Composites Part A 29A(1998)1145-115 1359-835X/98/S- see front mat c 1998 Elsevier Science Ltd. All rights reserved ELSEVIER PII:S1359835X(9700128-0 The design of the fibre-matrix interfacial zone in ceramic matrix composites Roger R. Naslain Laboratory for Thermostructural Composites, UMR-47(CNRS-SEP-UB1) University of Bordeaux, 3 Allee de La boetie, 33600 Pessac, france atrix interactions or deposited on the fibre surface prior to composite fabrication. It has several key functions, including crack defection, load transfer, diffusion barrier and residual stress relaxation. Four types of interphase are depicted involving weak interfaces, materials with a layered crystal structure(pyrocarbon, BN, micas and phyllosiloxides r materials with the B-alumina/magnetoplumbite structures), multilayers such as(PyC-SiC)n or(BN-SiC), or finally, porous materials. Achieving high mechanical properties and long lifetimes in severe environments require a subtle design of the fibre-matrix interfacial zone, which is depicted for Nicalon/glass-ceramic and Nicalon/SiC matrix composites. C 1998 Elsevier Science Ltd. All rights reserved eywords: A. ceram composites ( CMCs); B interface/interphase; pyrocarbon; boron nitride; multilayers INTRODUCTION FM interfacial zone in CMCs from a processing standpoint The first approach is through the in situ formation of a weak Ceramic matrix composites(CMCs)display, with respect to interphase resulting from some chemical reaction at the FM many other polymer or metal matrix composites, the key interfaces during the high temperature step of composite oroperty of being inverse composites, which is to say that processing. It has been extensively used during the early the strain at failure of the matrix, Em, is much lower than that development of glass-ceramic matrix composites,. The of the fibres, Ef. As a result in such composites the matrix second approach, which offers much more design flexibility undergoes microcracking, Since in CMCs the two inasmuch as it no longer relies on an in situ reaction, is constituents, namely the fibres and the matrix, are brittle, based on the use of precoated fibres, the weak interphase the matrix microcracks should not induce the early failure of being deposited on the fibre surface prior to the deposition he fibres by a notch effect, which means that the matrix of the matrix itself. This second approach was extensively microcracks should be defected at the fibre-matrix(FM) used for non-oxide composites, such as C/SiC and SiC/SiC interfaces. This key function is actually achieved when, composites (the fibre being designated first, as usual) after processing, the FM bonding is weak enough. Further, produced according to the cvi process. A variety of in CMCs the FM interfaces have another key function, i.e interphase materials was suggested, the most studied being that of load transfer as in any fibre-reinforced composite, pyrocarbon, h xagona which supposes conversely a strong enough FM bonding such as micas, B-aluminas and rare-earth phosphates Finally, CMCs are fabricated at relatively high temperatures The aim of the present contribution is, first, to present the (typically, -1000C in the so-called chemical vapour basis of the interphase concept in CMCs(in terms of infiltration (Cvn) process and even higher in the slurry terphase functions and interphase ma aterial nature)and impregnation/hot pressing technique) and they are used in second, to show how such a concept has been used to design severe conditions(high temperatures and oxidizing atmos the FM interfacial zone, focusing on some important pheres), a feature which introduces a constraint of presently used composites thermochemical and thermomechanical compatibilities The design of the FM interfacial zone, or interphase should take into account all these requirements, which are INTERPHASE CONCEPT IN CMCs often contradictory. It becomes an extremely difficult materials science challenge, it the lifetime of the composites Mechanical behaviour of CMCs required in advanced jet engines CMCs are inverse and damageable composites displaying There are two classical approaches to the design of the a non-linear stress-strain behaviour under tensile loading
