Available online w sciencedirect. com °scⅰ ence Direct CARBON ELSEVIER Carbon45(2007)21952204 www.elsevier.com/locate/carbon Modeling the effects of thermal and mechanical load cycling on a C/Sic composite in oxygen/argon mixtures Hui Mei", Aifei Cheng, Litong Zhang, Yongdong X National Key Laboratory of Thermostructure Composite Materials, Northwestern Polytechnical Uniuersity, Xi an Shaanxi 710072, PR China Received 27 March 2007: accepted 22 June 2007 Available online 3 July 200 Abstract Analytical solutions of and experimental results on the strain response of a carbon fiber reinforced Sic matrix composite under ther- nal and mechanical load cycling in O,Ar are presented. Thermal strain and mechanical strain were shown to approximately sustain linear relationships with temperature Tand stress a, respectively; whereas baseline strain was considered to be damage-dependent, result ing from a combination of two major contributing mechanisms:(a)a physical mechanism in the form of matrix microcracking accom- panied by fiber debonding, sliding or fracture and (b)a chemical mechanism in the form of the fiber oxidation associated with longitudinally increased compliance. Based on these analyses a theoretical model, taking into account the thermal strain, mechanical strain and baseline strain, was theoretically formulated with respect to the contribution of each on the overall total strain and to their generation mechanisms. The proposed model gave correct and reliable prediction C 2007 Elsevier Ltd. All rights reserved 1. Introduction and overview tal oxygen concentrations to which the C/SiC structure is The application of ceramic matrix composites(CMCs) In this regard, much work has been done to describe and to advanced airframe and propulsion systems for future to model the oxidation of fibres, matrices and interfaces pace transportation vehicles can provide benefits in life, the absence of loading by assuming steady state diffusion of performance, temperature margin, and weight savings [1- the oxidant to the site of oxidation [7-9). Additionally,sev- 3]. One CMC system of interest to the aerospace commu- eral stress-oxidation models have also been developed to nity is carbon fiber reinforced silicon carbide(C/SiC). investigate the effects of the oxidation of constituents on Compared to monolithic ceramics, this material presents the mechanical response or life of the composite material higher toughness and tolerance to the presence of cracks, accounting for the synergistic creep-oxidation interactions which implies a non-catastrophic mode of failure. But at high temperature [6, 10-13]. Although these previous one of the more formidable obstacles to the widespread studies have provided insight into the physics and mathe- use of C/SiC structures is that the carbon fibers within matics of carbon phase oxidation in ceramic composites the cracked matrix oxidize at medium to high temperatures under both unstressed and stressed conditions, these in oxidizing environments, especially in presence of exter- approaches are not readily and directly applicable to sup- nal temperature impacts and/or mechanical stress varia- port the complicated strain response analysis of C/sic tions [4-6]. It is therefore necessary to develop a tool that composites subjected to thermal and mechanical load is capable of determining the extent of oxidation and the cycling in oxidizing environments. Thus, a new model that residual strength and stiffness in the C/SiC component as treats the influences of thermal, mechanical and chemical a function of the time, temperature, stress and environmen- applied conditions on the strain evolution of C/Sic com- posite is needed Corresponding author. Fax: +86 29 88494620 Recently, Mei et al.[14-16] have experimentally E-mailaddress:phdhuimei@yahoo.com(H.Mei) reported some novel findings concerning real-time strain 0008-6223/S- see front matter 2007 Elsevier Ltd. All rights reserved. doi:l0.10l6/ carbon007.06.051
Modeling the effects of thermal and mechanical load cycling on a C/SiC composite in oxygen/argon mixtures Hui Mei *, Laifei Cheng, Litong Zhang, Yongdong Xu National Key Laboratory of Thermostructure Composite Materials, Northwestern Polytechnical University, Xi’an Shaanxi 710072, PR China Received 27 March 2007; accepted 22 June 2007 Available online 3 July 2007 Abstract Analytical solutions of and experimental results on the strain response of a carbon fiber reinforced SiC matrix composite under thermal and mechanical load cycling in O2/Ar are presented. Thermal strain and mechanical strain were shown to approximately sustain linear relationships with temperature T and stress r, respectively; whereas baseline strain was considered to be damage-dependent, resulting from a combination of two major contributing mechanisms: (a) a physical mechanism in the form of matrix microcracking accompanied by fiber debonding, sliding or fracture and (b) a chemical mechanism in the form of the fiber oxidation associated with longitudinally increased compliance. Based on these analyses a theoretical model, taking into account the thermal strain, mechanical strain and baseline strain, was theoretically formulated with respect to the contribution of each on the overall total strain and to their generation mechanisms. The proposed model gave correct and reliable predictions. 2007 Elsevier Ltd. All rights reserved. 1. Introduction and overview The application of ceramic matrix composites (CMCs) to advanced airframe and propulsion systems for future space transportation vehicles can provide benefits in life, performance, temperature margin, and weight savings [1– 3]. One CMC system of interest to the aerospace community is carbon fiber reinforced silicon carbide (C/SiC). Compared to monolithic ceramics, this material presents higher toughness and tolerance to the presence of cracks, which implies a non-catastrophic mode of failure. But one of the more formidable obstacles to the widespread use of C/SiC structures is that the carbon fibers within the cracked matrix oxidize at medium to high temperatures in oxidizing environments, especially in presence of external temperature impacts and/or mechanical stress variations [4–6]. It is therefore necessary to develop a tool that is capable of determining the extent of oxidation and the residual strength and stiffness in the C/SiC component as a function of the time, temperature, stress and environmental oxygen concentrations to which the C/SiC structure is exposed. In this regard, much work has been done to describe and to model the oxidation of fibres, matrices and interfaces in the absence of loading by assuming steady state diffusion of the oxidant to the site of oxidation [7–9]. Additionally, several stress-oxidation models have also been developed to investigate the effects of the oxidation of constituents on the mechanical response or life of the composite material accounting for the synergistic creep-oxidation interactions at high temperature [6,10–13]. Although these previous studies have provided insight into the physics and mathematics of carbon phase oxidation in ceramic composites under both unstressed and stressed conditions, these approaches are not readily and directly applicable to support the complicated strain response analysis of C/SiC composites subjected to thermal and mechanical load cycling in oxidizing environments. Thus, a new model that treats the influences of thermal, mechanical and chemical applied conditions on the strain evolution of C/SiC composite is needed. Recently, Mei et al. [14–16] have experimentally reported some novel findings concerning real-time strain 0008-6223/$ - see front matter 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.carbon.2007.06.051 * Corresponding author. Fax: +86 29 88494620. E-mail address: phdhuimei@yahoo.com (H. Mei). www.elsevier.com/locate/carbon Carbon 45 (2007) 2195–2204
H. Mei et al. Carbon 45(2007)2195-2204 response of a C/SiC composite which was subjected Thermal cycle number, N simultaneously to thermal cycling and mechanical stres 60 in oxidizing atmosphere using an 0790a onmental chamber fixed on the Instron servo-hydraulic ster. Many of these investigations have implied that increase in damage strain of the C/SiC composite was ascribed to the result of the combination of the following two major contributing mechanisms Physical damage mechanism, thermal cycling and 0335 mechanical stress induced matrix microcracking normal to tensile axis, known as crack opening displacement ccompanied by interfacial debonding or sliding(see 0120024003600480060007200840 igs. I and 2 in SM) Chemical damage mechanism, longitudinally increased compliance owing to reduction in the effective load bear- Fig. I. (a) The measured strain curve of the C/SiC composite subjecte ing area resulting from the progressive oxidation of the (b)a close-up of strain within several cycles. The red broken line reinforcing fibers from surface to interior through the represents the baseline strain. (For interpretation of the references in cracks created earlier(see Fig 3 in SM) colour in this figure legend the reader is referred to the web version of this article The aim of the present work is to incorporate the effects of the thermophysical and chemical environmental vari- stress-strain domain up to rupture. The linear deformation is limited up ables into a formulated strain response model on basis of to about 50 MPa(referred to as first-matrix cracking stress oe), after which the above damage mechanisms. In this paper, a strain evo- the behaviour becomes non-linear lution model for C/SiC composites subjected to thermal and mechanical load cycling in an O2/Ar mixture was pre- 2. 