Scripta Metallurgica et Materialia, Vol 32, No, 4, pp 505-509, 1995 Printed in the USA. All rights reserv THE ROLE OF COATING COMPLIANCE AND FIBER/MATRIX INTERFACIAL TOPOGRAPHY ON DEBONDING IN CERAMIC COMPOSITES RJKerans Wright Laboratory, Materials Directorate WL/MLLM, Wright-Patterson AFB, OH 45433 (Received July 13, 1994) vised September 14, 1994) It seems evident and is widely accepted that fiber- matrix interface properties are key determinants of composite properties; specifically, a relatively weak bond and low friction etween matrix and fiber are necessary conditions for tough, damage tolerant behavior (1,2). It is clear that for a matrix crack to bypass a fiber, two conditions are necessary. The fiber-matrix interface must fracture, and some sliding must occur for the crack to open(2). These events are governed by several properties. The post- fracture sliding is controlled by the residual stress coefficient of friction, interfacial fracture topography and wear properties. Interface fracture is controlled by all of these factors and the fracture toughness of the interface. The role of interfacial fracture topography is only now becoming appreciated. In the systems evaluated to date, fracture surface roughness accounts for a greater portion of the final friction than do the thermal residual stresses(3) All ceramic composites that demonstrate good damage tolerance have at least one layer of C or Bn between the fiber and matrix. It is widely held that the thickness of the layers are aportant parameters in determining the performance of a composite. This is based on immediately apparent why this should be if the role of the C layer is as generally presume solely to provide a weak layer to allow matrix cracks to deflect along the interface and thereby bypass the fiber. While this is a mechanics problem which remains unsolved, it seems that a few atomic layers should be sufficient to perform that function Another aspect of thickness effects is evidenced by the different requirements associated h diffcrent matrix systems. The naturally occurring C rich layer, typically less than 100 nm k, which forms during processing is sufficient in glass and glass ceramic composites(5) However, SiC matrix composites, which have similar naturally occurring layers, require an additional 100 nm to I um of C or BN(6) It has been suggested that the compliance of the coating layer may play a role in these eftects by adjusting the residual stress state(2, 6). However, within the conventional wisdom or the failure process, that has not seemed to adequately account for the magnitude of the effects In this work, the implications of roughness related coating thickness/compliance effects vere evaluated by estimating the maximu bable roughness of NICALONtm fibers, and calculating: 1. the effects on stress state after sliding of the fiber, 2. the implied ximum debond length and 3. the probable pullout length, as functions of applied coating thickness
Saipta Metallurgica et Matcrialia, Vol. 32, No. 4, pp. SOS-509,199s 1994 E%evier Science Ltd Printed in the USA. All rights resewed 0956-716X/95 $9.50 + .OO THE ROLE OF COATING COMPLIANCE AND FIBER/MATRIX INTERFACIAL TOPOGRAPHY ON DEBONDING IN CERAMIC COMPOSITES R. J. Kerans Wright Laboratory, Materials Directorate WL/MLLM, Wright-Patterson AFB, OH 45433 (Received July 13,1994) (Revised September 14,1994) Introduction It seems evident and is widely accepted that fiber - matrix interface properties are key determinants of composite properties; specifically, a relatively weak bond and low friction between matrix and fiber are necessary conditions for tough, damage tolerant behavior (1,2). It is clear that for a matrix crack to bypass a fiber, two conditions are necessary. The fiber - matrix interface must fracture, and some sliding must occur for the crack to open (2). These events are governed by several properties. The post - fracture sliding is controlled by the residual stress, coefficient of friction, interfacial fracture topography and wear properties. Interface fracture is controlled by all of these factors and the fracture toughness of the interface. The role of interfacial fracture topography is only now becoming appreciated. In the systems evaluated to date, fracture surface roughness accounts for a greater portion of the final friction than do the thermal residual stresses (3). All ceramic composites that demonstrate good damage tolerance have at least one layer of C or BN between the fiber and matrix. It is widely held that the thickness of the layers are important parameters in determining