Fibre-matrix interfacial zone in ceramic matrix composites: R. R. Naslain 400 300 200 microcrack 浦排时 0.6 0.8 .ONGITUDINAL TENSILE STRAIN ( % Figure 1 Tensile curves and crack deflection mechanisms for 2D-Nicalor/Py c/SiC composites corresponding to different fiber-matrix bonding(specimen I is a composite of type A and specimen J a composite of type B), according to Refs 4, 36,94 and a non-brittle failure. 2. However, these features, which crack density is low and the debonding length large, (ii)the are quite uncommon for ceramics, depend strongly on the stress-strain tensile curve displays a typical plateau-like FM bonding. Since in CMCS, R<Ef, the matrix fails first, feature, and(iv) at failure, the matrix microcracks are e it undergoes microcracking at low strain. When the load widely open exposing the fibre surface to the atmosphere, pplied along the fibre direction(1D-composites)or along the composite failing at a moderate stress with extensive one of the fibre directions (nD-composites ), the matrix fibre pullout(Figure 1). By contrast, in composites with microcracks usually propagate perpendicular to that fibre relatively strong bonding, which is the preferred situation direction(mode I). Even when the fibre volume fraction is from a mechanical standpoint and from that of the high(namely, for V,> V, where V( is the critical fibre environment, exactly opposite features are observed volume fraction), the formation and propagation of the first At saturation, the microcrack density is very high but the matrix microcrack(s)may induce the failure of the brittle crack opening and debond length small, rendering more fibres by a notch effect, if the FM bonding is too strong difficult the in-depth diffusion of the atmosphere constitu Therefore, weakening the FM bonding through appropriate ents(particularly, oxygen). Furthermore, since load transfer rocessing is a well-established requirement in CMCs. remains good up to failure, the tensile curve no longer When this requirement is fulfilled, the matrix microcracks exhibits the plateau-like feature, failure occurring at a propagate all over the composite cross-section without higher stress level and with limited fibre pullout. Hence, breaking the fibres, which are thus protected by the weak failure with extensive fibre pullout is not necessary, as often interfaces acting as mechanical fuses, the microcracks being stated as a valuable criterion for appreciating the quality of a deflected in the interfaces(mode il)over a distance, l which CMC. depends on the FM bonding(it can be of the order of 10- From the above discussion, it appears that the FM 20 um for relatively strong or of the order of 100 um or bonding should be neither too strong nor too weak. Hence, more, for very weak FM bonding). Matrix microcracking an important issue is the characteriz of the fm bonding and interface debonding occur within a given strain range up from a quantitative(or at least semiquantitative)standpoint, to a saturation state, the density of cracks, their spacing and which could be used to optimize processing conditions. The opening at saturation depending once again upon the FM FM bonding is usually depicted with two parameters bonding. Beyond this point, the applied load is mainly associated with fracture and slip, respectively. The arried by the fibres alone and the composite stiffness former is considered to involve an interfacial debond reduced as the result of damage energy, T, and the latter is expected to occur with The FM interfaces have also a load transfer function interfacial shear resistance, Ti, which is written as ti= which is fulfilled in an efficient manner when the FM T。μσr, where To constant'term associated with bonding is relatively strong, whereas a too weak FM roughness(most ceramic fibres exhibit some roughness ng yields poor characteristics at failure. Hence, the the nm scale), u is a Coulomb friction coefficient and ar is shape of the stress-strain tensile curve depends both on the the clamping stress normal to the interface For debonding damage extension and FM load transfer In composites with and sliding to occur, rather than brittle matrix crack a too weak FM bonding: (i)matrix microcracking occurs propagation through the fibres, the debond energy r;must within a relatively narrow strain range, (ii)at saturation the not exceed an upper limit, relative to the fibre fracture 1146
Fibre-matrix interfacial zone in ceramic matrix composites: R. R. Naslain bonding strengthened by clamping stress, depending on the sign of the Cte difference Aa=am-af. In modem CMCs, an interphase, i.e. a thin layer of a material with a low shear G strength, is systematically deposited on the fibre surface before composite processing or formed in situ at the FM interfaces, to control the FM bonding. The thickness of the to abou for fibres displaying a diameter of 7-20 um 10.24.25 However, the use of much thinner interphases, e.g. as low Debonding as 4-14 nm, was also reported The two main functions of the interphase are: (i)to act as 1.0 a mechanical fuse, i.e. to deflect the matrix microcracks and (ii)to maintain a good load transfer between the fibres and Figure 2 A debond diagram for CMCs, according to Ref. 2. The elastic the matrix as previously discussed. In addition, the interphase may act as a buffer, absorbing at least partially from materil li of meaterial ateriahere E I and E: are the plain strain the residual stresses at