2. Thermal cycling tests with externalstress sented. The model mainly accounted for the crack propa gation and multiplication of brittle sic matrix. and Thermal cycling ts were conducted in an oxygen/argon the oxidation of the reinforcing C fibers, both of which system, which was described in detail in Fig. 2 of (14) Thermal cycling increased the baseline strain of the material with time. At was carried out between temperatures of Tiower = 900Cand the same time, thermal strain owing to temperature varia- Tupper s 1200C(ATs 300C)over a period of 120 s and the mean heat tions and mechanical strain owing to stress changes were ing/cooling rates were about 5 K/s and -10 K/s, respectively. During the also taken into account and formulated. The proposed testing, a small fatigue stress of 60*20 MPa(sinusoidal wave, fre- model, without any attempt at fitting, gave correct and reli- dog-bone specimens. Strain was measured directly from the gauge length strain response behaviour r a specific case of a 3D C/ of the specim vere observed with a scanning electron microscope SiC composite subjected to repetitive temperature between (SEM, Hitachi $-4700 900and1200° and cyclic load of60±20 MPa in a 10.4 vol %O,/89.6 vol% Ar mixture 3. Strain results and analysis Following the test procedure described above, a typic 2. Experimental description strain response curve of the tested C/SiC composite speci men to cyclic temperature and fatigue stress was obtained 2.1. Materials and plotted in Fig. 1. Fig. Ib is a close-up observation of As described in [14 the same isothermal CVi process was employed to the strain curve within several cycles from the No. 17th fabricate the 3D C/SiC composites at M1000C in this investigation. Fiber to 20th cycle(i.e, from 2040 s to 2400 s). It should be noted architectures in the as-fabricated C/SiC composite preforms were show from Fig. I that the total strain should be a combined in Fig. 4 in SM. The dog-bone shaped specimens with a dimension of result of thermal strain due to heating/cooling, mechanical 185 x 3x 3 mm were cut from the fabricated composite plates, and fur. strain due to loading/unloading and baseline strain due to er coated with Sic by the same I-CVI processing(thickness s 50 um um ). damage accumulation. As a representative, the strain Table I in SM summarized the major properties of the as-received spec mensFig.5 in SM shown a typical tensile stress-strain curve of the 3d response data within the No. 18th thermal cycle is summa- C/SiC composite specimens. It can be seen that the composite behaves rized in Table 1. It can be seen from Fig. Ib and Table I as a typical damageable material, exhibiting an extensive non-linear thermal strain and mechanical strain linearly increase with 2 A relatively small stress is usually selected to apply on the C/Sic i Some figures and data from the previous investigations were put in the composite because the reinforcing fibers is susceptible to oxidation once ntary material (SM) se the cracks open
response of a C/SiC composite which was subjected simultaneously to thermal cycling and mechanical stress in oxidizing atmosphere using an induction heating environmental chamber fixed on the Instron servo-hydraulic tester. Many of these investigations have implied that increase in damage strain of the C/SiC composite was ascribed to the result of the combination of the following two major contributing mechanisms: • Physical damage mechanism, thermal cycling and mechanical stress induced matrix microcracking normal to tensile axis, known as crack opening displacement, accompanied by interfacial debonding or sliding (see Figs. 1 and 2 in SM1 ). • Chemical damage mechanism, longitudinally increased compliance owing to reduction in the effective load bearing area resulting from the progressive oxidation of the reinforcing fibers from surface to interior through the cracks created earlier (see Fig. 3 in SM). The aim of the present work is to incorporate the effects of the thermophysical and chemical environmental variables into a formulated strain response model on basis of the above damage mechanisms. In this paper, a strain evolution model for C/SiC composites subjected to thermal and mechanical load cycling in an O2/Ar mixture was presented. The model mainly accounted for the crack propagation and multiplication of the brittle SiC matrix, and the oxidation of the reinforcing C fibers, both of which increased the baseline strain of the material with time. At the same time, thermal strain owing to temperature variations and mechanical strain owing to stress changes were also taken into account and formulated. The proposed model, without any attempt at fitting, gave correct and reliable predictions in assessing the experimentally obtained strain response behaviour for a specific case of a 3D C/ SiC composite subjected to repetitive temperature between 900 and 1200 C and cyclic load of 60 ± 20 MPa in a 10.4 vol.% O2/89.6 vol.% Ar mixture. 2. Experimental description 2.1. Materials As described in [14], the same isothermal CVI process was employed to fabricate the 3D C/SiC composites at 1000 C in this investigation. Fiber architectures in the as-fabricated C/SiC composite preforms were shown in Fig. 4 in SM. The dog-bone shaped specimens with a dimension of 185 · 3 · 3 mm3 were cut from the fabricated composite plates, and further coated with SiC by the same I-CVI processing (thickness 50 lm). Table 1 in SM summarized the major properties of the as-received specimens. Fig. 5 in SM shown a typical tensile stress–strain curve of the 3D C/SiC composite specimens. It can be seen that the composite behaves as a typical damageable material, exhibiting an extensive non-linear stress–strain domain up to rupture. The linear deformation is limited up to about 50 MPa (referred to as first-matrix cracking stress rc), after which the behaviour becomes non-linear. 2.2. Thermal cycling tests with external stress Thermal cycling experiments were conducted in an oxygen/argon mixture of 10.4 vol.% O2/89.6 vol.% Ar using a newly developed integrated system, which was described in detail in Fig. 2 of [14]. Thermal cycling was carried out between temperatures of Tlower 900 C and Tupper 1200 C (DT 300 C) over a period of 120 s and the mean heating/cooling rates were about 5 K/s and 10 K/s, respectively. During the testing, a small fatigue stress of 60 ± 20 MPa2 (sinusoidal wave, frequency = 1 Hz, stress ratio R = 0.5) was simultaneously applied to the dog-bone specimens. Strain was measured directly from the gauge length by a contact Instron extensometer (Model A1452-1001B). Microstructures of the specimens were observed with a scanning electron microscope (SEM, Hitachi S-4700). 3. Strain results and analysis Following the test procedure described above, a typical strain response curve of the tested C/SiC composite specimen to cyclic temperature and fatigue stress was obtained and plotted in Fig. 1. Fig. 1b is a close-up observation of the strain curve within several cycles from the No. 17th to 20th cycle (i.e., from 2040 s to 2400 s). It should be noted from Fig. 1 that the total strain should be a combined result of thermal strain due to heating/cooling, mechanical strain due to loading/unloading and baseline strain due to damage accumulation. As a representative, the strain response data within the No. 18th thermal cycle is summarized in Table 1. It can be seen from Fig. 1b and Table 1, thermal strain and mechanical strain linearly increase with 1 Some figures and data from the previous investigations were put in the Supplementary material (SM) section. 2 A relatively small stress is usually selected to apply on the C/SiC composite because the reinforcing fibers is susceptible to oxidation once the cracks open. Fig. 1. (a) The measured strain curve of the C/SiC composite subjected simultaneously to thermal cycling and fatigue stress in O2/Ar mixture and (b) a close-up of strain within several cycles. The red broken line represents the baseline strain. (For interpretation of the references in colour in this figure legend, the reader is referred to the web version of this article.) 2196 H. Mei et al. / Carbon 45 (2007) 2195–2204