the performance of a composite. This is based on abundant anecdotal evidence as well as the more controlled study of Lowden (4). It is not immediately apparent why this should be if the role of the C layer is as generally presumed: solely to provide a weak layer to allow matrix cracks to deflect along the interface and thereby bypass the fiber. While this is a mechanics problem which remains unsolved, it seems that a few atomic layers should be sufficient to perform that function. Another aspect of thickness effects is evidenced by the different requirements associated with different matrix systems. The naturally occurring C-rich layer, typically less than 100 nm thick, which forms during processing is sufficient in glass and glass ceramic composites (5). However, Sic matrix composites, which have similar naturally occurring layers, require an additional 100 nm to 1 pm of C or BN (6). It has been suggested that the compliance of the coating layer may play a role in these effects by adjusting the residual stress state (2,6). However, within the conventional wisdom on the failure process, that has not seemed to adequately account for the magnitude of the effects. In this work, the implications of roughness related coating thickness/compliance effects were evaluated by estimating the maximum probable roughness of NICALONtm fibers, and calculating: 1. the effects on stress state after sliding of the fiber, 2. the implied maximum debond length and 3. the probable pullout length, as functions of applied coating thickness. 505
DEIBONDING IN CERAMIC COMPOSITE Interfacial Properties and Topography The materials considered as the basis for this study are C coated NiCalontm reinforced SiC by chemical vapor infiltration(CvI). In virtually all composites reinforced with NICalonm, the fibers develop one or more layers on the surface during processing as the result of degradation of the fiber(5, 6). There is generally a layer of essentially C 10 to 100 rum thick and usually a subsequent layer of Sio2 of comparable thickness. This is followed by the applied layer of c which is generally 0. 1 to 1 μ m thick(6)(Fig1). A thorough study of the surface topography of the Nicalon fiber has not yet been reported, however, there is good evidence that there is sufficient roughness to produce a significant effect, and to allow a reasonable estimation of the maximum amplitude. Preliminary Atomic Force Microscopy mplies that the amplitude A<24 nm(7). An estimate based on pushout behavior is consistent with that value. Jero et al. have observed that the frictional force observed during rogressive debonding(ie under small displacements)is much less than that measured after complete debonding( B). The approximate magnitude of this effect can be measured using the"pushback test"or comparing the final friction to that during progressive debonding, using a model with sufficient power to determine the latter. A suitable model of roughness effects can then be used to calculate the amplitude of the roughness. The elastic mismatch approach of Kerans and Parthasarathy ly(9),while simplistic and sure to overestimate the elastic effects somewhat, has proven quite consistent with bserved effects and has been utilized here Interface properties and roughness amplitude were determined using a representative well- behaved pushout specimen of NICALON/C/SiC (10). The thermal stresses were assumed to be given by AT=1000 C and the interface properties were calculated to be: strain energy release rate G =2 J/m2 and coeffecient of friction H=0.05. The difference in normal stress across the interface during and following progressive debonding, or(R), was found to be 240 MPa. The roughness implied by this difference was determined by the numerical solution of an exact elastic analysis, including coating layers, using the NDSANDS computer program(11, 12)and was found to be 20nm. The low value for coefficient of 0.05 implies that the roughness may be much less and is significantly greater than the calculated value. Nevertheless, this value is reasonably consistent with the afm value and since the objective of this work is to determine if roughness effects can be expected to be significant, a value that is somewhat high is preferable to one that is too low. Consequently analysis was conducted using A=24 nm, representing the probable maximum roughness of IICALONEm The effective period of the roughness is also problematical but a reasonable approximation can be developed from the"seating drop"of the pushback test results of jero. The length of the seating change of length of the loaded debonded fiber as compared to its length in the originally bondo drop phenomenon in a load-deflection curve should be equal to the period of the roughness plus position. The on for the later contribution is that the change in length will cause the reseating Phenomenon to occur at slightly different positions along the length of the fiber thereby spreading the phenomenon along the displacement axis by the difference in length. Eight pushback tests on Nicalon n1723 glass yieldλ=03±02μm Two sets of material representing the range of reaction layers and coatings generally (6)were considercd. Material I. comprises f=0. 