the FMinterfaces resulting from CTE compliant and thick enough. Further, in very reactive Cle tgy r. The upper limit of the r T ratio depends on the systems, such as non-oxide fibres embedded in a silica based stic mismatch a, as shown in Figure 2. Thus, the glass-ceramic matrix, the interphase may also act as a interface will act as mechanical fuse when the following diffusion barrier, which supposes that it is thermodynamic- nequality, T/ s 1/4, is satisfied for composites in which cally compatible with both the fibres and the matrix, on the the elastic mismatch is small (a=O). Since most ceramic one hand, and thick enough, on the other hand. Finally, most fibres have a fracture energy of the order of rr=20 J/m, an CMCs being used at high temperatures and in oxidizing upper limit for Ti is s 5J/m, which is broadly consistent atmospheres, the interphase should be preferably resistant to ith most experimental data although some higher values oxidation. This last requirement is especially important were mentio d-. Different tests were suggested to one remembers that CMCs are often microcracked. the measure the interface parameters, the most commonly used microcrack network facilitating the in-depth diffusion of being:()the push-through test performed with a flattened oxygen towards the interphases and the fibres.Unfortu- diamond tip which is applied under an increasing load to the nately, as will be apparent in the following sections, the best fibre end in a composite thin foil cut perpendicular to the interphase materials in terms of mechanical fuse function fibre axis4-2I and (ii tensile tests with unloading- are non-oxides, e.g. pyrocarbon or hex- BN. Hence, the loading hysteresis loops, performed on ID model effect of the environment on the interphases and the fibres. opposites in the non-linear stress-strain domain 22 both being usually non-oxides, is the major issue in the The former can be used on real composites whereas the design of modern CMCs latter, performed on model composites, yields data which may not be always representative of real composites. All of them require some skill when performed on small diameter Interphase types fibres. further, the treatment of the data relies on models As shown schematically in Figure 3, different kinds of which may not always be adequate. As an example, most t interphase have been suggested and tested in a variety of models assume that the FM interface has no thickness, CMCs, the main objective being to introduce a weak link in whereas in most composites, the interfacial zone is not a strongly bonded FM system. In type I interphases, a simple homogeneous and has a thickness ranging from 0. 1 to 1 um. weak interface, usually between the fibre and the interphase There is thus usually some discrepancy in the interface is introduced in the fm interfacial zone acting as parameter values derived from different tests, on the one mechanical fuse. Examples of such weak interfaces are hand, and these values should be regarded as an estimate, (i)the silica glass/anisotropic pyrocarbon interface which is useful for processing optimization, when assessed according often present in Nicalon/PyC/Sic or Nicalon/BN/SiC to a given experimental procedure and data treatment. on the composites fabricated by CVI from as-received Nicalon Si-C-O fibres'or (ii) the lanthanum phosphate LaPo4/ alumina interface In type II interphases, which are by far the most commonly used, the interphase is a lay Interphase functions material exhibiting a layered crystal structure, the layers CMCs being fabricated at high temperatures, strong being parallel to the fibre surface and weakly bonded to each interactions governed by solid state diffusion, may occur at the Fm interfaces during processing, which usually result in turbostratic pyrocarbon and hex-boron nitride, in the non strong FM bonding, fibre weakening and brittle behaviour. oxides family, as well as phyllosilicates (such as the Further, in specific FM systems displaying significant CTE fluorophlogopite mica, KMg3(AISi3)O1oF2)30.3 the struc- interfaces may become debonded when turally related synthetic phyllosiloxides(such as KMg2Al cooling the material to room temperature, or the FM Si4O12)34 and cleavable hexaluminate, such as hibonite 1147