H. Mei et al. Carbon 45(2007)2195-2204 Table I Strain response results of the C/SiC composites to thermal cycle and fatigue stress within the No. 18th thermal cycle 7()(K/s)h()(s-) ETh(%) (%) (MPa)H-(%)4a(%) 0.5940.71 0.117 0.682-0.658 0.024 Cooling1200-900°C -3.67×10 0.71l-0.601 60±20 0.663-0.634 0.029 =Fat- Fal gives the mean value of 12 maximum-minimum fatigue strains within the No. 18th thermal cycle. increase of T or o and then linearly decrease with decrease 3. 2. Mechanical strain of T or o. As cycle proceeds, the thermal strain and mechanical strain are periodically repeated with the fixed Generally, a relatively small stress can be considered to strain ranges and as the same period as their excitation apply on the C/sic composite because it is well-known that temperature or stress, although the baseline strain is con- the reinforcing carbon fibers is susceptible to oxidation tinuously increasing. These experimentally observed results once the cracks open. When an applied stress is below are advantageous in modeling the effects of temperature the so-called proportional limit (i.e, oc), the mechanical and stress on the strain strain of the composite is linear elastic and dependent on the stress as the classical hooke s law 3. Thermal strain (4) Assuming that thermal strain is absolutely temperature- Under a cyclic stress, the mechanical strain range can be dependent and follows a linear function as the temperature obtained through ET(o=ar(t) (1)=4 and, its rate can be expressed as where, E is the Youngs modulus of the composite and A stress range aeTn (t) aT(o)-oi(t (2) 3.3 Baseline strain where T(o is the heating/cooling rate, a signifies the coeffi The baseline strain, in fact, is a retained strain of the cient of thermal expansion (CTE)along the composite composite specimen once unloading both thermal and material axis direction. If the composite material is ther- mechanical loads to the bottoms of temperature and of mally cycled between two selected temperatures, the ther- stress. In this paper, it can be specially defined as thermal mal strain range eth between the lower and upper cycling fatigue strain(hereafter, TCF strain) temperatures at each cycle is determined as bottom th=EThpper-ETiower △T and its rate can be written as i is the mean CTE between two temperatures, and AT, ETCF(( dETc(o) temperature difference a0.014 0.012 0.010 0.0025 10K/s 0.008 0.0020 80 MPa 0.0005 0.0000 0050×10310×10415×1042.0×10 50x10310x1041.5×1042.0×104 Time(s) Fig. 2. The I TCF strain as a function of test time where the strain increase is governed by a physical damage mechanism, (a) at different stress of 60, 80, 120 MPa and(b)in different cooling rate of 5, 10, 15. 20 K/s
increase of T or r and then linearly decrease with decrease of T or r. As cycle proceeds, the thermal strain and mechanical strain are periodically repeated with the fixed strain ranges and as the same period as their excitation temperature or stress, although the baseline strain is continuously increasing. These experimentally observed results are advantageous in modeling the effects of temperature and stress on the strain. 3.1. Thermal strain Assuming that thermal strain is absolutely temperaturedependent and follows a linear function as the temperature T eThðtÞ ¼ aT ðtÞ ð1Þ and, its rate can be expressed as e_ThðtÞ ¼ oeThðtÞ ot ¼ a oT ðtÞ ot ¼ aT_ ðtÞ ð2Þ where T_ðtÞ is the heating/cooling rate, a signifies the coeffi- cient of thermal expansion (CTE) along the composite material axis direction. If the composite material is thermally cycled between two selected temperatures, the thermal strain range eR Th between the lower and upper temperatures at each cycle is determined as, e R Th ¼ e T upper Th e T lower Th ¼ aDT ð3Þ a is the mean CTE between two temperatures, and DT, temperature difference. 3.2. Mechanical strain Generally, a relatively small stress can be considered to apply on the C/SiC composite because it is well-known that the reinforcing carbon fibers is susceptible to oxidation once the cracks open. When an applied stress is below the so-called proportional limit (i.e., rc), the mechanical strain of the composite is linear elastic and dependent on the stress as the classical Hooke’s law eMe ¼ r E ð4Þ Under a cyclic stress, the mechanical strain range can be obtained through e R Me ¼ Dr E ð5Þ where, E is the Young’s modulus of the composite and Dr, stress range. 3.3. Baseline strain The baseline strain, in fact, is a retained strain of the composite specimen once unloading both thermal and mechanical loads to the bottoms of temperature and of stress. In this paper, it can be specially defined as thermal cycling fatigue strain (hereafter, TCF strain) eTCF ¼ e rbottom T bottom ð6Þ and its rate can be written as e_TCFðtÞ ¼ oeTCFðtÞ ot ð7Þ Table 1 Strain response results of the C/SiC composites to thermal cycle and fatigue stress within the No. 18th thermal cycle No. T T_ðtÞ ðK=sÞ e_ThðtÞ ðs1Þ eTh ð%Þ eR Th ð%Þ r (MPa) emax Fat –emin Fat ð%Þ eR Fat ð%Þ 18th Heating 900–1200 C 5 1.95 · 105 0.594–0.711 0.117 60 ± 20 0.682–0.658 0.024 Cooling 1200–900 C 10 3.67 · 105 0.711–0.601 0.110 60 ± 20 0.663–0.634 0.029 emax Fat –emin Fat gives the mean value of 12 maximum–minimum fatigue strains within the No. 18th thermal cycle. Fig. 2. The I TCF strain as a function of test time where the strain increase is governed by a physical damage mechanism, (a) at different stress of 60, 80, 100, 120 MPa and (b) in different cooling rate of 5, 10, 15, 20 K/s. H. Mei et al. / Carbon 45 (2007) 2195–2204 2197
H. Mei et al/ Carbon 45(2007)2195-2204 Hence, the TCF strain can be acceptable to act as a damage where n is the number of transverse cracks and UcoD indicator when the CMC composites are subjected to ther- crack opening displacement Crack opening displacement mal cycling and mechanical stress in an oxidizing atmo- of each crack 8 can be simply estimated as [17] sphere. As schemed in the Fig. 1, the slope of the red broken line represents the mean rate of the TCF strain. 8 dMeM cOS (9) The higher the rate, the faster the tCF strain increase 4VFtEr(ErR+ Em/m) and the severer the damage to the composites. If the rate where ga is the applied stress(MPa)and t, the shear sliding becomes zero with continued cycles, the composites go to stress of interface(MPa). The crack growth and multiplica a steady state in which initiation of the new cracks or prop- tion are taken into account to follow the exponential law as agation of the previous cracks is terminated transitorily. A progressively increasing TCF strain rate means that the composite will fracture eventuall B=nBs(am-ad)AT1-exp aeMM(or (EmIm+ErVe 4. Model formulation where Bs is the saturated crack density(m ),n and i are the constants, T is the maximum heating/cooling rate(K/ The derivation of the model relies on the principle that s). Substituting Eqs. (9)and(10)into(8) gives the following the total strain response behaviour of the composite can formulation of the type-I TCF strain for the physical dam be expressed as the sum of three terms: i.e., thermal strain, mechanical strain and TCF strain, For the formulation of age me linear laws as Eqs.()and (4). Thus, the development of ecr -AEm dnps(am-a+)AT cos p the model, the former two terms are simplified to follow the 4VFTErEr Vr+EmIm) the model will focus on the better understanding the phys- ical and chemical mechanisms of the successive tcf strain ×1-exp ZGAEm/m(af -%m) evolution during thermal cycling and mechanical stress in (EmVm+Erve oxidizing atmosphere Note that the above type-I TCF strain is a time-dependent function where the strain magnitude and shape are gov 4. 1. Matrix cracking and fiber sliding contribution erned by the environmental parameters, i.e., GA, T, AT, Consider a model composite of length L, that is rein- Table 2 forced with fibers of equal diameter d that occupy a frac- Parameters and values used in calculation tion Ve of the composite volume. It is assumed that theParameter Symbol Value Units fibers are aligned along the axis at braiding angles and Material coated with a proper Py C layer of finite thickness to impart Young's modulus of matrix 350 composite behaviour to the CFCC and that the distribu cTE of matrix tion of tensile strengths for the fibers and the fiber interfa- Young's modulus of fiber 0-/K 230 cial shear stress, t, are uniform and known. The matrix is volume fraction of fiber assumed to no plastic deformation and the inelastic strain CTE of fiber 0.5 10-/K is mainly derived from the sum of crack opening displace- Fiber diameter ment of transverse crack system, accompanied by the fiber Side of the square specimen cross section a debonding or sliding along a finite region adjacent to the Braiding angle matrix crack. It is also assumed that the composite consists Emvironmental of a matrix of volume fraction Vm with properties Em Applied stress tigue stress rang (Youngs modulus), om, which contains the reinforcing Constant fibers with properties Er, af onstan Consider the case when such the composite is subjected Oxidant partial pressure to thermal cycling and a tensile stress, larger than the Total pressure l01,325Pa matrix cracking stress, oc, so that a series of parallel and Molar density of carbon 150,000mol/m3 qually spaced transverse cracks are formed in the matrix Gas constant 8.31441J/molK Lower temperature limit To (see Fig. I in SM) TCF strain is simply the density Heating/cooling rate 5/-10K/s of transverse matrix microcracks B(i.e, number of cracks Thermal cycling temperature difference AT per m) multiplied by the crack opening displacement of micromechanical each crack sas Saturated crack density 阝s 7000 Crack opening displacement Variable m △L_n;UcoD=δ·β Sliding resistance of interface (8) CTE, Coefficient of thermal expansion
