JAm.cerm.Sc,845l136-42(2001) journal Raman Study of Hi-Nicalon-Fiber-Reinforced Celsian Composites IL Residual stress in Fibers Gwenael Gouadec, . f Philippe Colomban, and Narottam P. Bansal**, s Laboratoire Dynamique-Interactions-Reactivite (LADIR), UMR7075-CNRS and Universite Pierre et Marie Curie, Thiais, Val de marne, 94320, France Departement Materiaux et Systemes Composites(DMSC), Office National d'Etudes et de Recherches Aerospatiales(ONERA), Chatillon, Hauts de Seine, 92322, France ohn H. Glenn Research Center at Lewis Field, National Aeronautics and Space Administrat Cleveland, Ohio 4 Band shifts in Raman spectra were used to assess, on a I. Background microscopic scale, the residual strain existing in Hi-Nicalon silicon carbide fibers reinforcing celsian-matrix composites Raman-based stress measurements rely on the anharmonic Uncoated as well as p-BN/SiC-coated and p-B(SiN/SiC-coated nature of the chemical bonds, which make normal vibrations Hi-Nicalon fibers were used as the reinforcements. We unam- sensitive to any external disturbance of the potential well, say for biguously conclude that the fibers were in a state of compres nstance a pressure or a temperature change. The shift induced in sive residual stress. Quantitative determination of the residual mode i wavenumber i(em)when a strain E" is applied to a stress was made possible by taking into account the heating induced by laser probing and by using a reference line, of fixed wavenumber. We found fiber compressive residual stress v,=Do+SE values between 110 and 960 MPa, depending on the fiber/ nomenon was observed at the surface of p-BN/SiC-coated absolute"strain. io is a matrix coating in the composite. A stress relaxation-like phe expressed in units of cm 品 coated Hi-Nicalon fibers did not show any stress relaxation in anharmonicity., 4 If the inve the celsian-matrix composites. deformation theory, then v:+S:. . Introduction and s are in units of GPa and cm /GP C ERAMIC-MATRIX composites(CMCs)are light-weight refractory aterials which are of potential interest for high-temperature respectivel nde the fiber's Young modulus in GPa. S: and Si structural components in various aerospace and industrial applica- are negative in almost all cases but depend on the mode according tions In Part I, the phases present in celsian-matrix composites to the structure anisotropy. Many of the available values were obtained through diamond anvil cell experiments on crystals. '- reinforced with (desized) uncoated or p-BN/SiC-coated Hi- Nicalon fibers were identified and characterized by Raman mi- They cannot be applied to multiphase materials like Sic fiber rospectroscopy, from a chemical and structural point of view. In because the phases are amorphous/nanocrystalline. Wavenumber hift calibrations are then mandatory and usually obtained under this second part, we further interpret the fibers spectra in an axial stress, the control parameter being the strain attempt to assess the residual strain resulting from differences in Equation(2) supposes Youngs modulus e to be the same in the coefficients of thermal expansion( CTE)(a) between the fiber compression and in tension, which we shall discuss further, on and the matrix. Modeling this stress mathematically would be account of the bond nature and 3D symmetry in silicon carbide difficult, especially in the case of coated fibers. Indeed, interpha structures. Calibrations of S: were done for carbon Raman contri- materials promote stress relaxation(due to higher compliance, butions in polymeric, carbon, and Sic fibers. --In the latter case cracking, or a mismatch). Besides, they are usually partly crystal- the results could be compared with S: of silicon carbide optical line, often metastable, materials and their expansion is difficult to modes measure. The objective of this paper is to assess stresses in Theoretical predictions of strain-induced"Raman shifts'"exist Hi-Nicalon fibers embedded in celsian-matrix composites using for crystals. >, 4 They are based on the elastic constants tensor, and their use for the determination of biaxial stresses in Cvd on films produced consistent results. The Gruneisen coefficient is sometimes preferred to S: and S:to express the wavenumber sensitivity to external stresses. It is defined as a function of the volume(n)as D. R. Clarke--contributing editor Received June 19, 2000: approved December 4, 20( If only the pressure component changing the unit cell param nteractions-Reactivite )ffice National d Etudes et de Recherches Aerospatiale ters is considered( the so-called hydrostatic component, the National Aeronautics and Space Administration other one changes bond angles), then S; and y, are linked by
Raman Study of Hi-Nicalon-Fiber-Reinforced Celsian Composites: II, Residual Stress in Fibers Gwe´nae¨l Gouadec,†,‡ Philippe Colomban,† and Narottam P. Bansal**,§ Laboratoire Dynamique-Interactions-Re´activite´ (LADIR), UMR7075-CNRS and Universite´ Pierre et Marie Curie, Thiais, Val de Marne, 94320, France De´partement Mate´riaux et Syste`mes Composites (DMSC), Office National d’Etudes et de Recherches Ae´rospatiales (ONERA), Chatillon, Hauts de Seine, 92322, France John H. Glenn Research Center at Lewis Field, National Aeronautics and Space Administration, Cleveland, Ohio 44135 Band shifts in Raman spectra were used to assess, on a microscopic scale, the residual strain existing in Hi-Nicalon silicon carbide fibers reinforcing celsian-matrix composites. Uncoated as well as p-BN/SiC-coated and p-B(Si)N/SiC-coated Hi-Nicalon fibers were used as the reinforcements. We unambiguously conclude that the fibers were in a state of compressive residual stress. Quantitative determination of the residual stress was made possible by taking into account the heating induced by laser probing and by using a reference line, of fixed wavenumber. We found fiber compressive residual stress values between 110 and 960 MPa, depending on the fiber/ matrix coating in the composite. A stress relaxation-like phenomenon was observed at the surface of p-BN/SiC-coated Hi-Nicalon fibers whereas the uncoated or p-B(Si)N/SiCcoated Hi-Nicalon fibers did not show any stress relaxation in the celsian-matrix composites. I. Introduction CERAMIC-MATRIX composites (CMCs) are light-weight refractory materials which are of potential interest for high-temperature structural components in various aerospace and industrial applications. In Part I,1 the phases present in celsian-matrix composites reinforced with (desized) uncoated or p-BN/SiC-coated HiNicalon fibers were identified and characterized by Raman microspectroscopy, from a chemical and structural point of view. In this second part, we further interpret the fibers’ spectra in an attempt to assess the residual strain resulting from differences in the coefficients of thermal expansion (CTE) (a) between the fiber and the matrix. Modeling this stress mathematically would be difficult, especially in the case of coated fibers. Indeed, interphase materials promote stress relaxation (due to higher compliance, cracking, or a mismatch). Besides, they are usually partly crystalline, often metastable, materials and their expansion is difficult to measure. The objective of this paper is to assess stresses in Hi-Nicalon fibers embedded in celsian-matrix composites using Raman microspectroscopy. II. Background Raman-based stress measurements rely on the anharmonic nature of the chemical bonds, which make normal vibrations sensitive to any external disturbance of the potential well, say for instance a pressure or a temperature change.2 The shift induced in mode i wavenumber n# (cm21 ) when a strain e% is applied to a material is calibrated as follows:3 n#i 5 n#i 0 1 Si ε zε% (1) The superscript “%” is a reminder that we will not use the “absolute” strain. n#i 0 is the “stress-free wavenumber” and Si ε expressed in units of cm21 /%, is a direct measure of bond anharmonicity.3,4 If the investigated material obeys the “elastic deformation theory,” then n#i 5 n#i 0 1 Si ε S s ED 5 n#i 0 1 Si s zs (2) where the stress s and Si s are in units of GPa and cm21 /GPa, respectively, and E is the fiber’s Young modulus in GPa. Si ε and Si s are negative in almost all cases but depend on the mode according to the structure anisotropy. Many of the available values were obtained through diamond anvil cell experiments on crystals.5–7 They cannot be applied to multiphase materials like SiC fibers, because the phases are amorphous/nanocrystalline. Wavenumber shift calibrations are then mandatory and usually obtained under axial stress, the control parameter being the strain. Equation (2) supposes Young’s modulus E to be the same in compression and in tension, which we shall discuss further, on account of the bond nature and 3D symmetry in silicon carbide structures. Calibrations of Si ε were done for carbon Raman contributions in polymeric, carbon, and SiC fibers.8–12 In the latter case, the results could be compared with Si ε of silicon carbide optical modes.12 Theoretical predictions of strain-induced “Raman shifts” exist for crystals.13,14 They are based on the elastic constants tensor, and their use for the determination of biaxial stresses in CVDdiamond15 or silicon16 films produced consistent results. The Gru¨neisen coefficient is sometimes preferred to Si s and Si ε to express the wavenumber sensitivity to external stresses. It is defined as a function of the volume (V) as gi 5 2 ] log n#i ] log V (3) If only the pressure component changing the unit cell parameters is considered (the so-called hydrostatic component, the other one changes bond angles17), then Si s and gi are linked by D. R. Clarke—contributing editor Manuscript No. 188455. Received June 19, 2000; approved December 4, 2000. **Fellow, American Ceramic Society. † Laboratoire Dynamique-Interactions-Re´activite´. ‡ Office National d’Etudes et de Recherches Ae´rospatiales. § National Aeronautics and Space Administration. J. Am. Ceram. Soc., 84 [5] 1136–42 (2001) 1136 journal