4 volume fraction Nicalon fiber R-Sum with layers of Inm C followed by 10nm silica followed by a C layer of thickness t=0, 50nm, 100nm, 500nm, or 1um. Material Il comprises f=0. 4 volume fraction Nicalon fiber R=8um with layers of 100nm C followed by Onm silica followed by a C layer of thickness t=0, 0.5um, lum, or 2um. The inner two layers represent the degradation induced layers and the outer one the applied C layer. The assumed elastic properties of the constituent materials are indicated in Table 1. The largest sources of error are
506 DEBONDINOIN cERAMIccoMPosm Vol. 32, No. 4 Interfacid Pronerties and ToDoeraDhy The materials considered as the basis for this study are C coated NICALONtm reinforced Sic by chemical vapor infiltration (CVI). In virtually all composites reinforced with NICALONtm, the fibers develop one or more layers on the surface during processing as the result of degradation of the fiber (5,6). There is generally a layer of essentially C 10 to 100 run thick and usually a subsequent layer of Si02 of comparable thickness. This is followed by the applied layer of C which is generally 0.1 to 1 urn thick (6) (Fig 1). A thorough study of the surface topography of the Nicalon fiber has not yet been reported, however, there is good evidence that there is sufficient roughness to produce a significant effect, and to allow a reasonable estimation of the maximum amplitude. Preliminary Atomic Force Microscopy (AFM) work by Jero implies that the amplitude A 5 24 nm (7). An estimate based on pushout behavior is consistent with that value. Jero et al. have observed that the frictional force observed during progressive debonding (i.e. under small displacements) is much less than that measured after complete debonding (8). The approximate magnitude of this effect can be measured using the “pushback test” or by comparing the final friction to that during progressive debonding, using a model with sufficient power to determine the latter. A suitable model of roughness effects can then be used to calculate the amplitude of the roughness. The elastic mismatch approach of Kerans and Parthasarathy (9), while simplistic and sure to overestimate the elastic effects somewhat, has proven quite consistent with observed effects and has been utilized here. Interface properties and roughness amplitude were determined using a representative wellbehaved pushout specimen of NICALON/C/SiC (10). The thermal stresses were assumed to be given by AT = 1000 C and the interface properties were calculated to be: strain energy release rate G = 2.6 J/m* and coeffecient of friction u = 0.05. The difference in normal stress across the interface during and following progressive debonding, or(R), was found to be 240 MPa. The roughness implied by this difference was determined by the numerical solution of an exact elastic analysis, including coating layers, using the NDSANDS computer program (11,12) and was found to be 20nm. The low value for friction coefficient of 0.05 implies that the roughness may be much less and is probably not significantly greater than the calculated value. Nevertheless, this value is reasonably consistent with the AFM value and since the objective of this work is to determine if roughness effects can be expected to be significant, a value that is somewhat high is preferable to one that is too low. Consequently, analysis was conducted using A=24 nm, representing the probable maximum roughness of NICALONt”‘. The effective period of the roughness is also problematical but a reasonable approximation can be developed from the “seating drop” of the pushback test results of Jero. The length of the seating drop phenomenon in a load-deflection curve should be equal to the period of the roughness plus the change of length of the loaded debonded fiber as compared to its length in the originally bonded position. The reason for the later contribution is that the change in length will cause the reseating phenomenon to