Hence, the TCF strain can be acceptable to act as a damage indicator when the CMC composites are subjected to thermal cycling and mechanical stress in an oxidizing atmosphere. As schemed in the Fig. 1, the slope of the red broken line represents the mean rate of the TCF strain. The higher the rate, the faster the TCF strain increases and the severer the damage to the composites. If the rate becomes zero with continued cycles, the composites go to a steady state in which initiation of the new cracks or propagation of the previous cracks is terminated transitorily. A progressively increasing TCF strain rate means that the composite will fracture eventually. 4. Model formulation The derivation of the model relies on the principle that the total strain response behaviour of the composite can be expressed as the sum of three terms: i.e., thermal strain, mechanical strain and TCF strain. For the formulation of the model, the former two terms are simplified to follow the linear laws as Eqs. (1) and (4). Thus, the development of the model will focus on the better understanding the physical and chemical mechanisms of the successive TCF strain evolution during thermal cycling and mechanical stress in oxidizing atmosphere. 4.1. Matrix cracking and fiber sliding contribution Consider a model composite of length L, that is reinforced with fibers of equal diameter d that occupy a fraction Vf of the composite volume. It is assumed that the fibers are aligned along the axis at braiding angles /, and coated with a proper PyC layer of finite thickness to impart composite behaviour to the CFCC and that the distribution of tensile strengths for the fibers and the fiber interfacial shear stress, s, are uniform and known. The matrix is assumed to no plastic deformation and the inelastic strain is mainly derived from the sum of crack opening displacement of transverse crack system, accompanied by the fiber debonding or sliding along a finite region adjacent to the matrix crack. It is also assumed that the composite consists of a matrix of volume fraction Vm with properties Em (Young’s modulus), am, which contains the reinforcing fibers with properties Ef, af. Consider the case when such the composite is subjected to thermal cycling and a tensile stress, larger than the matrix cracking stress, rc, so that a series of parallel and equally spaced transverse cracks are formed in the matrix (see Fig. 1 in SM). The TCF strain is simply the density of transverse matrix microcracks b (i.e., number of cracks per m) multiplied by the crack opening displacement of each crack d, as e I TCF ¼ DLin L ¼ n UCOD L ¼ d b ð8Þ where n is the number of transverse cracks and UCOD, crack opening displacement. Crack opening displacement of each crack d can be simply estimated as [17] d ¼ dV 2 mE2 m cos u 4V 2 f sEfðEfV f þ EmV mÞ 2 r2 A ð9Þ where rA is the applied stress (MPa) and s, the shear sliding stress of interface (MPa). The crack growth and multiplication are taken into account to follow the exponential law as b ¼ gbsðam afÞDT 1 exp krAEmV mðaf amÞ ðEmV m þ EfV fÞ 2 T_ " #t ( ) ! ð10Þ where bs is the saturated crack density (m1 ), g and k are the constants, T_ is the maximum heating/cooling rate (K/ s). Substituting Eqs. (9) and (10) into (8) gives the following formulation of the type-I TCF strain for the physical damage mechanism, e I TCF ¼ r2 AV 2 mE2 mdgbsðam afÞDT cos u 4V 2 f sEfðEfV f þ EmV mÞ 2 1 exp krAEmV mðaf amÞ ðEmV m þ EfV fÞ 2 T_ " #t ( ) ! ð11Þ Note that the above type-I TCF strain is a time-dependent function where the strain magnitude and shape are governed by the environmental parameters, i.e., rA, T_ , DT, Table 2 Parameters and values used in calculation Parameter Symbol Value Units Material Young’s modulus of matrix Em 350 GPa Volume fraction of matrix Vm 0.6 CTE of matrix am 4.6 106 /K Young’s modulus of fiber Ef 230 GPa Volume fraction of fiber Vf 0.4 CTE of fiber af 0.5 106 /K Fiber diameter d 7 lm Side of the square specimen cross section a 0.003 m Braiding angle / 22 Environmental Applied stress rA 60 MPa Fatigue stress range Dr 40 MPa Constant g 14,346 Constant k 120,050 Oxidant partial pressure v 0.104 % Total pressure P 101,325 Pa Molar density of carbon qc 150,000 mol/m3 Gas constant R 8.31441 J/mol K Lower temperature limit T0 T 1173 K Heating/cooling rate T_ 5/10 K/s Thermal cycling temperature difference DT 300 K Micromechanical Saturated crack density bs 7000 m1 Crack opening displacement d Variable m Sliding resistance of interface s 3 MPa CTE, Coefficient of thermal expansion. 2198 H. Mei et al. / Carbon 45 (2007) 2195–2204
H. Mei et al./ Carbon45(2007)2195-2204 assuming that the material parameters remain constant (ErVr+EmMa(t) Due to page limitation, the effect laws of only those rep- resentative environmental parameters on the strain are cal culated and plotted in the following. The readers may (ErVr+EmIm)a-2x/(Vr cos )I (12) predict other controlled laws of the environmental param- Equivalently, eters of interest by using the proposed models Using the ated for a specific case of the C/SiC composite. Fig. 2 TCE (ErR+ EmIm)1-2x/aVr cos o ) (13) shows the predicted type-I TCF strains for tests conducted at different applied stresses of 60, 80, 100, 120 MPa where x is the recession distance of carbon fibers from the Fig. 2a)and in different cooling rate of 5, 10, 15, 20 K/s surface into the interior of the composite(as also schemed Fig. 2b). The effect of stress can be seen clearly in Fig. 2a. As expected, the higher the applied stress, the phase is assumed to be controlled by diffusion of oxygen TCF strain to saturation(defined as a specific inflexion et al. [7] suggested that the recession distance x can be strain whose first derivative is equal to zero). As stress properly developed, related to the exposure time f,as quadratically. It can be also seen from Fig. 2b that the x=Kp=(4D_P(+2(D/D)+ increases linearly, the TCF strain to saturation is promoted D/D+I (14) higher the cooling rate, the greater the TCF strain rate where Kp is referred to as the oxidation rate. z is the oxi- and the shorter the time to crack saturation. In summary, dant partial pressure, P is the total pressure(Pa),Pc is the parameters in the amplitude of the eq.(11),e.g. the the molar density of carbon(mol/m), R is the gas constant applied stress, significantly influence the magnitude of the (J/mol K), Tis the absolute temperature(K), Dk and d are TCF strain while the parameters in the exponential argu- Knudsen diffusion coefficient and Fick diffusion coefficient, ent, e.g. the heating/cooling rate. mail respectively. The type-II TCF strain, integrating Eqs. (13) and(14), can be rewritten (ErV+ EmVm)[1-2VKpt/(avr cos p) (15) 4.2. Fiber oxidation contribution Eckel et al. also suggested that the kp may be simply devel- Matrix cracks will serve as avenues for the ingress of the oped as a function of the oxidant partial pressure z, the environment atmosphere into the composite. In this analy temperature T and the characteristic dimension of the sis we are concerned with the effects of oxidizing environ- crack opening displacement a ments on the reliability of a CMc through reacting with the carbonaceous fibers in microcracks and the oxidation Kp=6.263/ In 3.8 x 10(1+x)8+11 of the matrix is neglected 3.8×10°6+1 When the oxidizing gas ingresses into the composite a As mentioned previously, 8 could be mainly determined by sequence of events is triggered starting first with the oxida- the external applied stress aa according to Eq. (9). The in tion of the fiber. Especially, in presence of a small applied creased stresses widen the crack and the increased temper- stress aA, the fiber oxidation can result in increase of the atures narrow the cracks. USing the Eq. (16), the effects of TCF strain by reducing the effective load bearing area of dimension of 8 on oxidation rate Kp at different tempera- le reinforcing fibers in a square cross-section of the com- ture and in different oxidant partial pressure are plotted posite specimen with a side length of a(as schemed in in Fig 4a. It is evident that temperature only weakly affects Fig 3). Therefore, the type-II TCF strain will increase with the magnitude and shape of the kp curve in the small 8 the reduction both in effective cross-section area A(n)and in (Sc)regime, the speed the volume-averaged Young modulus by of gas diffusion is much greater than the reaction rate of the ig. 3. Schematic diagrams of fiber oxidation in the cross-section of C/SiC composites
