Naslain R. Carbon 1993;31:1273. [2] Cavalier JC, Lacombe A, Rouges JM. In: Bunsell AR, Lamicq P, Massiah A, editors. Developments in the science and technology of composite materials. London: Elsevier; 1989. p. 99 (in French). [3] Biernacki K, Szyszkowski W, Yannacopoulos S. Compos Part A 1999;30A:1027. [4] Figiel Ł, Kamin´ski M. Comp Struct 2003;81:1865. [5] Webb JE, Singh RN. J Am Ceram Soc 1996;79:2857. [6] Singh RN, Wang H. Compos Eng 1995;5:1287. [7] Yin XW, Cheng LF. Carbon 2002;40:905. [8] Panda PK, Kannan TS, Dubois J, Olagnon C, Fantozzi G. Sci Technol Adv Mater 2002;3:327. [9] Reynaud P, Rouby D, Fantozzi G. Scripta Metall Mater 1994;31:1061. [10] Fantozzi G, Reynaud P, Rouby D. Silic Indus 2001;66:109. [11] Cherouali H, Fantozzi G, Reynaud P, Rouby D. Mater Sci Eng 1998;250A:169. [12] Miyashita Y, Kanda K, Zhu S, Mutoh Y, Mizuno M, McEvily AJ. Int J Fatigue 2002;24:241. [13] Zhu S, Mizuno M, Kagawa Y, Cao JW, Nagano Y, Kaya H. Mater Sci Eng 1997;225A:69.