occur at slightly different positions along the length of the fiber thereby spreading the phenomenon along the displacement axis by the difference in length. Eight pushback tests on Nicalon in 1723 glass yield h = 0.3 + 0.2 urn. Two sets of material representing the range of reaction layers and coatings generally reported (6) were considered. Material I. comprises f=O.4 volume fraction Nicalon fiber R=8pm with layers of 1Onx-n C followed by 1Onm silica followed by a C layer of thickness t=O, 5Onm, lOOrun, 5OOnm, or lprn. Material II. comprises f=O.4 volume fraction Nicalon fiber R=Bum with layers of 1OOnm C followed by 1OOnm silica followed by a C layer of thickness t=O, 0.5um, lum, or 2um. The inner two layers represent the degradation induced layers and the outer one the applied C layer. The assumed elastic properties of the constituent materials are indicated in Table 1. The largest sources of error are
Vol 32, No 4 DEBONDING IN CERAMIC COMPOSITES probably Poissons ratios, part he fiber, for which there are a variety of values in the literature. The elastic moduli of as-deposited Cvi SiC are also subject to speculation. Elastic Effects of Fiber Sliding When a fiber with random roughness of amplitude a and effective wavelength i slides within a atching matrix hole the effect of the elastic mismatch due to roughness will increase from zero to A/R with increasing sliding distance to \/2. Ignoring abrasion, it will remain at A/R for sliding displacements beyond 72. Therefore, the portion of the fiber ahead of the debonding crack, not yet displaced (Region I, Fig. 2), is subject to the thermal residual strains. The portion immediately behind the crack, Region IL, is subject to increasingly compressive radial stress with increasing distance from the crack tip. In region Ill, wherein the fiber is displaced more than 1/2, it is subject to the thermal train plus the full mismatch strain A/R The magnitude of the resulting stresses is dependent upon the geometry and elastic properties of the constituents; of particular interest here are the effects of coating thickness. The total stresses in the matrix at the matrix/coating interface in both Region I (thermal stress only) and Region Ill (the maximum stress) have been calculated as functions of applied C coating thickness for both materials, quite similar in the two materials, i.e., the thinner, reaction layers of C and Sio have relatively little effect, hence only material I is illustrated. The magnitudes of the hoop stresses after sliding(Region )are such that a high incidence of radial cracking(crack surface normal in the 0 direction)would be expected if the fiber were to slide that far. Effects On Debond Length While the behavior in Region Il, d< 1 /2, is greatly complicated by the progressive contribution of the roughness, the problem has recently been solved for at least one type of roughness and uncoated bers( 13). The results imply that for NICALON/C/SiC composites, a reasonable approximation of bond length can be obtained using the simpler treatment with the total misfit stress. This is not generally true for all composites. While this analysis strictly applies to pullout of a single fiber from a much larger(infinite)matrix, it is slightly simpler than more rigorous analyses and the uncertainties in the Poisson's ratios dominate any effects of boundary conditions. Finally, it provides what is for this purpose a conservative estimate by overestimating the debond length. Debond length is then(after R Where R is the fibers radius, u is the friction coefficient, k= E+m)+1-vr) Ei and Vi are moduli and Poissons ratios with f and m denoting fiber and matrix, P*=Sn兀R On is the radial stress at the interface, pr is the residual axial tension in the fiber 马-[
Vol. 32, No. 4 DEBONDING IN CERAMIC COMPOSITES 507 probably Poisson’s ratios, particularly that of the fiber, for which there are a variety of values in the literature. The elastic moduli of as-deposited CVI Sic are also subject to speculation. -Effects When a fiber with random roughness of amplitude A and effective wavelength h slides within a matching matrix hole, the effect of the elastic mismatch due to roughness will increase from zero to A/R with increasing sliding distance to h/2. Ignoring abrasion, it will remain at A/R for sliding displacements beyond h/2. Therefore, the portion of the fiber ahead of the debonding crack, not yet displaced (Region I, Fig. 2), is subject to the thermal residual strains. The portion immediately behind the crack, Region II, is subject to increasingly compressive radial stress with increasing