May 2001 Raman Study of Hi-Nicalon-Fiber-Reinforced Celsian Composites 1137 a proportionality factor: -vdP/dr is the materials bulk modulus. In the G ase, equations of state linking v(or the cell parameters) oulder plied stress are mandatory to compare S: with Y,. In cubi Y, can also be obtained from the wavenumber shifts measured for unidirectional stresses along high symmetry The Hi-Nicalon fiber consists of several phases, which are all somewhat amorphous. It would be unrealistic to theoreticall predict strain-induced Raman shifts to them or to use reported Gruneisen coefficients. We will therefore use s: coefficients owever, the rather high measurement uncertainty necessitates the Experimental Procedure The samples and the experimental equipment have been fully described in Part I. The spectra were recorded on an"XY spectrograph(Dilor, France) with a back-illuminated nitrogen cooled CCD detector. A motorized X-Y displacement table was used for a 2D mapping of any given surface and the 556.3 nm line f a neon lamp that was installed in the spectrograph chamber was a wavenumber reference when working with the 514.5 nm laser line(the corresponding shift was 1460.4 cm, peaking in the middle of the carbon signal) Our interest focused on three unidirectional celsian-matrix composites(12 plies), prepared by hot pressing. Composite I was reinforced by desized Hi-Nicalon fibers (fiber volume fraction Vr 35%), while ce te 2 incorporated double-coated Hi Nicalon fibers(a layer of 0. 4 um pyrolytic bn(p-BN) over (b) oated by a 0. 2-0.3 um thick SiC diffusion barrier, V= 28% Composite 3 will refer to a composite similar to composite 2, but with a 12 wt% silicon doping of the p-BN layer(0. 4 um thick Wavenumber and an overcoating of sic (-0 2-0.3 um thick). Sections of each omposite were polished perpendicular and parallel to the fiber ectral deconvolutions: (a) carbon spectra b)SiC spectra;(O)optic mode, (T)transverse mode, (L) longitudinal direction. Some "reference" fibers were also extracted from mode) composite 2 by matrix crushing in an agate mortar. Such extraction was impossible in composite I because of fiber strength degrada tion from mechanical damage during composite processing ected locations and ept for the 2) Choice of the"Stress Probe"and Fiting Procedure “sp3” carbon mode( D band) and the optic SiC, which are given pure Lorentzian sh Because of the high electronic absorption of C-C bonds in the parameters are the wavenumber(v), the full half-height harmonics (pure or combined) are enhanced by a so-called (), the intensity, and the band area. resonance” phenomenon.lHi- Nicalon“ "silicon carbide” fibers ave a rather large excess of carbon(C: Si stoichiometric ratio of (3) Power and Wavelength-Induced Measurement 1. 4, that is, 40% excess carbon); thus the signal C-C will be the Disturbances nost convenientstress probe. Only in the nearly stoichiometric Thermal expansion has the same lengthening effect on bonds as fibers might the Sic spectrum be used for stress assessment. Ou tensile stress. Any localized heating induced by the laser impact sults will be based on the so-called"D band whose attribution might therefore lead one to overestimate tensile stresses an bonds has been discussed in Part I In the past, a underestimate compressive ones. Under unfavorable conditions "G band"("sp--like C-C bonds)has been preferred in fibers compression might even be confused with tensile stress. There is whose D band was either weak or less defined. Yet the carbon thus a need, before stress analysis, for a preliminary study intended atoms contributing to the d band are carbon moiety surface atoms to assess the influence of working parameters on sample heating. and should be better incorporated into the Sic network than those The first parameter is the material itself (M), which includes the corresponding to G, in the bulk. Besides, the fine structure of G fiber composition and its environment; fibers are either free- reveals a doublet in the new generations of SiC fibers, one standing in air, for S; calibrations, or embedded in a given matrix component of which, D, has wavelength-dependent intensity an for in situ measurements. A large influence of fiber surroundings position. 0 This component is only a shoulder in highly amorphous on thermal dissipation is anticipated. The matrix should act as a arbon, but is truly pronounced in SiC fibers. Besides, the double huge heat sink and dissipate most of the accumulated heat. Other fitting depends to some extent on a smaller(but much wider) ban parameters that might have an effect on heating are the wavelengt around 1530 cm(probably carbon linked to oxygen atoms"). A of the laser(because of carbon resonance), the laser power Figure I illustrates typical decompositions performed using the (P), the recording time(n), and the surface area impacted by the Labspec software(Dilor, France). The first step is the systematic spot(A) subtraction of a linear base line attached to the spectral window (i Wavenumbers are highly sensitive to the power. Eight tests limits. all known contributions are then entered close to the on the carbon d band showed a linear mode softening with
a proportionality factor: Si s 5 n#i 0 z gi B (4) where B 5 2V(dP/dV)T is the material’s bulk modulus. In the general case, equations of state linking V (or the cell parameters) to the applied stress are mandatory to compare Si s with gi . In cubic materials (V 5 a3 ), gi can also be obtained from the wavenumber shifts measured for unidirectional stresses along high symmetry directions.18 The Hi-Nicalon fiber consists of several phases, which are all somewhat amorphous. It would be unrealistic to theoretically predict strain-induced Raman shifts to them or to use reported Gru¨neisen coefficients. We will therefore use Si ε coefficients; however, the rather high measurement uncertainty necessitates the determination of optimum working conditions. III. Experimental Procedure (1) Samples and Equipment The samples and the experimental equipment have been fully described in Part I.1 The spectra were recorded on an “XY” spectrograph (Dilor, France) with a back-illuminated nitrogencooled CCD detector. A motorized X-Y displacement table was used for a 2D mapping of any given surface and the 556.3 nm line of a neon lamp that was installed in the spectrograph chamber was a wavenumber reference when working with the 514.5 nm laser line (the corresponding shift was 1460.4 cm21 , peaking in the middle of the carbon signal). Our interest focused on three unidirectional celsian-matrix composites (12 plies), prepared by hot pressing.1 Composite 1 was reinforced by desized Hi-Nicalon fibers (fiber volume fraction Vf 5 35%), while composite 2 incorporated double-coated HiNicalon fibers (a layer of ;0.4 mm pyrolytic BN (p-BN) overcoated by a ;0.2–0.3 mm thick SiC diffusion barrier; Vf 5 28%). Composite 3 will refer to a composite similar to composite 2, but with a 12 wt% silicon doping of the p-BN layer (;0.4 mm thick) and an overcoating of SiC (;0.2–0.3 mm thick). Sections of each composite were polished perpendicular and parallel to the fiber direction. Some “reference” fibers were also extracted from composite 2 by matrix crushing in an agate mortar. Such extraction was impossible in composite 1 because of fiber strength degradation from mechanical damage during composite processing.19 (2) Choice of the “Stress Probe” and Fitting Procedure Because of the high electronic absorption of C–C bonds in the visible–UV range, the symmetric stretching modes and their harmonics (pure or combined) are enhanced by a so-called “resonance” phenomenon.1 Hi-Nicalon “silicon carbide” fibers have a rather large excess of carbon (C:Si stoichiometric ratio of 1.4, that is, 40% excess carbon); thus the signal C–C will be the most convenient “stress probe.” Only in the nearly