etc., associated with the physical damage mechanism by assuming that the material parameters remain constant. Due to page limitation, the effect laws of only those representative environmental parameters on the strain are calculated and plotted in the following. The readers may predict other controlled laws of the environmental parameters of interest by using the proposed models. Using the data listed in Table 2, TCF strain predictions were generated for a specific case of the C/SiC composite. Fig. 2 shows the predicted type-I TCF strains for tests conducted at different applied stresses of 60, 80, 100, 120 MPa (Fig. 2a) and in different cooling rate of 5, 10, 15, 20 K/s (Fig. 2b). The effect of stress can be seen clearly in Fig. 2a. As expected, the higher the applied stress, the shorter the time to crack saturation and the larger the TCF strain to saturation (defined as a specific inflexion strain whose first derivative is equal to zero). As stress increases linearly, the TCF strain to saturation is promoted quadratically. It can be also seen from Fig. 2b that the higher the cooling rate, the greater the TCF strain rate and the shorter the time to crack saturation. In summary, the parameters in the amplitude of the Eq. (11), e.g. the applied stress, significantly influence the magnitude of the TCF strain while the parameters in the exponential argument, e.g. the heating/cooling rate, mainly change its shape. 4.2. Fiber oxidation contribution Matrix cracks will serve as avenues for the ingress of the environment atmosphere into the composite. In this analysis we are concerned with the effects of oxidizing environments on the reliability of a CMC through reacting with the carbonaceous fibers in microcracks and the oxidation of the matrix is neglected. When the oxidizing gas ingresses into the composite a sequence of events is triggered starting first with the oxidation of the fiber. Especially, in presence of a small applied stress rA, the fiber oxidation can result in increase of the TCF strain by reducing the effective load bearing area of the reinforcing fibers in a square cross-section of the composite specimen with a side length of a (as schemed in Fig. 3). Therefore, the type-II TCF strain will increase with the reduction both in effective cross-section area A(t) and in the volume-averaged Young modulus by e II TCF ¼ rAA0 AðtÞ A0 ðEfV f þ EmV mÞ h iAðtÞ ¼ rAa4 ðEfV f þ EmV mÞ½a 2x=ðV f cos uÞ4 ð12Þ Equivalently, e II TCF ¼ rA ðEfV f þ EmV mÞ½1 2x=ðaV f cos uÞ4 ð13Þ where x is the recession distance of carbon fibers from the surface into the interior of the composite (as also schemed in Fig. 3). In the present study, the oxidation of carbon phase is assumed to be controlled by diffusion of oxygen gas through the matrix microcracks. In this case, Eckel et al. [7] suggested that the recession distance x can be properly developed, related to the exposure time t, as x2 ¼ Kpt ¼ 4D P qcRT ln ð1 þ vÞðDk=DÞ þ 1 Dk=D þ 1 t ð14Þ where Kp is referred to as the oxidation rate. v is the oxidant partial pressure, P is the total pressure (Pa), qc is the molar density of carbon (mol/m3 ), R is the gas constant (J/mol K), T is the absolute temperature (K), Dk and D are Knudsen diffusion coefficient and Fick diffusion coefficient, respectively. The type-II TCF strain, integrating Eqs. (13) and (14), can be rewritten as e II TCF ¼ rA ðEfV f þ EmV mÞ½1 2 ffiffiffiffiffiffiffi Kpt p =ðaV f cos uÞ4 ð15Þ Eckel et al. also suggested that the Kp may be simply developed as a function of the oxidant partial pressure v, the temperature T and the characteristic dimension of the crack opening displacement d Kp ¼ 6:263 1010T 1=2 ln 3:8 106 ð1 þ vÞd þ 1 3:8 106 d þ 1 ð16Þ As mentioned previously, d could be mainly determined by the external applied stress rA according to Eq. (9). The increased stresses widen the crack and the increased temperatures narrow the cracks. Using the Eq. (16), the effects of dimension of d on oxidation rate Kp at different temperature and in different oxidant partial pressure are plotted in Fig. 4a. It is evident that temperature only weakly affects the magnitude and shape of the Kp curve in the small d (6dc = 1 lm) regime. In the large d (Pdc) regime, the speed of gas diffusion is much greater than the reaction rate of the a x x Fig. 3. Schematic diagrams of fiber oxidation in the cross-section of C/SiC composites. H. Mei et al. / Carbon 45 (2007) 2195–2204 2199
0012 3.50E-008 1200c,1atm b 2.50E-008 TCF strain fm 200°c,0.3atm c2.00E-008 0.3atm 0.006 .50E-008 900,0.3atr I Resctiocontrolled 000E+000 08171E8161430020140x°6ox8010710100 Crack opening displacement, &/m Time(s) Fig 4.(a) Effect of crack opening displacement a on recession rate kp and (b) relationships of the Il TCF stain and ArR of the composite with test time t under the stress of 80 MPa and the oxidant partial pressure of o carbon phase leading to the transformation from the diffu-. 76 MPa), at which point the time to failure ff 22.2 h sion-controlled to the reaction-controlled regime. In- (i.e, 80,000 s)and the type-II TCF strain to failure approx creased oxidant partial pressure is shown to accelerate imates to 0.53% as determined in Fig. 4b. Through the the gas diffusion both in small and large 8 regimes. Thus, model of Eq (15 )and the data listed in the Table 2, pre- becomes a significant microstructural parameter and dicted distributions of the type-II TCF strain at different plays a key role in the oxidation kinetics mechanism of car- stress of 40, 60, 80, 100, 120 MPa(Fig. Sa) and in different bon phase(hereafter, 8-controlled regime"). Environmen- oxidant partial pressure of 0.1, 0.3, 0.5, 0.7( Fig 5b)are ob- tal temperature and stress change the oxidation kinetics tained. The effect laws of the various stresses on the type-II mechanism of the carbonaceous CMCs by firstly changing TCF strain in Fig. Sa are qualitatively similar to the oxida the characteristic dimension of the s tion-assisted stress-rupture results of the C/SiC composites Fig. 4b illustrates relationships of the predicted type-II obtained experimentally by Halbig et al. [6]. Namely, the TCF strain using the above model of Eq (15)and the effec- higher the applied stress, the shorter the time to failure tive load bearing area reduction ratio(ARR, i.e. A(O/A(0)) and the faster the TCF strain increases. Note that these with test time t under the stress of 80 MPa and the oxidant curves, similar to the classical creep of the stressed metals, partial pressure of 0.1%. It is obvious that there exists a also show three well-defined regimes: transient, steady close correlation between the Arr and the type -lI TCF state, and tertiary stages. During the transient stage the ini- strain,i.e,where the type-II TCF strain increases most, tial sudden loading is responsible for the linear increase in the arr of the composite decreases most as well. This is TCF strain. Hence, these TCF strain curves start from the a strong indication that the reduction in the effective load different primary strain by the different stress levels sepa- bearing area of the fibers plays a critical role in the rating them. The steady-state stage exhibited by the tCF increased compliance behaviour of the C/SiC composites strain curves indicates that the composite experiences less exposed in the oxidizing atmosphere. This process can be change in compliance and its duration is as long as the time best described using the 'snow ball effect'analogy. When required to slow consume carbon fibers through the cyclic the fibers are subjected to a chemical recession at a con- opening-closing matrix microcracks. The tertiary stage stant tensile stress, the superficial carbon fibers will fail associated with an accelerated deformation as a result of firstly, and then because of global load sharing assump- the oxidation-assisted failure of a large number of fibers. tions, the load originally carried by the now broken fibers Failure of the composite occurs when a critical number will be transferred to the surviving fibers. Consequently, of fibers fail leading to the critical ARR of r the surviving fibers are now subjected to a larger tensile It is evident from Fig. 5b that the higher oxidant partial ress. The composite will eventually fail when the progres- pressure can directly promote the TCf strain rate and ively increasing tensile stress reaches the UTS of the mate- shorten the time to failure by solely affecting the parabolic rials with decreasing ARR as rate constant Kp according to Eq (15). It can be actually A(o observed that the same failure strain of 0.53%o determined A (17) by the critical ARR corresponds to the different time to failure of In≈22.2,ln≈7.6,l1≈4.6,t≈3.3 h in the where R signifies the critical ARR, Amax represents the horizontal abscissa for the different oxidant partial pres- maximum value of the applied stress. That is to say, in sure of 0. 1, 0.3, 0.5 and 0.7. Additionally, as oxidant partial he present experiment the tested C/SiC composite will fail pressure increases its effect on the time to failure is remark once Re reaches 0. 193(aMax=80 MPa and ours= ably weaken, taking on an apparent deactivation because