distance from the crack tip. In Region III, wherein the fiber is displaced more than h/2, it is subject to the thermal strain plus the full mismatch strain A/R. The magnitude of the resulting stresses is dependent upon the geometry and elastic properties of the constituents; of particular interest here are the effects of coating thickness. The total stresses in the matrix at the matrix/coating interface in both Region I (thermal stress only) and Region III (the maximum stress) have been calculated as functions of applied C coating thickness for both materials, using NDSANDS. The effects on both radial (Fig. 3) and hoop stresses (Fig. 4) are pronounced and quite similar in the two materials, i.e., the thinner, reaction layers of C and 502 have relatively little effect, hence only material I is illustrated. The magnitudes of the hoop stresses after sliding (Region III) are such that a high incidence of radial cracking (crack surface normal in the tl direction) would be expected if the fiber were to slide that far. Effects On Debond Length While the behavior in Region II, d< h/2, is greatly complicated by the progressive contribution of the roughness, the problem has recently been solved for at least one type of roughness and uncoated fibers (13). The results imply that for NICALON/C/SiC composites, a reasonable approximation of debond length can be obtained using the simpler treatment with the total misfit stress. This is not generally true for all composites. While this analysis strictly applies to pullout of a single fiber from a much larger (infinite) matrix, it is slightly simpler than more rigorous analyses and the uncertainties in the Poisson’s ratios dominate any effects of boundary conditions. Finally, it provides what is for this purpose a conservative estimate by overestimating the debond length. Debond length is then (after ref. 9): l= 2+ ln ( p*-pd -% ) p*-P* Where R is the fibers radius, u is the friction coefficient, k= Emvf W+y,J + F&-q) Ei and Vi are moduli and Poisson’s ratios with f and m denoting fiber and matrix, on is the radial stress at the interface, Pr is the residual axial tension in the fiber, p*= [4&&_!+ f
DEBONDING IN CERAMIC COMPOSITES and Pa is the tension in the fiber at fiber failure The total debond length at fiber failure for various coating thicknesses is given in Fig. 5. while this is a very approximate calculation of debond length, it is in good agreement with the approximate 1 mm inferred by Lamon et al. from testing of microcomposites with 1 um C coatings ( 14). Fig. 5 indicates that the application of compliant coatings can change the debond length from a few tens of fiber diameters to effectively infinity with the addition of coatings y greater than lum thickness due to elastic effects alor expected pull out lengths, using Curtins analysis(15). otted in Fig 5, and range from less than 10 fiber diameters to the length of the specimen. It is Probable that, due to roughness effects, the sole role of the major fraction of the thickness of c and in current composites is elastic, This suggests that a weak a very thin layer of weak material nay be sufficient for debonding purposes if sufficient ance is provided by a thicker, preferably oxidation resistant, coating(Fig. 6) It might even sible to use a sufficiently thin layer of C such that it can be protected through a combination of diffusion path limitation and modest glass formation. Conversely dea, n ing good composite behavior. It will be necessary to design the correct compliance based upon ond fracture surface roughness and constituent elastic properties, as well as deb into any effective coating system Table 1. Constituent Elastic Properties Fiber Matrix E)_200 015 016 References 1. A. G. Evans and D B. Marshall, Acta Metall, V37, No. 10, P. 2567(1989) 2. R.J. Kerans, R. S. Hay, N. J. Pagano and T. A. Parthasarathy, Am. Ceram. Soc. Bull., v68, No 2, P429(1989 3. P. D. Jero, R J Kerans and T.A. Parthasarathy, J. Am. Ceram. Soc. v74[111, p. 2793(1991) 4. R A Lowden, Adv Composite Mater, P 619(1991) 5. J.J. Brennan, Materials Science Res v20, Plenum Press, New York, P 549( 1986) 6. R. Naslain, Composites Interfaces, v1, N 3, Pp. 253(1993) 7. P D Jero, T.A. Parthasarathy and R J Kerans, HT-CMC 1 High Temperature Ceramic Composites, Eds, R. Naslain, J. Lamon, D. Doumenigts, Goodhead Pub. Ltd, Cambridge, P 401 (1993) 8. P D Jero and R.J. Kerans, Scripta Metall.& Mater, V 24, P 2315(1990) 9. R J. Kerans and T. A Parthasarathy, ]. Am. Ceram Soc., v74[7]p 1585( 1991) 10. C. Labrugere, Doctoral Thesis, University of Bordeaux( 1994) 11. N.J. Pagano and G P. Tandon, Compos. Sci. Tech 31, P. 273(1988) 12. AdTech Systems Research, Dayton OH (1993) 13. T.A. Parthasarathy, D B. Marshall and R J Kerans, Acta Metall Mater, in press 14. J. Lamon, N. Lissart, C. Rechiniac, D H. Roach and J M. Jouin, Ceram. Eng. Sci. Proc., V14, [7-8] p.1115(1993) 15. w.A. Curtin, J. Am. Ceram Soc., VZ, (11)2837(1991)