stoichiometric fibers might the SiC spectrum be used for stress assessment. Our results will be based on the so-called “D band” whose attribution to “Csp3–Csp2” bonds has been discussed in Part I. In the past, a “G band” (“sp2 ”-like C–C bonds) has been preferred in fibers whose D band was either weak or less defined. Yet the carbon atoms contributing to the D band are carbon moiety surface atoms and should be better incorporated into the SiC network than those corresponding to G, in the bulk. Besides, the fine structure of G reveals a doublet in the new generations of SiC fibers, one component of which, D9, has wavelength-dependent intensity and position.20 This component is only a shoulder in highly amorphous carbon, but is truly pronounced in SiC fibers. Besides, the doublet fitting depends to some extent on a smaller (but much wider) band, around 1530 cm21 (probably carbon linked to oxygen atoms3 ). Figure 1 illustrates typical decompositions performed using the Labspec software (Dilor, France). The first step is the systematic subtraction of a linear base line attached to the spectral window limits. All known contributions are then entered close to their expected locations and are given a Gaussian shape, except for the “sp3 ” carbon mode (D band) and the optical (TO/LO) modes of SiC, which are given pure Lorentzian shapes. The adjusted parameters are the wavenumber (n), the full width at half-height (w), the intensity, and the band area. (3) Power and Wavelength-Induced Measurement Disturbances Thermal expansion has the same lengthening effect on bonds as tensile stress. Any localized heating induced by the laser impact might therefore lead one to overestimate tensile stresses and underestimate compressive ones. Under unfavorable conditions, compression might even be confused with tensile stress. There is thus a need, before stress analysis, for a preliminary study intended to assess the influence of working parameters on sample heating. The first parameter is the material itself (M), which includes the fiber composition and its environment; fibers are either freestanding in air, for Si ε calibrations, or embedded in a given matrix for in situ measurements. A large influence of fiber surroundings on thermal dissipation is anticipated. The matrix should act as a huge heat sink and dissipate most of the accumulated heat. Other parameters that might have an effect on heating are the wavelength l of the laser (because of carbon resonance), the laser power (P), the recording time (t), and the surface area impacted by the spot (A). (i) Wavenumbers are highly sensitive to the power. Eight tests on the carbon D band showed a linear mode softening with Fig. 1. Typical examples of spectral deconvolutions: (a) carbon spectra, (b) SiC spectra; (O) optic mode, (T) transverse mode, (L) longitudinal mode). May 2001 Raman Study of Hi-Nicalon-Fiber-Reinforced Celsian Composites 1137
1138 Journal of the American Ceramic Society--Gouadec et al. Vol 84. No 5 increasing powers, for a variety of t A values. This will be Table L. Ordinate at Origin vi and Slope s: of Linear aracterized by negative S; parameters, in units of cm/mw Regressions calculated after the fitted wavenumbers of Modes i, Plotted as a Function of the Laser Power P(in T= T+ S-P mw, Measured on the Sample) Mode i/fiber/matrix/(nm)t Fo(cm-) Examples of S: measurement for different test conditions(fiber Si(cm/mw) setting, wavelength, ... )are shown in Fig. 2 and Table I Only for Carbon D band/Hi-S 1326.11 SiC (TO mode) was a"plateau clearly identified, up to 17 mW on//647 SiC bonds being nonresonant in the visible range there is there is"D"Hi-Nicalon//514.5 1353.96 therefore no direct conversion of the photonic energy into heat. "D/Hi-Nicalon/ /514.5 1357 Heating is postponed, and measurement confidence would proba- 1354.18 bly be better using SiC as the stress probe when new generations D/SCS6/Ti alloy/514.5 "D"/Hi-Nicalon/celsian/514.5 of nearly stoichiometric SiC fibers are studied (instead of carbon"D"/Hi-Nicalon/celsian/457.9 phase, which becomes so dispersed it no longer probes the actual "D"/Hi-Nicalon/celsian/6471I stress). As for carbon, regressions were extrapolated down to P=Omw although it looks as though a very short "plateau"exists Sic to modet/SCS6 791.49 below 0.5 mw(we cannot rule out heating, but noise contribution (SiC)Ti alloy/514.5 tops being negligible at such low powers and might be disturbing mahm①mm Textron, U.S.A. )is a 33 D One expected result that Table I confirms is the poor fiber edge. #so. 0.07 cm-I/mw for dissipation ability of free-standing fibers, due to the low thermal conductivity of air. In contrast, the matrix surrounding a fiber in a omposite acts as a heat sink, the dissipating effect being greater for a metallic matrix(Ti6242 alloy) than for ceramics(celsian). observed with the A= 514.5 nm line(Table 2 caption wrong- (iii) The effect, if there is one, on changing A on the surface fully mentioned air annealing ). Note that changing the power from of free-standing fibers from I um"(spot size)to 100 um(spot with the 9.2 cm-I shift predicted based on Table 1(6X 1.54).In scanning by a mobile mirror) when (A, P, 1) are set at given values seems to be negligible addition, Sp measurement showed a better reproducibility for v) If there is any"time effect fibers analyzed under equivalent conditions than SG. Figure 3(a) equilibrium state is reached very quickly since wavenumbers did shows that for a pure carbon fiber(FT700 grade ), both vp and Sp change linearly with the exciting laser line energy. A similar linea not systematically decrease(at least for power below 10 mw) behavior is found for n in Hi-Nicalon fibers(Fig. 3(b)and can Power clearly is the most important parameter and data in Table consequently be expected also for their Se when I increased for a given set of (A, M, A) I will help us compensate the thermally induced Raman shift as a (B) In Situ Results, Comparison with Stress-Free References function of the sample and the waveler Once S: is known, it becomes possible to interpret wavenumber shifts from one place to another in terms of stress difference. The actual loading at each point requires a vistress-free"reference measured at room temperature. vi might unfortunately differ from Result ne sample to another and a value found in the sample itself is (I Fiber Analysis expected to be more reliable. Besides, measuring the stress in the (A SD Calibration under Axial Tension: A-2.7(+0.4) ame sample as the reference would avoid compensating for the cm / value has been obtained by gouadec et al. for Sp in annealed Hi-Nicalon fibers (1000.C in reducing atmosphere) Sic mw 5 mw 17 mw 15--920s8 10 786 1320 11 802002202402602.80 1340 776 ▲Free- standing H 13310occs6m6242A 1325 1340 1330 o Hi-Nicalon 20 Fig. 2. Wavenumber vs laser power(A 514.5 nm) for the D band 802002202.40260280 free-standing in air or embedded in composite 1; left-hand scale)and for SiC TO, mode in an SCS-6 fiber(right-hand scale). The labels indicate the Fig 3.(a)vo (left-hand scale, in cm)and SD(right-hand scale, in umber of measured values( when more than one)for the corresponding ser line energy (in electronvolts) points. The displayed regression lines would be straight lines for a linear for FT700 carbon fibers. (b) Comparison of vo dependency to the lin ower scale energy in Hi-Nicalon and FT700 fibers
increasing powers, for a variety of {t;A} values. This will be characterized by negative Si P parameters, in units of cm21 /mW: n#i 5 n#i 0 1 Si P zP (5) Examples of Si P measurement for different test conditions (fiber, setting, wavelength, . . . ) are shown in Fig. 2 and Table I. Only for SiC (TO mode) was a “plateau” clearly identified, up to 17 mW. Si–C bonds being nonresonant in the visible range, there is therefore no direct conversion of the photonic energy into heat. Heating is postponed, and measurement confidence would probably be better using SiC as the stress probe when new generations of nearly stoichiometric SiC fibers are studied (instead of carbon phase, which becomes so dispersed it no longer probes the actual stress12). As for carbon, regressions were extrapolated down to P 5 0 mW although it looks as though a very short “plateau” exists below 0.5 mW (we cannot rule out heating, but noise contribution stops being negligible at such low powers and might be disturbing the fitting). (ii) One expected result that Table I confirms is the poor dissipation ability of free-standing fibers, due to the low thermal conductivity of air. In contrast, the matrix surrounding a fiber in a composite acts as a heat sink, the dissipating effect being greater for a metallic matrix (Ti6242 alloy) than for ceramics (celsian). (iii) The effect, if there is one, on changing A on the surface of free-standing fibers from 1 mm2 (spot size) to 100 mm2 (spot scanning by a mobile mirror) when {l,P,t} are set at given values seems to be negligible. (iv) If