carbon phase leading to the transformation from the diffusion-controlled to the reaction-controlled regime. Increased oxidant partial pressure is shown to accelerate the gas diffusion both in small and large d regimes. Thus, d becomes a significant microstructural parameter and plays a key role in the oxidation kinetics mechanism of carbon phase (hereafter, ‘‘d-controlled regime’’). Environmental temperature and stress change the oxidation kinetics mechanism of the carbonaceous CMCs by firstly changing the characteristic dimension of the d. Fig. 4b illustrates relationships of the predicted type-II TCF strain using the above model of Eq. (15) and the effective load bearing area reduction ratio (ARR, i.e. A(t)/A(0)) with test time t under the stress of 80 MPa and the oxidant partial pressure of 0.1%. It is obvious that there exists a close correlation between the ARR and the type-II TCF strain, i.e., where the type-II TCF strain increases most, the ARR of the composite decreases most as well. This is a strong indication that the reduction in the effective load bearing area of the fibers plays a critical role in the increased compliance behaviour of the C/SiC composites exposed in the oxidizing atmosphere. This process can be best described using the ‘snow ball effect’ analogy. When the fibers are subjected to a chemical recession at a constant tensile stress, the superficial carbon fibers will fail firstly, and then because of global load sharing assumptions, the load originally carried by the now broken fibers will be transferred to the surviving fibers. Consequently, the surviving fibers are now subjected to a larger tensile stress. The composite will eventually fail when the progressively increasing tensile stress reaches the UTS of the materials with decreasing ARR as Rc ¼ AðtÞ A0 ¼ rA max rUTS ð17Þ where Rc signifies the critical ARR, rAmax represents the maximum value of the applied stress. That is to say, in the present experiment the tested C/SiC composite will fail once Rc reaches 0.193 (rAmax = 80 MPa and rUTS = 413.76 MPa), at which point the time to failure tf 22.2 h (i.e., 80,000 s) and the type-II TCF strain to failure approximates to 0.53% as determined in Fig. 4b. Through the model of Eq. (15) and the data listed in the Table 2, predicted distributions of the type-II TCF strain at different stress of 40, 60, 80, 100, 120 MPa (Fig. 5a) and in different oxidant partial pressure of 0.1, 0.3, 0.5, 0.7 (Fig. 5b) are obtained. The effect laws of the various stresses on the type-II TCF strain in Fig. 5a are qualitatively similar to the oxidation-assisted stress-rupture results of the C/SiC composites obtained experimentally by Halbig et al. [6]. Namely, the higher the applied stress, the shorter the time to failure and the faster the TCF strain increases. Note that these curves, similar to the classical creep of the stressed metals, also show three well-defined regimes: transient, steadystate, and tertiary stages. During the transient stage the initial sudden loading is responsible for the linear increase in TCF strain. Hence, these TCF strain curves start from the different primary strain by the different stress levels separating them. The steady-state stage exhibited by the TCF strain curves indicates that the composite experiences less change in compliance and its duration is as long as the time required to slow consume carbon fibers through the cyclic opening–closing matrix microcracks. The tertiary stage is associated with an accelerated deformation as a result of the oxidation-assisted failure of a large number of fibers. Failure of the composite occurs when a critical number of fibers fail leading to the critical ARR of Rc. It is evident from Fig. 5b that the higher oxidant partial pressure can directly promote the TCF strain rate and shorten the time to failure by solely affecting the parabolic rate constant Kp according to Eq. (15). It can be actually observed that the same failure strain of 0.53% determined by the critical ARR corresponds to the different time to failure of tf1 22.2, tf2 7.6, tf3 4.6, tf4 3.3 h in the horizontal abscissa for the different oxidant partial pressure of 0.1, 0.3, 0.5 and 0.7. Additionally, as oxidant partial pressure increases its effect on the time to failure is remarkably weaken, taking on an apparent deactivation because Fig. 4. (a) Effect of crack opening displacement d on recession rate Kp and (b) relationships of the II TCF stain and ARR of the composite with test time t under the stress of 80 MPa and the oxidant partial pressure of 0.1%. 2200 H. Mei et al. / Carbon 45 (2007) 2195–2204
H. Mei et al./ Carbon 45(2007)2195-2204 b0012 0008120MPa 60 MPa 0010 x07//x∞s t x=0.1 0.002 0000 0.000 0.020×10401060×1080x10 0020×1040×10460×10480×10310×105 Fig. 5. The type-ll TCF strain as a function of test time where the strain increase is governed by a chemical damage mechanism; (a) at different stress of 40, 60, 80, 100, 120 MPa and(b) in different oxidant partial pressure of 0. 1, 0.3. 0.5, 0.7 the effect of the oxidizing gas concentration is restrained in As demonstrated through Eq (20), the total modelled TCF he diffusion-controlled kinetics regime strain behaviour of the composite depends on 43. Total strain response model (1)Environmental parameters including space such as σA,T,To,△ Tand z,etc.), and time t. The theoretically expected tCF strain response behav (2) Material parameters he properties of the matrix iour of the composite can be expressed as the sum of such as Em, am, of the fibres such as Ef, af, braiding Eqs. (l1)and(15) angle and fiber diameter d, and of the composite uch as volume fractions, Vm and ve GAVEEmdnBs (am -ao)AT cos (3)Micromechanical parameters associated with the 4FTEr(ErR+Em/m) interfaces and cracks. such as the interfacial shear stress t, crack opening displacement 8 and the satu p LGAEm'm(axr -)rl rated crack density, Bs (EmVm+ErVe) (E+EaVm)-2√Kp/(ac (18) The total tcf strain distribution at the different stress of 60, 80, 100, 120 MPa and its constitutive principle, tak- ing from an example at 100 MPa, are illustrated in Fig. 6 However, when (=0, the total TCF strain gives As shown in Fig 6a, it is easy to understand the contribu G tions of the initial heating strain INitial, the physical damage CFIt=0 (ErR+Emm) strain ETc and the chemical damage strain tce to the total TCF strain ETCE. The similar three stages can be introduced It must be notable that the thermal expansion of the com- to interpret the total TCF strain evolution with respect to posite because of the initial heating up to the lower temper- the controlling laws of each mechanism. ature limit To in the first thermal cycle is not taken into Firstly. the sudden strain increase of the composites in account in Eq.(18). Consequently, considering the effect the first cycle is ascribed in large part to the initial thermal of the initial thermal strain INitial yield expansion strain up to the selected lower temperature limit Of course, another slight reason for this phenomenon should be the first loading which is taken into account in 3V2 EndiNg(xm-x)△Tcos the chemical damage strain tce when t=0. Secondly, ther- 4τEf(Evr+EmVm)2 mal cycling induced matrix cracking, multiplication, fiber debonding and sliding can be considered to be responsible {1-ex1 for the secondary larger exponential-like strain growth (EmIm +Er until the physical damage reaches saturated. Finally, the tertiary stage is associated with a continuous and slow ErVr+EmVm)[l-2vKpt/aRcos o) increase in the compliance of the composite as a result of Emmm erroe oxidation-assisted fiber failure in a s-controlled kinetics EmIm+ erve regime. Evidently, the three sequential stages are governed by the initially selected lower temperature limit To and K=6.263×10-10m121n 38×10°(1+x)6+ 3.8×10°8+1 ( 20) applied stress level oo, by the subsequently thermal and mechanical cycling induced physical damage, and by the
the effect of the oxidizing gas concentration is restrained in the diffusion-controlled kinetics regime. 4.3. Total strain response model The theoretically expected TCF strain response behaviour of the composite can be expressed as the sum of Eqs. (11) and (15) e T TCF ¼ r2 AV 2 mE2 mdgbsðam afÞDT cos u 4V 2 f sEfðEfV f þ EmV mÞ 2 1 exp krAEmV mðaf amÞ ðEmV m þ EfV fÞ 2 T_ " #t ( ) ! þ rA ðEfV f þ EmV mÞ½1 2 ffiffiffiffiffiffiffi Kpt p =ðaV f cos uÞ4 ð18Þ However, when t = 0, the total TCF strain gives e T TCF t¼0 ¼ rA ðEfV f þ EmV mÞ ð19Þ It must be notable that the thermal expansion of the composite because of the initial heating up to the lower temperature limit T0 in the first thermal cycle is not taken into account in Eq. (18). Consequently, considering the effect of the initial thermal strain eInitial yields e T TCF ¼ e I TCF þ e II TCF þ eInitial ¼ r2 AV 2 mE2 mdgbsðam afÞDT cos u 4V 2 f sEfðEfV f þ EmV mÞ 2 1 exp krAEmV mðaf amÞ ðEmV m þ EfV fÞ 2 T_ " #t ( ) ! þ rA ðEfV f þ EmV mÞ½1 2 ffiffiffiffiffiffiffi Kpt p =ðaV f cos uÞ4 þ EmV mam þ EfV faf EmV m þ EfV f T 0 and Kp ¼ 6:263 1010T 1=2 ln 3:8 106 ð1 þ vÞd þ 1 3:8 106 d þ 1 ð20Þ As demonstrated through Eq. (20), the total modelled TCF strain behaviour of the composite depends on (1) Environmental parameters including space such as rA, T_ , T0, DT and v, etc.), and time t. (2) Material parameters, i.e., the properties of the matrix such as Em, am, of the fibres such as Ef, af, braiding angle / and fiber diameter d, and of the composite such as volume fractions, Vm and Vf. (3) Micromechanical parameters associated with the interfaces and cracks, such as the interfacial shear stress s, crack opening displacement d and the saturated crack density, bs. The total TCF strain distribution at the different stress of 60, 80, 100, 120 MPa and its constitutive principle, taking from an example at 100 MPa, are illustrated in Fig. 6. As shown in Fig. 6a, it is easy to understand the contributions of the initial heating strain eInitial, the physical damage strain eI TCF and the chemical damage strain eII TCF to the total TCF strain eT TCF. The similar three stages can be introduced to interpret the total TCF strain evolution with respect to the controlling laws of each mechanism. Firstly, the sudden strain increase of the composites in the first cycle is ascribed in large part to the initial thermal expansion strain up to the selected lower temperature limit. Of course, another slight reason for this phenomenon should be the first loading which is taken into account in the chemical damage strain eII TCFwhen t = 0. Secondly, thermal cycling induced matrix cracking, multiplication, fiber debonding and sliding can be considered to be responsible for the secondary larger exponential-like strain growth until the physical damage reaches saturated. Finally, the tertiary stage is associated with a continuous and slow increase in the compliance of the composite as a result of oxidation-assisted fiber failure in a d-controlled kinetics regime. Evidently, the three sequential stages are governed by the initially selected lower temperature limit T0 and applied stress level r0, by the subsequently thermal and mechanical cycling induced physical damage, and by the Fig. 5. The type-II TCF strain as a function of test time where the strain increase is governed by a chemical damage mechanism; (a) at different stress of 40, 60, 80, 100, 120 MPa and (b) in different oxidant partial pressure of 0.1, 0.3, 0.5, 0.7. H. Mei et al. / Carbon 45 (2007) 2195–2204 2201