Fibre-matrix interfacial zone in ceramic matrix composites: R. R. Naslain and(ii)the(PyC-SiC)n multilayer interphases(with typically I<n<10) used in interphases, the interphase is a layer of a porous material Examples of such interphases are porous alumina(or zirconia layers in alumina fibre/alumina matrix composites..49.One simple way to form such porous oxides is first to deposit a carbon/oxide mixture on the fibre surface then to embed the coated fibres in the matrix and finally, to burn out the carbon of the interphase. Other approaches have been proposed to (a)Type I t b) Type Il weaken the FM bonding in CMCs but have not been applie yet to real composites or have not yielded improved M d F mechanical properties or/and lifetimes n Interesting example, Kriven et al have recently suggested the use of shear induced phase transformations occurring with volume shrink age in some materials, e.g. the proto to clino transformation in enstatite MgSiO3(AV/Vo = -5.5%), to induce a weakening effect (which is thus the opposite of that exploited metastable tetragonal zirconia)and hence to promote debond ing in the(enstatite)interphase 0.51 c】 Type In mposites The very first interphases reported to weaken FM bonding and to yield tough CMCs were probably formed in a fortuitous manner. as the result of an in situ chemical reaction at the FM interfaces during processing, for both CaAl2O19, which crystallizes in the magnetoplumbite glass-ceramic matrix(hot pressing) and SiC-matrix(CVi) structure-type related to B-alumina4-37. Ideally, these reinforced with Si-C-O Nicalon fibres*. These fibres are interphases should be strongly bonded to the fibre surface now known to be metastable at high temperatures, under and the layers perfectly parallel to the fibre surface. Under going decomposition beyond about 1100C and reacting such conditions, the weak fibre surface/interphase interface with most silica based glass-ceramic compositions, with,in is no longer the mechanical fuse and the matrix cracks are either case formation of free carbon, which may act as a deflected in a diffuse manner within the interphase itself. mechanical fuse, as discussed in more detail in the next However, as discussed in the next section, such interphases section. This interphase formation approach, though it has are rarely used in an optimized state. In other words, they limited flexibility, is still successfully used in various are often poorly crystallized and oriented (not to say reactive systems. As an example, rare-earth hexa-aluminate amorphous)and too weakly bonded to the fibre surface, with nterphases, approximating the structural formula the result that they behave as a type I interphase or as a type CeI-1Al12-yO19-(magnetoplumbite alumina) and display I/type II hybrid interphase. Finally, the number of materials ng roughly the proper cleavage plane orientation, have been with a layered crystal structure, displaying an easy cleavage formed within dual coatings, by in situ solid state reaction in parallel to the layers and hence which can be used as a low po2 atmosphere between a ceria-doped zirconia interphases in CMCs, is extremely limited. In type Ill coating deposited on a Saphikonf corundum single crystal interphases, the concept used in type II interphases is fibre and an outer alumina coating(simulating an alumina extended to the scale of the nano-or microstructure. These matrix) interphases consist of a stack of layers of different nature Chemical vapour deposition(CVD)or chemical vapour (say,(X- Yn), strongly bonded to the fibre surface but infiltration(CVI) are the most commor d techniques with weak internal interfaces which can be either the X/Y for the deposition of non-oxide interphases such as PyC. interfaces or even atomic planes if one of the layers, say X, hex-BN or related multilayers, whereas their use is more has a layered crystal structure, as for the type II interphase. difficult(but still possible in most cases) for oxides 8.5 With respect to the latter, type Ill interphases can be widely Their success is caused by several important advantages: (i) tailored, the adjustable parameters being the nature of X and simple volatile precursors are available for the interphase Y, the number of X-Y sequences, n, and the thicknesses of materials of interest, i.e. hydrocarbons such as CH4; C 3 h6 or and y layers in the sequence. Another important C3Hg for pyrocarbon, CH] SiCl3 (MTS)H2 for Sic and advantage of these interphases lies in the fact that the BX3-NH3(with X= F, cl, Br) for BN, (ii) they are low terphase functions can now be decoupled. As an example, temperature and low pressure processes with hence no layer X can act as mechanical fuse and layer Y as diffusion significative degradation of the fibres,(iii) they can be barrier. At least two interphases of this type have been xtensively studied (i)the dual BN-SiC (n= I)interphases From Nippon Carbon, Tokyo. used in silica based glass-ceramic matrix composites 8-4>+From Saphikon Inc,Milford(NH),USA 1148