there is any “time effect” on heat accumulation, the equilibrium state is reached very quickly since wavenumbers did not systematically decrease (at least for power below 10 mW) when t increased for a given set of {l,M,A}. Power clearly is the most important parameter and data in Table I will help us compensate the thermally induced Raman shift as a function of the sample and the wavelength used. IV. Results (1) Fiber Analysis (A) SD ε Calibration under Axial Tension: A 22.7 (60.4) cm21 /% value has been obtained by Gouadec et al. for SD ε in annealed Hi-Nicalon fibers (1000°C in reducing atmosphere) observed with the l 5 514.5 nm line11 (Table 2 caption wrongfully mentioned air annealing). Note that changing the power from 2 to 8 mW provoked a 9.1 cm21 shift of n#D, which agrees well with the 9.2 cm21 shift predicted based on Table I (6 3 1.54). In addition, SD ε measurement showed a better reproducibility for fibers analyzed under equivalent conditions than SG ε . Figure 3(a) shows that for a pure carbon fiber (FT700 grade), both n#D and SD ε change linearly with the exciting laser line energy. A similar linear behavior is found for n#D in Hi-Nicalon fibers (Fig. 3(b)) and can consequently be expected also for their SD ε . (B) In Situ Results; Comparison with Stress-Free References: Once Si ε is known, it becomes possible to interpret wavenumber shifts from one place to another in terms of stress difference. The actual loading at each point requires a n#i 0 “stress-free” reference measured at room temperature. n#i 0 might unfortunately differ from one sample to another and a value found in the sample itself is expected to be more reliable. Besides, measuring the stress in the same sample as the reference would avoid compensating for the Fig. 2. Wavenumber vs laser power (l 5 514.5 nm) for the D band (“sp3 –sp2 /sp3 ”-hybridized carbon atoms) of a Hi-Nicalon fiber (either free-standing in air or embedded in composite 1; left-hand scale) and for SiC TO1 mode in an SCS-6 fiber (right-hand scale). The labels indicate the number of measured values (when more than one) for the corresponding points. The displayed regression lines would be straight lines for a linear power scale. Table I. Ordinate at Origin n#i 0 and Slope Si P of Linear Regressions Calculated after the Fitted Wavenumbers of Modes i, Plotted as a Function of the Laser Power P (in mW, Measured on the Sample) Mode i/fiber/matrix/l (nm)† n#i 0 (cm21 ) Si P (cm21 /mW) Carbon D band/Hi-S Nicalon/‡ /647.1 1326.11 21.30 “D”/Hi-Nicalon/‡ /514.5 1353.96 21.54 “D”/Hi-Nicalon§ / ‡ /514.5 1357.76 21.60 “D”/FT700/2/514.5 1354.18 20.20 “D”¶ /SCS6/Ti alloy/514.5 1354.26 20.47 “D”/Hi-Nicalon/celsian/514.5 1355.06 20.55 “D”/Hi-Nicalon/celsian/457.9 1368.90 20.70 “D”/Hi-Nicalon/celsian/647.1 1328.88 20.57 SiC TO mode††/SCS6 (SiC)/Ti alloy/514.5 791.49 0‡‡ † FT700 is a pure carbon fiber (Tonen, Japan); SCS6 fiber (Textron, U.S.A.) is a 33 mm diameter carbon filament surrounded by a 50 mm thick SiC layer. ‡ Free-standing fiber. § Scanning over 100 mm. ¶ In the center of carbon filament. ††12.5 mm from the fiber edge. ‡‡STO-SiC P 5 20.07 cm21 /mW for P . 17 mW. Fig. 3. (a) n#D (left-hand scale, in cm21 ) and SD ε (right-hand scale, in cm21 /%) as a function of the exciting laser line energy (in electronvolts) for FT700 carbon fibers. (b) Comparison of n#D dependency to the line energy in Hi-Nicalon and FT700 fibers. 1138 Journal of the American Ceramic Society—Gouadec et al. Vol. 84, No. 5
May 2001 Raman Study of Hi-Nicalon-Fiber-Reinforced Celsian Composites 1139 laser heating"issue. The two spectra of the longitudinal mapping 2) Interphase Materials of Fig. 4(458 nm) that were recorded on the fiber crack in composite 2 are almost stress-free internal references. With 5 mw view in Part 1. 'We found that the 6H-polytype dominate exciting power, their D band mean wavenumber was 135880 SiC layer of composite 2. The wavenumbers will be (+1.00)cm-I, the mean for all other probed points being 1365.75 below in terms of possible strains, which will not be poss\ (+0. 15)cm. There is thus strong evidence that the fiber is in BN, because of a disturbing signal scattering phenomeno ompression in the matrix. The fitting of spectra that were recorded under the very same conditions on cross sections gave an verage 1366.20(+0.30)cm wavenumber(seven measurement Discussion points). Some matrix being removed on polished cross sections, one might expect greater stress effects on spectra recorded through ( Fiber Stress in the Composites the matrix. Such recording was not possible, because of celsian's The above-mentioned apparent compression is in perfect agree low transparency(20 um). But, Wu and Colomban found ment with the CTE of the Hi-Nicalon fiber (a=3.5 x 10-6roC equivalent results for fiber-reinforced mullite probed on cross between room temperature(RT) and 500C, from Nippon carbon sections and through the matrix. 2I This peculiar behavior must data sheets) and the celsian matrix(a=5.28 X 10/C between result from CMCs having reloading lengths of a few micrometers RT and 1200C). For CMCs, the reinforcement CTE should be nly, much smaller than the typical 500-1000 um values encoun- greater, to the matrix, whose tensile strength is loy tered in organic matrix composites. 22 residual compression and, in this way, prevent microcracking Yet, low a matrixes usually have loosely packed frameworks and compared with the experimental accuracy and we decided to the best matrix choice requires a chemical and mechanical analysis record additional series of spectra with the 514.5 nm wavelength, of the multiphase materials. Note that a of the monoclinic celsian for the neon line to be systematically included as a reference ase of srAlSio. was measured to be 2.5x 10o/oc- which The power was fixed at I mW on the sample, the lowest possible confirms that thermal expansion of celsian is a function of the value still giving acceptable spectra quality for reasonable record- alkali/alkaline-earth content. 26 (A) Anticipated Stress: The axial (o) and transverse(o 900 s recording did not produce different values, neither for residual stresses in an"infinite"fiber, that is to say, a long and 2四购以mF5 Spectra were recorded lated using the following expressions: 27 wavenumbers nor for bandwidths fiber sections or sections of extracted fibers the latter constitute an internal reference. Different regions were tested in composites I and 2 for statistical analysis. Plotted values are averages with error bars indicating the extremes. Figure 6 gives an example of a E7/21 1+v two-dimensional mapping performed on composite I under the conditions used for Fig. 5. Again, a correction of the apparatus Cr-am+ va(oa-amjAT The subscripts in Eqs. (6)and (7) are relative to the matrix(m)and the axial(A)or transverse(T)properties of the fiber v are Poisson ratios and AT represents the difference between room temperature and the so-called"induction temperature, Ti, at which the matrix becomes too "soft" to constrain the fiber 28 Because of the lack of availability of experimental values of Ti, we will replace it by Is= 1300C, a typical sintering temperature for aluminosilicate ma- tries. Besides, given the relative isotropy of Hi-Nicalon fibers and, again, because of the lack of data, axial and transverse properties should be considered the same(with a subscript f). If w 10 now put all Poisson ratios to zero, which is partly justified as a first approximation by the very local probing scale of Raman spectros- Length X (um) copy, Eqs. (6)and(7)simplify to GA=(am-ax△FEr (8) On crack Taking Em =96 GPa and the Er= 270 GPa Youngs modulus given by Nippon Carbon Co for Hi-Nicalon fibers, the compres sive stress should be about 600-650 MPa, axially, and greater than 150 MPa, radially. The stress measurement derived from(unpo- Far from crack larized) Raman spectra should fall in between since the method is not sensitive to the loading direction, at least in amorphous materials 800100012001400160018002000 L(B) Experimental Stress Assessment: Looking at Fig. 5, the venumbers of the Hi-Nicalon fibers before matrix embedding Wavenumber. cm-1 are rather close: uncoated and coated fibers are suitable to serve as posite 2 polished parallel to stress-free references. Their bandwidths are slightly different, direction. Spectra were recorded with the 458 nm line (5 mw, 60 but Fig. 7 in Part I showed that structural evolution of Hi-Nicalon int,(b) Examples of spectra recorded along the fiber and fibers begins around 1300-1400%C, which happens to be the temperature range of p-BN/SiC chemical vapor deposition