2202 H. Mei et al/ Carbon 45(2007)2195-2204 b0.024 0.018 0.02 c0.018 0015 0.009 90012 0009 60 MPa 0.006 0003 0000 1x104 01x1042×1043x1044×1045×1046x104 Time(s) Time(s) Fig. 6.(a)Contributions of the initial heating strain initial, physical damage strain EtCE and chemical damage strain etc to the total TCF strain ETC. (b)A amily of the total TCF strain vs time curves using the comprehensive TCF strain model at the different stress of 60, 80, 100, 120 MPa. continuously oxidation-assisted chemical recession, respec- strains at the higher stresses exhibit the pronounced higher tively. Fig. 6b shows that all the predicted total TCF strains to saturation in the secondary regime strains at the different stress levels have alike three-stage characteristics. These TCF strain curves start from the almost same initial thermal expansion strain, and then 4.4. Comparison with experimental results are separated by the different strain rate, the different strain to saturation and the different time to failure. As we know, Using the model of Eq (20) and data listed in the Table when a constant load is applied to C/Sic at elevated oxidiz- 2, the calculational and experimental results of the tcF ing temperatures, the applied stress opens the as-fabricated strain for the tested C/SiC composite specimen during 70 cracks and allows for easier ingress of oxygen to the fibers. thermal cycles with AT= 300C and fatigue stress of At sufficiently high stress, cracks may be open too wide for 60+ 20 MPa are presented in Fig. 7a. It can be seen that crack closure and sealing to occur. Thus the applied stress the model gives very accurate prediction. Furthermore, in will in turn play a critical role on the oxidation process of the present investigation, the total strain of the material the internal fibers by forming the wider crack opening dis- is depicted as the sum of a line ear cont lution correspond placement 8. In these TCF strain curves, the effects of stres- ing to reversible deformation(i. e, thermal strain owing to s on the lives of the composites can be clearly seen. As heating/cooling and mechanical strain owing to reloading. expected, the predicted TCF strains at higher stresses fall unloading) and a non-linear contribution corresponding to to the left side of the plot while those at lower stresses the irreversible deformation resulting from the physical and are concentrated to the right side of the plot. Another effect chemical damage. Hence, the throughout strain response from stress is the strain to saturation. The predicted TCF curve of the composite can be greatly simplified as the Thermal cycle number, N Thermal cycle number, N 6070 0.70 0.65- 725 8 b0a85 0.50 00400 Calculational Expermental 0.35 0205 01200240036004800600072008400 01200240036004800600072008400 Time(s) Fig. 7. Comparison of calculational results to experimental observations for the tested C/SiC composite specimen during 70 thermal cycles with AT=300C and fatigue stress of 60+ 20 MPa; (a) TCF strain and(b)entire strain response curve
continuously oxidation-assisted chemical recession, respectively. Fig. 6b shows that all the predicted total TCF strains at the different stress levels have alike three-stage characteristics. These TCF strain curves start from the almost same initial thermal expansion strain, and then are separated by the different strain rate, the different strain to saturation and the different time to failure. As we know, when a constant load is applied to C/SiC at elevated oxidizing temperatures, the applied stress opens the as-fabricated cracks and allows for easier ingress of oxygen to the fibers. At sufficiently high stress, cracks may be open too wide for crack closure and sealing to occur. Thus the applied stress will in turn play a critical role on the oxidation process of the internal fibers by forming the wider crack opening displacement d. In these TCF strain curves, the effects of stresses on the lives of the composites can be clearly seen. As expected, the predicted TCF strains at higher stresses fall to the left side of the plot while those at lower stresses are concentrated to the right side of the plot. Another effect from stress is the strain to saturation. The predicted TCF strains at the higher stresses exhibit the pronounced higher strains to saturation in the secondary regime. 4.4. Comparison with experimental results Using the model of Eq. (20) and data listed in the Table 2, the calculational and experimental results of the TCF strain for the tested C/SiC composite specimen during 70 thermal cycles with DT = 300 C and fatigue stress of 60 ± 20 MPa are presented in Fig. 7a. It can be seen that the model gives very accurate prediction. Furthermore, in the present investigation, the total strain of the material is depicted as the sum of a linear contribution corresponding to reversible deformation (i.e., thermal strain owing to heating/cooling and mechanical strain owing to reloading/ unloading) and a non-linear contribution corresponding to the irreversible deformation resulting from the physical and chemical damage. Hence, the throughout strain response curve of the composite can be greatly simplified as the Fig. 6. (a) Contributions of the initial heating strain einitial, physical damage strain eI TCF and chemical damage strain eII TCF to the total TCF strain eT TCF. (b) A family of the total TCF strain vs. time curves using the comprehensive TCF strain model at the different stress of 60, 80, 100, 120 MPa. Fig. 7. Comparison of calculational results to experimental observations for the tested C/SiC composite specimen during 70 thermal cycles with DT = 300 C and fatigue stress of 60 ± 20 MPa; (a) TCF strain and (b) entire strain response curve. 2202 H. Mei et al. / Carbon 45 (2007) 2195–2204
H. Mei et al./ Carbon45(2007)2195-2204 sum of the thermal strain, mechanical strain and TCf on environmental parameters, material parameters as well strain. related to test time t. as as on micromechanical properties of fibres, matrix and Mam error interface. The model takes into account thermophys ETcE(n0)+ Emm+Ev. r(t-ne) and chemical effects on the strain evolution of the compos- ite with time, and its formulation relies on the concept of △G 2(EmVm+Erve) (21) contribution of the matrix cracking, multiplication, of the fiber debonding, sliding and oxidation, both individuals where n is thermal cycle number (i.e, 0, 1, 2,,. )and can and combinations. The model was utilized in analysing be obtained from a integer conversion, with respect to the the effect of the environmental parameters such as stress, time t and period 0, as cooling rate, oxidant partial pressure, etc. on composite performance, as well as in linking the micromechanical (22) Interactions with the macroscopic response of the material The model with Fig 7b plots the experimental variation and predicted peak fitting, was successful in assessing the experimentally results of the total strain for the tested C/SiC composite obtained thermal cycling strain response behaviour for a pecimen as a function of test time. Despite slight over pre- specific case of a C/Sic composite in presence of mechan diction at the end of the curve, it is clear, without any at- ical fatigue and oxidizing atmosphere tempt at fitting, that the strain model of Eq.(21)not only matches the order of the observed strain, but also fol- Acknowledgements lows the correct trend with increasing strain. The slight deviation between the model and experimental data is The work is supported by the Natural Science Founda- ikely to be ascribed to the fact that the significant CTE tion of China ( Contract No. 90405015)and National parameter in this model also undergoes degradation with Young Elitists Foundation(Contract No. 50425208) damage aggravation of the composites cycle by cycle. In Appreciation is also extended to the Program for Changi the interest of simplicity but without losing generality, in ang Scholars and Innovative Research Team in university contrast to the experimental observations in Fig. 1, the en (PCSIRT tire strain prediction illustrated in Fig. 8 exhibits a intrinsic nature of the thermal cycling strain response of the C/Sic Appendix A Supplementary material composite under mechanical stress and oxidizing atmo- sphere, although neglecting the actual thermal inertia and Supplementary data associated with this article can be linearly treating thermal strain and mechanical strain. found, in the online version, at doi: 10.1016/j carbon. 