Fibre-matrix interfacial zone in ceramic matrix composites: R. R. Naslain EXAMPLES OF TAILORED INTERFACIAL ZONE MIS+H posites posites with a glass-ceramic matrix rei Nicalon# fibres, namely the Si-C-O ceramic grade fibres (NLM-202)and more recently Hi-Nicalon"fibres 60, hav been developed during the last two decades for applications at medium or high temperatures. The most commonly used H2 matrix compositions belong to the LAS (Li,o-A12 03-SiO2) MAS(MgO-Al2O3-SiO2), CAS(CaO-Al2OSiO2)and BMAS(Bao-MgO-Al203-SiO2) systems and also contain small amounts of various additives. The composites are processed according to a prepreg route comprising slurry Figure 4 Deposition of (Pyc-Sic), multilayer interphases by PCVDI impregnation and hot pressing steps. The design of their FM interfaces has been performed in two steps, first by taking advantage of in situ reactions occurring at the FM interfaces during hot pressing and which result in tough materials and performed with the same equipment as that used for the CVI second, through fibre CVD precoating, in order to of non-oxide matrices, and(iv) they yield well-controlled improve the composite lifetime under stress in oxidizing interphase deposits, in terms of thickness, composition and environments 6 7.41 structure. Furthermore, in one of their last versions, namely Nicalon/glass-ceramic composites are reactive systems pressure pulsed CVD or CVI (P-CVD/CVi)the interphase in the temperature range corresponding to the hot pressing can be deposited layer by layer, with a layer thickness which step typically 1200-1400C. Chemical and structural can be as low as I nm if necessary, giving an extremely analyses at the nm scale have shown that a complex large flexibility to these processes. As an example, if the multilayer FM interfacial zone is formed in situ as the result nature of the gaseous precursor is periodically changed, of an oxidation of the fibre surface(which will be assumed multilayer interphases such as(PyC-SiC)m, are obtained, the to consist of SiC, for simplicity) by oxygen from the thicknesses of the PyC and Sic layers beir matrix. 7.41.60-69. The nature of the interfacial zone, its controlled by the number of hydrocarbon and kinetics of growth and thus its thickness, depend mainly on pulses, respectively(Figure 4) the matrix composition and hot pressing conditions. Thus, Solution/gel or sol/gel processes are particularly appro the FM bonding and hence the mechanical properties of the priate for the deposition of simple, binary or ternary oxide composites, can be tailored. The key point is the presence in hase is the FM interfacial zone, of a thin layer of carbon, often formed on the fibre surface through the repetition of dip strongly textured, which acts as mechanical fuse(with low oating/gelification/drying/firing sequences. Metal alkoxides in and sometimes extremely low r; values, and low to medium water-alcohol solution are often chosen as precursors T, values depending on the state of residual stresses, i. e. CTE inasmuch as different alkoxides can be mixed together at mismatch. 5. 63. 66.67.70-72). The mechanism responsible for the molecular scale with a view to form ultimately complex the formation of the FM interfacial zone is still a matter of oxide interphases, yielding after hydrolysis/polycondensa- controversy. The most commonly accepted is a passive tion, homogeneous singie phase gels. Conversely, the use of oxidation of SiC (from the fibre) by oxygen (or carbon mixtures of sols or mixtures of sols and organometallic monoxide)from the matrix, according to one of the species yields diphasic gels. As an example, the fluorophlo- following overall equations gopite mica interphase, KMg3 (AlSi ,)O oF2 and the related phyllosiloxide interphase KMg 12, were both SiC+O2→C+SiO2 prepared according to an all-alkoxide route. 303233.For SiC+2CO→3C+SiO the latter, the precursor was a mixture of KOCH Ag(OC2H5)2; Al(OC,H9)3 and Si(oc?Hs)4 in 2-methoxy the source of oxygen in eqn (1)being either oxygen dis- ethanol. The oxide equivalents of alkoxide solutions being lved in the slurry glass particles or oxygen generated by low, several sequences are necessary to achieve a gel specific oxides used as additives.73-76 However. more deposit of significant thickness, e.g. 5 for a phyllosiloxide complex mechanisms involving e.g. the active oxidation of Sic were also proposed". The processing parameters gel thickness of I um. Further, the gel-oxide conversion, which can be adjusted are the matrix composition (and par- including solvent vapourization, removal of volatiles and sintering/crystallization of the porous amorphous deposit, ticularly the nature and concentration of the specific oxides occurs with an important shrinkage and needs to be sed as additives )and the hot pressing conditions( tempera- conducted with care in order to avoid crack formation in ture, duration and atmosphere). Based on thermodynamic the deposit. The solution/gel or sol/gel processes could be considerations, it has been shown that oxides which are extended to multilayer interphases, (X-Y)m, where X and Y are now oxide layers of different compositions f Nicalon and Hi-Nicalon fibres from Nippon Carbon, Tokyo 1149