“laser heating” issue. The two spectra of the longitudinal mapping of Fig. 4 (458 nm) that were recorded on the fiber crack in composite 2 are almost stress-free internal references. With 5 mW exciting power, their D band mean wavenumber was 1358.80 (61.00) cm21 , the mean for all other probed points being 1365.75 (60.15) cm21 . There is thus strong evidence that the fiber is in compression in the matrix. The fitting of spectra that were recorded under the very same conditions on cross sections gave an average 1366.20 (60.30) cm21 wavenumber (seven measurement points). Some matrix being removed on polished cross sections, one might expect greater stress effects on spectra recorded through the matrix. Such recording was not possible, because of celsian’s low transparency (#20 mm). But, Wu and Colomban found equivalent results for fiber-reinforced mullite probed on cross sections and through the matrix.21 This peculiar behavior must result from CMCs having reloading lengths of a few micrometers only, much smaller than the typical 500–1000 mm values encountered in organic matrix composites.22 All shifts that we measured on composites 1 and 2 remained low compared with the experimental accuracy and we decided to record additional series of spectra with the 514.5 nm wavelength, for the neon line to be systematically included as a reference.11 The power was fixed at 1 mW on the sample, the lowest possible value still giving acceptable spectra quality for reasonable recording times. We retained 180 s per spectrum, but we checked that a 900 s recording did not produce different values, neither for wavenumbers nor for bandwidths. The in situ results are presented in Fig. 5. Spectra were recorded on cross-sectional views (on polished composites), freshly cut fiber sections or sections of extracted fibers; the latter constitute an internal reference. Different regions were tested in composites 1 and 2 for statistical analysis. Plotted values are averages with error bars indicating the extremes. Figure 6 gives an example of a two-dimensional mapping performed on composite 1 under the conditions used for Fig. 5. Again, a correction of the apparatus shift was done using the neon line as a reference. (2) Interphase Materials SiC and Si spectra were presented from a qualitative point of view in Part I.1 We found that the 6H-polytype dominates in the SiC layer of composite 2. The wavenumbers will be discussed below in terms of possible strains, which will not be possible for BN, because of a disturbing signal scattering phenomenon.1 V. Discussion (1) Fiber Stress in the Composites The above-mentioned apparent compression is in perfect agreement with the CTE of the Hi-Nicalon fiber (a 5 3.5 3 1026 /°C between room temperature (RT) and 500°C, from Nippon carbon data sheets) and the celsian matrix (a 5 5.28 3 1026 /°C between RT and 1200°C23). For CMCs, the reinforcement CTE should be greater, to put the matrix, whose tensile strength is low, under residual compression and, in this way, prevent microcracking.24 Yet, low a matrixes usually have loosely packed frameworks and the best matrix choice requires a chemical and mechanical analysis of the multiphase materials. Note that a of the monoclinic celsian phase of SrAl2Si2O8 was measured to be 2.5 3 1026 /°C,25 which confirms that thermal expansion of celsian is a function of the alkali/alkaline-earth content.26 (A) Anticipated Stress: The axial (sA) and transverse (sT) residual stresses in an “infinite” fiber, that is to say, a long and nonfragmented fiber, embedded in a given matrix, can be calculated using the following expressions:27 sA 5 F 2nA EA ~aT 2 am! 1 S 1 2 nT ET 1 1 1 nm Em D~aA 2 am!G 3 EAzDTYS 2nA 2 EA 2 1 2 nT ET 2 1 1 nm Em D (6) sT 5 @aT 2 am 1 nA~aA 2 am!#zDT 2nA 2 EA 2 1 2 nT ET 2 1 1 nm Em (7) The subscripts in Eqs. (6) and (7) are relative to the matrix (m) and the axial (A) or transverse (T) properties of the fiber. n are Poisson ratios and DT represents the difference between room temperature and the so-called “induction temperature,” Ti , at which the matrix becomes too “soft” to constrain the fiber.28 Because of the lack of availability of experimental values of Ti , we will replace it by TS 5 1300°C, a typical sintering temperature for aluminosilicate matrixes. Besides, given the relative isotropy of Hi-Nicalon fibers and, again, because of the lack of data, axial and transverse properties should be considered the same (with a subscript f). If we now put all Poisson ratios to zero, which is partly justified as a first approximation by the very local probing scale of Raman spectroscopy, Eqs. (6) and (7) simplify to sA 5 ~am 2 af!zDTzEf (8) sT 5 ~am 2 af!zDTzS EmEf Em 1 Ef D (9) Taking Em 5 96 GPa23 and the Ef 5 270 GPa Young’s modulus given by Nippon Carbon Co. for Hi-Nicalon fibers, the compressive stress should be about 600–650 MPa, axially, and greater than 150 MPa, radially. The stress measurement derived from (unpolarized) Raman spectra should fall in between since the method is not sensitive to the loading direction, at least in amorphous materials. (B) Experimental Stress Assessment: Looking at Fig. 5, the wavenumbers of the Hi-Nicalon fibers before matrix embedding are rather close: uncoated and coated fibers are suitable to serve as “stress-free” references. Their bandwidths are slightly different, but Fig. 7 in Part I showed that structural evolution of Hi-Nicalon fibers begins around 1300–1400°C, which happens to be the temperature range of p-BN/SiC chemical vapor deposition.1 Fig. 4. (a) Photomicrograph of composite 2 polished parallel to the fiber direction. Spectra were recorded with the 458 nm line (5 mW, 60 s) at each point. (b) Examples of spectra recorded along the fiber and on the fiber crack. May 2001 Raman Study of Hi-Nicalon-Fiber-Reinforced Celsian Composites 1139
1140 Journal of the American Ceramic Society--Gouadec et al Vol 84. No 5 Comp Comp#3 Pristine fiber 135280cm 1352 Coated fibe Extracted 1350 Composites #1 &3 te #2 Fiber extracted from Composite #2 16 Measurement Series Fig. 5. Hi-Nicalon D band wavenumber(corrected by a neon reference line)and bandwidth determined by fitting the spectra recorded in 180 s(A=514.5 nm, P= I mW)on cross sections of different samples(drawings on the right-hand side). Each point corresponds to a region seled sample. The number of recorded spectra is given. perimental wavenumber shift AvExD has three possible p-BN/SiC interphase apparently relaxes the residual stress(#2) It can result from a mechanical stress(Distress), but laser but not the silicon-doped one(#3). Besides, working with the (AVHeating)and/or chemical alterations(Avchemistry )must 514.5 nm line, we found that vp in a fiber surrounded by a also be considered preserved p-BN/SiC double coating was lower than in a nearby △p= fiber with a broken SiC ring, where p-BN layer had dissolved in the celsian. The BN layer acts as a compliant material which (10) protects the fibers and results in better o and m(Weibull modulus) values in the composite. The stress-related shift, the one we want to know, is given by The compressive stresses that we found(100 and 950 MPa) (11) are not exactly in the range anticipated from our micromechanical modeling(150-650 MPa). Yet, the modeling relied on many Concerning the laser heating, it is straightforward to write approximations and did not consider interphase materials at all. On the other hand, we possibly incorrectly estimated the value of s △ilan=P(S (12) Indeed, measurements on carbon and polymeric fibers revealed As for the chemical term. it can be neglected in our case. on a slight but steady softening when stresses are changed from ccount of the small bandwidth difference between embedded and reference fibers. This approximation is ascertained by the fact that means a correlative decrease of S; in compression. Equation(2)is AvEn 0 for the fiber extracted from composite 2 (it has the same only an approximation and second- or third-order polynomial wavenumber in Fig. 5 as the "stress-free"references of the expressions are necessary to unify compressive and tensile behav left-hand side), for which Avstress and AvHeating must vanish. All iors of Courtaulds Graphil (U. K )XAS carbon fibers. However, all, the residual compressive stress will be obtained from the e dependency remains linear for fully crystalline polydiacetylene following expression fibers"and the same must apply to Hi-Nicalon fibers. Indeed, ceramics consist of highly covalent bonds and have a 3D structure ()x-(1)(=) and no failure-initiating defects in compression. Besides, in the △o(GPa)=E△e= peculiar case of the Hi-Nicalon fibers, there is a uniform distribu- tion of C and SiC species. We therefore assume our Sp value is E△甲一P(S=mk- presland almost the same whatever the sign of the applied strain, at least in (13) the explored range ( C) Statistical Relevance of the Results: It must be pointed After Fig. 5, for which P is I mW, the experimental shifts are out that not only the recording conditions, but also the statistical dispersion between the fibers(batch, size, coating, environment, △=△=135365-1351.70=1.95cm ) must be taken into account to ascertain the effect of chemical degradation or stress concentration on the Raman spectra. For △BB=1352.80-1351.70 instance, core wavenumbers in Figs. 5 and 6 differ despite erfectly identical working conditions. Yet, there is good consis If we put these shifts into Eq.(13), E 70G tency in the series of experiments that we conducted on composites I and 2 (Fig. 5). The extreme values that we obtained with 488 nm 1.54cm (Table I), excitation on nine different fibers of the: ame part of a sampl the calculated stress is 110 MPa compression in composite 2, and were -1359.5 and 1360.3 cm. A tenth fiber, isolated from a 960 MPa compression in composites 1 and 3. Hence, the the others by at least 10 diameters, showed a higher value of