508 DEBGNDING IN CERAMIC COMF’OSlTES Vol. 32, No. 4 and Pa is the tension in the fiber at fiber failure. The total debond length at fiber failure for various coating thicknesses is given in Fig. 5. While this is a very approximate calculation of debond length, it is in good agreement with the approximate 1 mm inferred by Lamon et al. from testing of microcomposites with 1 pm C coatings (14). Fig. 5 indicates that the application of compliant coatings can change the debond length from a few tens of fiber diameters to effectively infinity with the addition of coatings slightly greater than 1~ thickness, due to elastic effects alone. The corresponding expected pull out lengths, using Curtin’s analysis (15), are also plotted in Fig 5, and range from less than 10 fiber diameters to the length of the specimen. It is evidently probable that, due to roughness effects, the sole role ofthe majorfraction of the thickness of C and BN coatings in current composites is elastic. This suggests that a weak interface or a very thin layer of weak material may be sufficient for debonding purposes if sufficient compliance is provided by a thicker, preferably oxidation resistant, coating. (Fig. 6) It might even be possible to use a sufficiently thin layer of C such that it can be protected through a combination of diffusion path limitation and modest glass formation, Conversely, and perhaps of greater importance, this result implies that a weak interface alone is insufficient for obtaining good composite behavior. It will be necessary to design the correct compliance based upon debond fracture surface roughness and constituent elastic properties, as well as debonding properties, into any effective coating system. Table 1. Constituent Elastic Properties Fiber Matrix C SO2 E (GPa) 200 400 9 73 Y 0.15 0.15 0.11 0.16 References 1. A. G. Evans and D. B. Marshall, Acta Metall., ~37, No. 10, p. 2567 (1989). 2. R. J. Kerans, R. S. Hay, N. J. Pagan0 and T. A. Parthasarathy, Am. Ceram. Sot. Bull., ~68, No. 2, p. 429 (1989). 3. P. D. Jero, R. J. Kerans and T. A. Parthasarathy, J . Am. Ceram. Sot. v74 [11], p. 2793 (1991) 4. R. A. Lowden, Adv. Composite Mater., p. 619 (1991) 5. J. J. Brennan, Materials Science Res. ~20, Plenum Press, New York, p. 549 (1986). 6. R. Naslain, Composites Interfaces, vl, N” 3, pp. 253 (1993). 7. P. D. Jero, T. A. Parthasarathy and R. J. Kerans, HT-CMC 1 High Temperature Ceramic Composites, Eds, R. Naslain, J. Lamon, D. Doumenigts, Goodhead Pub. Ltd., Cambridge, p. 401 (1993) 8. P. D. Jero and R. J. Kerans, Scripta Metall. & Mater., v 24, p. 2315 (1990). 9. R. J. Kerans and T. A. Parthasarathy, J. Am. Ceram. Sot., v74 [7] p. 1585 (1991). 10. C. Labrugere, Doctoral Thesis, University of Bordeaux (1994) 11. N. J. Pagan0 and G. P. Tandon, Compos. Sci. Tech 31, p. 273 (1988) 12. AdTech Systems Research, Dayton OH (1993) 13. T. A. Parthasarathy, D. B. Marshall and R. J. Kerans, Acta Metal1 & Mater, in press. 14. J. Lamon, N. Lissart, C. Rechiniac, D. H. Roach and J. M. Jouin, Ceram. Eng. Sci. Proc., V14, [7-S], p. 1115 (1993) 15. W.A. Curtin, J. Am. Ceram. Sot., Va (11) 2837 (1991)
Vol 32, No 4 DEBONDING IN CERAMIC COMPOSITES Fig 1. Typical (idealized) NICALONtm/SiC composite Fig. 2. Progressively increasing elastic mismatch etween fiber and matrix during fiber pullout. 当苏品8 Coating Thickness(micrometer Fig. 3. Interface radial stress vs C coating thickness: Fig. 4. Matrix hoop stress vs. C coating thickness: naterial D) 55000 Fig. 6. Possible altermate approach to interface
Vol. 32,No. 4 DEBONDINGINCERAMICCOMPOSlTES so9 A=24nm -‘- Fig. 1. Typical (idealized) NICALONtm/SiC composite structure. 0 -100 i 2 ____.__.H z -200 r _---- 0.2 0.4 0.6 0.8 1 1.2 C Coating Thickness (micrometers) Fig. 3. Interface radial stress vs. C coating thickness: (material I) . -550 -500 450 4M -350 -3M) -250 -200 L/F Normal Stress (MPa) Fig. 5. Debond and pullout lengths vs. interfacial radial stress and C coating thickness. Fig. 2. Progressively increasing elastic mismatch between fiber and matrix during fiber pullout. 0 02 0.4 06 OS 1 C Coating Thickness (micrometers)) Fig. 4. Matrix hoop stress vs. C coating thickness: (material I) * = 24 nm 7- Fig. 6. Possible alternate approach to interface control