2007.06.051 5. Summary and conclusion rences A strain response model has been developed for carbo- naceous fiber-reinforced ceramic matrix composites based [I Naslain R. preparation and properties of non-oxide CMCs for application in engines and nuclear reactors: an overview. Comp Sci Technol2004:64:155-7 CN [2]Christin F Design, fabrication C/C application of C/SiC and SiC/SiC 010203040506070 composites. In: Krenkel W, Naslain R, Schneider H, editors. High temperature ceramic matrix composites, vol. 4. Weinheim: Wiley VCH Press;2001.p.731-43. 3] Schmidt S, Beyer S, Knabe H, Immich H, Meistring R, Gessler A Advanced ceramic matrix composite materials for current and future propulsion technology applications. Acta Astronautica 2004: 55 0.530 [4] Yin xW, Cheng LF Thermal shock behavior of 3-dimensional C/SiC Carbon2002(40):905-10 5] Mall S, Engesser JM. Effects of frequency on fatigue behavior of CVI C/SiC at elevated temperature. Comp Sci Techne [6] Halbig MC, Brewer DN, Eckel AJ. Degradation nuous fiber composites under constant load AS NASA TM-209681, January, 2000 1200240036004800600072008400 [7 Eckel AJ, Cawley JD, Parthasarathy TA. Oxidation kinetics of a continuous carbon phase in a nonreactive matrix. J Am Ceram Soc 1995:78(4)972-80. Fig 8. Predicted entire strain response curve and a close-up of several [8] Sullivan RM. A model for the oxidat carbon silicon carbide representative thermal cycles, using the comprehensive strain model that is omposite structures. Carbon 2005: 43 depicted as the sum of the TCF strain baseline, thermal strain and fatigue [9] Lamouroux F, Naslain R, Jouin JM. Kinetics and mechanisms of strain, when the composite is subjected to 70 thermal cycles with oxidation of 2D woven C/SiC composites: Il, theoretical approach. J △T=300° C and fatigue stress of60±20MP Am Ceram Soc1994;77(8):2058-68
sum of the thermal strain, mechanical strain and TCF strain, related to test time t, as e ¼ e T TCFðnhÞ þ EmV mam þ EfV faf EmV m þ EfV f T_ðt nhÞ DrA 2ðEmV m þ EfV fÞ ð21Þ where n is thermal cycle number (i.e., 0, 1, 2, 3,...) and can be obtained from a integer conversion, with respect to the time t and period h, as n ¼ t h h i ð22Þ Fig. 7b plots the experimental variation and predicted peak results of the total strain for the tested C/SiC composite specimen as a function of test time. Despite slight over prediction at the end of the curve, it is clear, without any attempt at fitting, that the strain model of Eq. (21) not only matches the order of the observed strain, but also follows the correct trend with increasing strain. The slight deviation between the model and experimental data is likely to be ascribed to the fact that the significant CTE parameter in this model also undergoes degradation with damage aggravation of the composites cycle by cycle. In the interest of simplicity but without losing generality, in contrast to the experimental observations in Fig. 1, the entire strain prediction illustrated in Fig. 8 exhibits a intrinsic nature of the thermal cycling strain response of the C/SiC composite under mechanical stress and oxidizing atmosphere, although neglecting the actual thermal inertia and linearly treating thermal strain and mechanical strain. 5. Summary and conclusion A strain response model has been developed for carbonaceous fiber-reinforced ceramic matrix composites based on environmental parameters, material parameters as well as on micromechanical properties of fibres, matrix and interface. The model takes into account thermophysical and chemical effects on the strain evolution of the composite with time, and its formulation relies on the concept of contribution of the matrix cracking, multiplication, of the fiber debonding, sliding and oxidation, both individuals and combinations. The model was utilized in analysing the effect of the environmental parameters such as stress, cooling rate, oxidant partial pressure, etc. on composite performance, as well as in linking the micromechanical interactions with the macroscopic response of the material to applied deformation. The model, without any attempt at fitting, was successful in assessing the experimentally obtained thermal cycling strain response behaviour for a specific case of a C/SiC composite in presence of mechanical fatigue and oxidizing atmosphere. Acknowledgements The work is supported by the Natural Science Foundation of China (Contract No. 90405015) and National Young Elitists Foundation (Contract No. 50425208). Appreciation is also extended to the Program for Changjiang Scholars and Innovative Research Team in university (PCSIRT). Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.carbon. 2007.06.051. References [1] Naslain R. preparation and properties of non-oxide CMCs for application in engines and nuclear reactors: an overview. Comp Sci Technol 2004;64:155–7. [2] Christin F. Design, fabrication C/C application of C/SiC and SiC/SiC composites. In: Krenkel W, Naslain R, Schneider H, editors. High temperature ceramic matrix composites, vol. 4. Weinheim: WileyVCH Press; 2001. p. 731–43. [3] Schmidt S, Beyer S, Knabe H, Immich H, Meistring R, Gessler A. Advanced ceramic matrix composite materials for current and future propulsion technology applications. Acta Astronautica 2004;55: 409–20. [4] Yin XW, Cheng LF. Thermal shock behavior of 3-dimensional C/SiC composite. Carbon 2002(40):905–10. [5] Mall S, Engesser JM. Effects of frequency on fatigue behavior of CVI C/SiC at elevated temperature. Comp Sci Technol 2006;66:863–74. [6] Halbig MC, Brewer DN, Eckel AJ. Degradation of continuous fiber ceramic matrix composites under constant load conditions. NASA/ TM-209681, January, 2000. [7] Eckel AJ, Cawley JD, Parthasarathy TA. Oxidation kinetics of a continuous carbon phase in a nonreactive matrix. J Am Ceram Soc 1995;78(4):972–80. [8] Sullivan RM. A model for the oxidation of carbon silicon carbide composite structures. Carbon 2005;43:275–85. [9] Lamouroux F, Naslain R, Jouin JM. Kinetics and mechanisms of oxidation of 2D woven C/SiC composites: II, theoretical approach. J Am Ceram Soc 1994;77(8):2058–68. Fig. 8. Predicted entire strain response curve and a close-up of several representative thermal cycles, using the comprehensive strain model that is depicted as the sum of the TCF strain baseline, thermal strain and fatigue strain, when the composite is subjected to 70 thermal cycles with DT = 300 C and fatigue stress of 60 ± 20 MPa. H. Mei et al. / Carbon 45 (2007) 2195–2204 2203
H. Mei et al. Carbon 45(2007)2195-2204 [10 Casas L, Martinez-Esnaola JM. Modelling the effect of oxidation on [14] Mei H, Cheng LF, Zhang LT, Xu YD. Effect of temperature he creep behaviour of fibre- reinforced ceramic matrix composites. gradients and stress levels on damage of C/SiC composites in Acta Mater2003;51:3745-57 xidizing atmosphere. Mater Sci Eng A 2006: 430: 314-9. [ll] Lara-Curzio E. Analysis of oxidation-assisted stress-rupture of [15] Mei H, Cheng LF, Zhang LT. Thermal cycling damage mechanisms continuous fiber-reinforced ceramic matrix composites at intermedi- of C/SiC composites in displacement constraint and oxidizing ate temperatures. Comp Part A 1999: 30: 549-54 mosphere. J Am Ceram Soc 2006: 89(7): 2330-4 12] Reynaud P. Rouby D, Fantozzi G. Effects of temperature and of [16] Mei H, Cheng LF, Zhang LT. Damage mechanisms of C/SiC xidation on the interfacial shear stress between fibres and matrix in composites subjected to constant load and thermal cycling in ceramic-matrix composites. Acta mater 1998: 46: 2461-9. xidizing atmosphere. Scripta Mater 2006: 54: 163-8 13]Dutta I Role of interfacial and matrix creep during thermal cycling of [17] Begley MR, Cox BN, McMeeking RM. Creep crack growth with ontinuous fiber reinforced metal matrix composites. Acta mater small scale bridging in ceramic matrix composites. Acta mater 2000:48:1055-74 1997:45(7):2897-909
[10] Casas L, Martinez-Esnaola JM. Modelling the effect of oxidation on the creep behaviour of fibre-reinforced ceramic matrix composites. Acta Mater 2003;51:3745–57. [11] Lara-Curzio E. Analysis of oxidation-assisted stress-rupture of continuous fiber-reinforced ceramic matrix composites at intermediate temperatures. Comp Part A 1999;30:549–54. [12] Reynaud P, Rouby D, Fantozzi G. Effects of temperature and of oxidation on the interfacial shear stress between fibres and matrix in ceramic-matrix composites. Acta mater 1998;46:2461–9. [13] Dutta I. Role of interfacial and matrix creep during thermal cycling of continuous fiber reinforced metal matrix composites. Acta mater 2000;48:1055–74. [14] Mei H, Cheng LF, Zhang LT, Xu YD. Effect of temperature gradients and stress levels on damage of C/SiC composites in oxidizing atmosphere. Mater Sci Eng A 2006;430:314–9. [15] Mei H, Cheng LF, Zhang LT. Thermal cycling damage mechanisms of C/SiC composites in displacement constraint and oxidizing atmosphere. J Am Ceram Soc 2006;89(7):2330–4. [16] Mei H, Cheng LF, Zhang LT. Damage mechanisms of C/SiC composites subjected to constant load and thermal cycling in oxidizing atmosphere. Scripta Mater 2006;54:163–8. [17] Begley MR, Cox BN, McMeeking RM. Creep crack growth with small scale bridging in ceramic matrix composites. Acta mater 1997;45(7):2897–909. 2204 H. Mei et al. / Carbon 45 (2007) 2195–2204