The experimental wavenumber shift Dn#Exp D-band has three possible origins. It can result from a mechanical stress (Dn#Stress), but laser heating (Dn#Heating) and/or chemical alterations (Dn#Chemistry) must also be considered: Dn# Exp D 5 n#Sample 2 n#Free-standing fiber 5 Dn#Stress 1 Dn#Heating 1 Dn#Chemistry (10) The stress-related shift, the one we want to know, is given by Dn# Stress 5 SHi-Nicalon ε zDε% (11) Concerning the laser heating, it is straightforward to write Dn# Heating 5 P~Ssample P 2 SFree-standing P ! (12) As for the chemical term, it can be neglected in our case, on account of the small bandwidth difference between embedded and reference fibers. This approximation is ascertained by the fact that Dn#Exp D > 0 for the fiber extracted from composite 2 (it has the same wavenumber in Fig. 5 as the “stress-free” references of the left-hand side), for which DnStress and DnHeating must vanish. All in all, the residual compressive stress will be obtained from the following expression: Ds~GPa! 5 EfzDε 5 S Ef 100DDε% 5 S Ef 100DS Dn#Stress SHi-Nicalon ε D 5 Ef@Dn#Exp D 2 P~Ssample P 2 SFree-standing P !# 100SHi-Nicalon ε (13) After Fig. 5, for which P is 1 mW, the experimental shifts are the following: Dn# Exp #1 5 Dn#Exp #3 5 1353.65 2 1351.70 5 1.95 cm21 Dn# Exp #2 5 1352.80 2 1351.70 5 1.10 cm21 If we put these shifts into Eq. (13), taking Ef 5 270 GPa, SHi-Nicalon ε 5 22.7 cm21 /% (see above), SD in celsian P/514.5 nm 5 20.55 cm21 /mW, and SD Free-standing P/514.5 nm 5 21.54 cm21 /mW (Table I), the calculated stress is 110 MPa compression in composite 2, and a 960 MPa compression in composites 1 and 3. Hence, the p-BN/SiC interphase apparently relaxes the residual stress (#2), but not the silicon-doped one (#3). Besides, working with the 514.5 nm line, we found that n#D in a fiber surrounded by a preserved p-BN/SiC double coating was lower than in a nearby fiber with a broken SiC ring, where p-BN layer had dissolved in the celsian. The BN layer acts as a compliant material which protects the fibers and results in better sr and m (Weibull modulus) values in the composite.29 The compressive stresses that we found (>100 and 950 MPa) are not exactly in the range anticipated from our micromechanical modeling (150–650 MPa). Yet, the modeling relied on many approximations and did not consider interphase materials at all. On the other hand, we possibly incorrectly estimated the value of Si ε . Indeed, measurements on carbon30 and polymeric fibers31 revealed a slight but steady softening when stresses are changed from tensile ones to compressive ones (Etensile . Ecompression), which means a correlative decrease of Si ε in compression. Equation (2) is only an approximation and second- or third-order polynomial expressions are necessary to unify compressive and tensile behaviors of Courtaulds Graphil (U.K.) XAS carbon fibers.32 However, the dependency remains linear for fully crystalline polydiacetylene fibers8 and the same must apply to Hi-Nicalon fibers. Indeed, ceramics consist of highly covalent bonds and have a 3D structure and no failure-initiating defects in compression. Besides, in the peculiar case of the Hi-Nicalon fibers, there is a uniform distribution of C and SiC species. We therefore assume our SD ε value is almost the same whatever the sign of the applied strain, at least in the explored range. (C) Statistical Relevance of the Results: It must be pointed out that not only the recording conditions, but also the statistical dispersion between the fibers (batch, size, coating, environment, . . . ) must be taken into account to ascertain the effect of chemical degradation or stress concentration on the Raman spectra. For instance, core wavenumbers in Figs. 5 and 6 differ despite perfectly identical working conditions. Yet, there is good consistency in the series of experiments that we conducted on composites 1 and 2 (Fig. 5). The extreme values that we obtained with 488 nm excitation on nine different fibers of the same part of a sample were ;1359.5 and 1360.3 cm21 . A tenth fiber, isolated from the others by at least 10 diameters, showed a higher value of Fig. 5. Hi-Nicalon D band wavenumber (corrected by a neon reference line) and bandwidth determined by fitting the spectra recorded in 180 s (l 5 514.5 nm, P 5 1 mW) on cross sections of different samples (drawings on the right-hand side). Each point corresponds to a region selected in the corresponding sample. The number of recorded spectra is given. 1140 Journal of the American Ceramic Society—Gouadec et al. Vol. 84, No. 5
May 2001 Raman Study of Hi-Nicalon-Fiber-Reinforced Celsian Composites 1141 compressive nature of the stress in composite 2, but with a constant ADExbandental(about 1 cm, between"stress-free"references- either desized fibers or extracted fibers- and the wavenumber Laser measured in situ). If Sp really depends linearly on the laser energy (in electronvolts), according to Fig. 3 comments, then we must conclude after Eq.(13)that the stress is wavelength-dependent (Ssample has almost no dependency on wavelength in Table I). As a matter of fact, light penetration in resonant materials is correlated with the wavelength. Assuming that the absorption of Hi-Nicalon fibers should be close to that of amorphous silicon carbide(a-SIC) films, which is ascertained by Raman spectra similitude, the maximal penetrations can be assumed to be about 25 and 75 nm for the 458 and 647.1 nm lines, respectively. There might exist actual stress differences as a function of the distance to the sample surface or, in other words, the analyzed volume. Our observation would indicate a stress gradient with the highest stress measured at the very surface of the sample(blue laser observation). More 1355.51 detailed results would require precise measurements of the laser power and using reference lamps for red and blue excitations 13550 (2) Overcoating Analysis For SiC and Si materials, the only values at our disposal were 4 found in the literature for compressive experiments in diamond 1354,5 anvils on single-crystal pieces. Linear dependencies of 3.53 + 0.21 and 4.28 0.22 cm/GPa were calculated respectively for th TO and o modes of 6H-siC, which our SiC layer mainly consists of(see Part D). TO, and Lo modes are at their expected values(796 and 966 cm, respectively) but TO, mode points at 785-788 cm, while its reported stress-free value is about 789.2 cm-. This would correspond to a huge stress, but this"shift attribution"would be dubious and the great variety of recorded spectra are rather a consequence of great structural variations(SiC Figure 5 in Part I suggested the more intense the silicon signal the weaker the TOz contribution. Silicon unsplit (stress-free wavenumber peaks at 5206cm" It is a triply degenerate mode whose splitting has been fully reported in the literature for stresses along the [111] and [001] directions, biaxial strains in a polycrystalline silicon thin layer. 36,37 The mean value(weighted by the degeneracy)must be considered in disordered silicon and the splitting explains band widening observed in strained silicon. In our samples we detected a mean 2 cm upward shift, would indicate compressive stresses of nearly 1 GPa(th se mean 0 sensitivity is"450 MPa/cm for [100] stress). Such a would be very high for a macroscopic stress, but what is measured here is the very local stress of cross-section two-dimensional mapping(0.5 um directions)performed on a fiber/matrix interface ed fibers)using a 514.5 nm excitation(P= I mW,I VI. S 00s):(a) raph, (b) mapping based on the fitted wavenumbers for the spectra in the black rectangle, (c) bandwidth mapping. The fibers are under compressive residual stress of 950 MPa celsian-matrix composites reinforced with uncoated or p-B(Si)N/ SiC coated Hi-Nicalon fibers. In another celsian-matrix composite <1360.7cm-. Thus, the compression would be attenuated by the where Hi-Nicalon fibers have been coated with undoped p-BN/ presence of surrounding fibers, which would mean, as expected, SiC, residual stress in the fibers seems to have significantly relaxed that the matrix plays a more important role than the interphase to 110 MPa. In addition, the compression seems to reach a higher ssing the fibers level at the fiber core than in the vicinity of the fiber-matrix apparent stress relaxation about 3 m from the fiber-matrix account the effectiveAt results would require (O)taking into D Possible Edge efect: Figure 6(b) shows evidence of an s modulus in compression, (ii) per interphase in composite 1. a band widening is observed at the very forming Sn calibration compression, and (iii) using fibers interface in Fig. 6(c), which suggests a chemical alteration(and, annealed under the ons of composite fabrication as a possibly, fitting errors due to intensity lowering), but the wave- rererence number shift must have a mechanical effect for the region of onstant bandwidth. The best way to ascertain why the c-C bonds Refere edge region are different would be to record spectra on ct33A smallcompression relaxation"was detected in com- Reinforced Cel Ph Colomban, andN.P.Bansal, "Raman Study of Hi-Nicalon-Fiber- -embedded fibers, which would rule out any mechanical posite 2 when moving away from the fiber core, but no proper s" Am ceram.Soc,845]112935(2001) -Ph. Colomban and J. Corset, "Special Issue on Raman(Micro Spectrometry and edge analysis"could be performed where BN contribution was affecting the bandwidths destructive Mechanical Characterization of (E) Wavelength Influence: Rapid mappings performed with SiC Fibers by Raman Spectroscopy, J. Eur. Ceram. Soc., in pres "M. A. White,"Thermal Properties of Solids: Etude in Three-Part Anharmony, four different wavelengths(458/488/514/647 nm) confirmed the Camn.J.Chem,74,l916-21(1996
;1360.7 cm21 . Thus, the compression would be attenuated by the presence of surrounding fibers, which would mean, as expected, that the matrix plays a more important role than the interphase in compressing the fibers. (D) Possible Edge Effect: Figure 6(b) shows evidence of an apparent stress relaxation about 3 mm from the fiber–matrix interphase in composite 1. A band widening is observed at the very interface in Fig. 6(c), which suggests a chemical alteration (and, possibly, fitting errors due to intensity lowering), but the wavenumber shift must have a mechanical effect for the region of constant bandwidth. The best way to ascertain why the C–C bonds of the edge region are different would be to record spectra on nickel-embedded fibers, which would rule out any mechanical effect.33 A small “compression relaxation” was detected in composite 2 when moving away from the fiber core, but no proper “edge analysis” could be performed where BN contribution was affecting the bandwidths. (E) Wavelength Influence: Rapid mappings performed with four different wavelengths (458/488/514/647 nm) confirmed the compressive nature of the stress in composite 2, but with a constant Dn#Experimental D-band (about 1 cm21 , between “stress-free” references— either desized fibers or extracted fibers—and the wavenumber measured in situ). If SD ε really depends linearly on the laser energy (in electronvolts), according to Fig. 3 comments, then we must conclude after Eq. (13) that the stress is wavelength-dependent (Ssample P has almost no dependency on wavelength in Table I). As a matter of fact, light penetration in resonant materials is correlated with the wavelength. Assuming that the absorption of Hi-Nicalon fibers should be close to that of amorphous silicon carbide (a-SiC) films,34 which is ascertained by Raman spectra similitude, the maximal penetrations can be assumed to be about 25 and 75 nm for the 458 and 647.1 nm lines, respectively. There might exist actual stress differences as a function of the distance to the sample surface or, in other words, the analyzed volume. Our observation would indicate a stress gradient with the highest stress measured at the very surface of the sample (blue laser observation). More detailed results would require precise measurements of the laser power and using reference lamps for red and blue excitations. (2) Overcoating Analysis For SiC and Si materials, the only values at our disposal were found in the literature for compressive experiments in diamond anvils on single-crystal pieces. Linear dependencies of 3.53 6 0.21 and 4.28 6 0.22 cm21 /GPa were calculated respectively for the TO and LO modes of 6H-SiC,35 which our SiC layer mainly consists of (see Part I). TO2 and LO modes are at their expected values (796 and 966 cm21 , respectively) but TO1 mode points at 785–788 cm21 , while its reported stress-free value is about 789.2 cm21 . This would correspond to a huge stress, but this “shift attribution” would be dubious and the great variety of recorded spectra are rather a consequence of great structural variations (SiC polytypism).1 Figure 5 in Part I suggested the more intense the silicon signal, the weaker the TO2 contribution. Silicon unsplit (stress-free) wavenumber peaks at 520.6 cm21 . 36 It is a triply degenerate mode whose splitting has been fully reported in the literature for stresses along the [111] and [001] directions,18 or biaxial strains in a polycrystalline silicon thin layer.36,37 The mean value (weighted by the degeneracy) must be considered in disordered silicon and the splitting explains band widening observed in strained silicon. In our samples we detected a mean 2 cm21 upward shift, which would indicate compressive stresses of nearly 1 GPa (the mean sensitivity is ;450 MPa/cm21 for [100] stress16). Such a value would be very high for a macroscopic stress, but what is measured here is the very local stress. VI. Summary The fibers are under compressive residual stress of 950 MPa in celsian-matrix composites reinforced with uncoated or p-B(Si)N/ SiC coated Hi-Nicalon fibers. In another celsian-matrix composite, where Hi-Nicalon fibers have been coated with undoped p-BN/ SiC, residual stress in the fibers seems to have significantly relaxed to 110 MPa. In addition, the compression seems to reach a higher level at the fiber core than in the vicinity of the fiber–matrix interface. More precise results would require (i) taking into account the effective Young’s modulus in compression, (ii) performing SD ε calibrations under compression, and (iii) using fibers annealed under the conditions of composite fabrication as a reference. References 1 G. Gouadec, Ph. Colomban, and N. P. Bansal, “Raman Study of Hi-Nicalon-FiberReinforced Celsian Composites: I, Distribution and Nanostructure of Different Phases,” J. Am. Ceram. Soc., 84 [5] 1129–35 (2001). 2 Ph. Colomban and J. Corset, “Special Issue on Raman (Micro)Spectrometry and Materials Science,” J. Raman Spectrosc., 30 [10] 861–947 (1999). 3 G. Gouadec and Ph. Colomban, “Nondestructive Mechanical Characterization of SiC Fibers by Raman Spectroscopy,” J. Eur. Ceram. Soc., in press. 4 M. A. White, “Thermal Properties of Solids: Etude in Three-Part Anharmony,” Can. J. Chem., 74, 1916–21 (1996). Fig. 6. Example of cross-section two-dimensional mapping (0.5 mm steps in X and Y directions) performed on a fiber/matrix interface in composite 1 (uncoated fibers) using a 514.5 nm excitation (P 5 1 mW, t 5 300 s): (a) photomicrograph; (b) mapping based on the fitted wavenumbers for the spectra in the black rectangle, (c) bandwidth mapping. May 2001 Raman Study of Hi-Nicalon-Fiber-Reinforced Celsian Composites 1141
1142 Journal of the American Ceramic Society--Gouadec et al. Vol 84. No 5 B. A. Weinstein and R. Zallen, ""Pressure Raman Effects in Covalent and J. Wu and Ph Colomban, " Raman Spectroscopy Study or DIstribution in the Continuous Fibre-Reinforced CMC, J. Raman Spectre in Solids 1. Edited by M. Cardona and G. Guntherodt. Springer-Verlag, 1. S. Amer and L. S. Schadler. "The Effect Fibre/Fibre Interact .A Jayaraman, "Diamond Anvil Cell and High-Pressure Physical Investigations," 19-28(1999) 2N. P. Bansal, "Solid State Synthesis and Properties of Monoclinic Celsian, "J. e.Sc,33[19471l-15(1 Amil Cell Press, New York, 1984. T Scanu, J Guglielmi, and Ph. Colomban, "lon Exchange and Hot Corrosion L. S. Schadler and C, Galiotis, "Fundamentals and Applications of Micro Raman trices: A Vibrational and Microstructural Study, Solid State Spectroscopy to Strain Measurements in Fibre Reinforced Composites, Int Mater. tonics,7071,109-20(199 Re,403]116-34(1995 25N. P Bansal. "Influence of Fibre Volume Fraction on Mechanical behaviour( R J. Young, Analysis of C es Using Raman and Fluorescence Micros- CVD SiC Fiber/SrAl2Si2O% Glass-Ceramic Matrix Composites, "J. Adv. Mater, 28 Review," Microsc,1852]199-205(1996 148-58(1 pRm间hm419 Colomban, H. Courret, F. Romain, G. Ge Prepared Pure and Li-Doped Hexacelsian Polymorphs: An IR, Raman, and Thermal ouadec, S. Karlin, and Ph, Colomban, "Raman Extensometry Study of Expansion Study of the B-Phase Stabilization by Short-Range Disorder NLM202 and Hi-Nicalon SiC Fibres," Composites, 29B, 251-61(1998). -G. Gouadec and Ph. Colomban,""Raman Extensometry. Anharmonicity and 27W. Wu, M. Desaeger, I. Verpoest, and J. Varna, "An Improved Analysis of the Stress"(in Fr ) pp. 759-66 in Proceedings of the JNCII(1 Ith Journees Nationales augmentation Test: L, Two-Phase Model, Compos. Sci. mmm2b上Lm Technol.,57,809-19(1997 J. R. Wood, Y Huang, R. J. Young, and G Marom, "Measurement of Thermal tein,"Morphic Effects: ll, Effects of External Fo on the Frequencies of the g-0 Optical Phonons,J. Phys. Chem. Solids, 32, 563-70 ure of Hi-Nicalon Fibers H. Shin, F. H. Pollak, and P. M. Raccah, "Effects of Uniaxial Stress on the Extracted from Celsian Matrix Composites", pp. 3-16 in Ceramic Transactions, Vo Raman Frequencies of T1]O3 and Al2O3, pp. 401-405 in Proceedings of the 3rd 96, Advances in Ceramic Matrix Composites 1. Edited by J. P, Singh and N. P Scattering in Solids( Paris, Fibres. Pa J. w. Ager Ill and M. D. Drory, "Quantitative Measurement of Residual Biaxial 3C. Vlattas and C Galictis, the Behavior of Polymer Fibres Under DhRB我用209ma NM里色6121190(9 Axial Compression,”P 6l. De Wolf, "Stress Measurements in Si Microelectronics Devices Using Raman and S. B. Smith,"Compressional Behaviour of Carbon Fibres. Part II. Softening, "J Mater. Sci, 29, 786-99 Spectroscopy,J. Raman Spectrosc., 30[10]877-83(1999) "Comparative Measurement of Residual Stress in Diamond Coatings by Low- 3G. Gouadec and Ph. Colomban, "De I'Analyse Micro/Nano-Structurale ncident-Beam-Angle-Diffraction and Micro-Raman Spectroscopy, J. Mater. Res. Micromecanique a I'Imagerie des Fibres de renfort d'un Composite a Matrice nalysis to Imaging of Anastassakis, A. Pinczuk, E. Burnstein, F, H. Pollak, and M. Cardona, "Effect $G. Derst, C. Wilbertz, K. L Bhatia, W. Kratschmer, and S. Kalbitzer, " Optica of Static Uniaxial Stress on the Raman Spectrum of Silicon, Solid State Commun N. P Bansal and J 1. Eldridge,"Hi-Nicalon Fiber-Reinforced Celsian Matrix Residual stress in 6H-Sic Composites: Influence of Interface Modification,J. Mater. Res, 13 16] 1530-3 Particles within AL,O / SiC Composites Through Raman Spectroscopy,J.Am. G. Gouadec, J.-P. Forgerit, and Ph Colomban, "Raman Extensometry of C an M. Cheong, H L Marcus, and F. Adar, "Raman Microp Surement SiC Fibers: Conditions Choice by 2D-Correlation"(in Fr ) pp 551-60 in Proceed of Residual Strains at the Interfaces of Si on Quartz,J Mater. Res perieure, Cachan, France, November 2000), Vol. I. Edited by O. Allix, C. Cluzel 37M. Holtz, J. C. Carty, and W M. Duncan, "Ultraviolet Raman Stress Mapping in and J. Lamon. Association pour les Materiaux Composites, Paris, France, 2000 Silicon,"4 ppl. Phys.let,741412008-10(1999)
5 B. A. Weinstein and R. Zallen, “Pressure Raman Effects in Covalent and Molecular Solids”; pp 463–527 in Topics in Applied Physics, Vol. 54, Light Scattering in Solids IV. Edited by M. Cardona and G. Gu¨ntherodt. Springer-Verlag, Heidelberg, Germany, 1984. 6 A. Jayaraman, “Diamond Anvil Cell and High-Pressure Physical Investigations,” Rev. Mod. Phys., 55 [1] 65–108 (1983). 7 J. R. Ferraro, Vibrational Spectroscopy at High External Pressures: The Diamond Anvil Cell. Academic Press, New York, 1984. 8 L. S. Schadler and C. Galiotis, “Fundamentals and Applications of Micro Raman Spectroscopy to Strain Measurements in Fibre Reinforced Composites,” Int. Mater. Rev., 40 [3] 116–34 (1995). 9 R. J. Young, “Analysis of Composites Using Raman and Fluorescence Microscopy—A Review,” J. Microsc., 185 [2] 199–205 (1996). 10D. Le´veˆque and M. H. Auvray, “Study of Carbon-Fibre Strain in Model Composites by Raman Spectroscopy,” Compos. Sci. Technol., 56, 749–54 (1996). 11G. Gouadec, S. Karlin, and Ph. Colomban, “Raman Extensometry Study of NLM202 and Hi-Nicalon SiC Fibres,” Composites, 29B, 251–61 (1998). 12G. Gouadec and Ph. Colomban, “Raman Extensometry: Anharmonicity and Stress” (in Fr.); pp. 759–66 in Proceedings of the JNC11 (11th Journe´es Nationales sur les Composites, Arcachon, France, November 1998), Vol. 2. Edited by J. Lamon and D. Baptiste. Association pour les Materiaux Composites, Paris, France, 1998. 13E. Anastassakis and E. Burstein, “Morphic Effects: II, Effects of External Forces on the Frequencies of the q ; 0 Optical Phonons,” J. Phys. Chem. Solids, 32, 563–70 (1971). 14S. H. Shin, F. H. Pollak, and P. M. Raccah, “Effects of Uniaxial Stress on the Raman Frequencies of Ti2O3 and Al2O3”; pp. 401–405 in Proceedings of the 3rd International Conference on Light Scattering in Solids (Paris, France, 1975). Flammarion, Paris, France, 1975. 15J. W. Ager III and M. D. Drory, “Quantitative Measurement of Residual Biaxial Stress by Raman Spectroscopy in Diamond Grown on a Ti Alloy by Chemical Vapor Deposition,” Phys. Rev. B, 48 [4] 2601–607 (1993). 16I. De Wolf, “Stress Measurements in Si Microelectronics Devices Using Raman Spectroscopy,” J. Raman Spectrosc., 30 [10] 877–83 (1999). 17H. Mohrbacher, K. V. Acker, B. Blanpain, P. V. Houtte, and J.-P. Celis, “Comparative Measurement of Residual Stress in Diamond Coatings by LowIncident-Beam-Angle-Diffraction and Micro-Raman Spectroscopy,” J. Mater. Res., 11 [7] 1776–82 (1996). 18E. Anastassakis, A. Pinczuk, E. Burnstein, F. H. Pollak, and M. Cardona, “Effect of Static Uniaxial Stress on the Raman Spectrum of Silicon,” Solid State Commun., 8, 133–38 (1970). 19N. P. Bansal and J. I. Eldridge, “Hi-Nicalon Fiber-Reinforced Celsian Matrix Composites: Influence of Interface Modification,” J. Mater. Res., 13 [6] 1530–37 (1998). 20G. Gouadec, J.-P. Forgerit, and Ph. Colomban, “Raman Extensometry of C and SiC Fibers: Conditions Choice by 2D-Correlation” (in Fr.); pp. 551–60 in Proceedings of the JNC12 (12th Journe´es Nationales sur les Composites, Ecole Nationale Superieure, Cachan, France, November 2000), Vol. 1. Edited by O. Allix, C. Cluzel, and J. Lamon. Association pour les Materiaux Composites, Paris, France, 2000. 21J. Wu and Ph. Colomban, “Raman Spectroscopy Study on the Stress Distribution in the Continuous Fibre-Reinforced CMC,” J. Raman Spectrosc., 28, 523–29 (1997). 22M. S. Amer and L. S. Schadler, “The Effect of Interphase Toughness on Fibre/Fibre Interaction in Graphite/Epoxy Composites: An Experimental and Modelling Study,” J. Raman Spectrosc., 30 [10] 919–28 (1999). 23N. P. Bansal, “Solid State Synthesis and Properties of Monoclinic Celsian,” J. Mater. Sci., 33 [19] 4711–15 (1998). 24T. Scanu, J. Guglielmi, and Ph. Colomban, “Ion Exchange and Hot Corrosion of Ceramic Composites Matrices: A Vibrational and Microstructural Study,” Solid State Ionics, 70/71, 109–20 (1994). 25N. P. Bansal, “Influence of Fibre Volume Fraction on Mechanical Behaviour of CVD SiC Fiber/SrAl2Si2O8 Glass-Ceramic Matrix Composites,” J. Adv. Mater., 28 [1] 48–58 (1996). 26Ph. Colomban, H. Courret, F. Romain, G. Gouadec, and D. Michel, “Sol-Gel Prepared Pure and Li-Doped Hexacelsian Polymorphs: An IR, Raman, and Thermal Expansion Study of the b-Phase Stabilization by Frozen Short-Range Disorder,” J. Am. Ceram. Soc., 83 [12] 2974–82 (2000). 27W. Wu, M. Desaeger, I. Verpoest, and J. Varna, “An Improved Analysis of the Stresses in a Single-Fibre Fragmentation Test: I, Two-Phase Model,” Compos. Sci. Technol., 57, 809–19 (1997). 28J. R. Wood, Y. Huang, R. J. Young, and G. Marom, “Measurement of Thermal Strains During Compressive Fragmentation in Single-Fibre Composites by Raman Spectroscopy,” Compos. Sci. Technol., 55, 223–29 (1995). 29N. P. Bansal, “Tensile Strength and Microstructure of Hi-Nicalon Fibers Extracted from Celsian Matrix Composites”; pp. 3–16 in Ceramic Transactions, Vol. 96, Advances in Ceramic Matrix Composites IV. Edited by J. P. Singh and N. P. Bansal. American Ceramic Society, Westerville, OH, 1999. 30N. Melanitis and C. Galiotis, “Compressional Behavior of Carbon Fibres, Part, 1: A Raman Spectroscopy Study,” J. Mater. Sci., 25 [12] 5081–90 (1990). 31C. Vlattas and C. Galiotis, “Monitoring the Behavior of Polymer Fibres Under Axial Compression,” Polymer, 32 [10] 1788–93 (1991). 32N. Melanitis, P. L. Tetlow, C. Galiotis, and S. B. Smith, “Compressional Behaviour of Carbon Fibres, Part II: Modulus Softening,” J. Mater. Sci., 29, 786–99 (1994). 33G. Gouadec and Ph. Colomban, “De l’Analyse Micro/Nano-Structurale et Microme´canique a` l’Imagerie des Fibres de Renfort d’un Composite a` Matrice Me´tallique (From Micro/Nano-Structure and Mechanical Analysis to Imaging of Fibers Reinforcing an MMC),” J. Phys. IV, Part 4, 10, 69–74 (2000). 34G. Derst, C. Wilbertz, K. L. Bhatia, W. Kra¨tschmer, and S. Kalbitzer, “Optical Properties of SiC for Crystalline/Amorphous Pattern Fabrication,” Appl. Phys. Lett., 54 [18] 1722–24 (1989). 35J. F. DiGregorio and T. E. Furtak, “Analysis of Residual Stress in 6H-SiC Particles within Al2O3/SiC Composites Through Raman Spectroscopy,” J. Am. Ceram. Soc., 75 [7] 1854–57 (1992). 36Y. M. Cheong, H. L. Marcus, and F. Adar, “Raman Microprobe Measurements of Residual Strains at the Interfaces of Si on Quartz,” J. Mater. Res., 2 [6] 902–909 (1987). 37M. Holtz, J. C. Carty, and W. M. Duncan, “Ultraviolet Raman Stress Mapping in Silicon,” Appl. Phys. Lett., 74 [14] 2008–10 (1999). M 1142 Journal of the American Ceramic Society—Gouadec et al. Vol. 84, No. 5