CARBON PERGAMON Carbon39(2001)635-645 Longitudinal compressive behaviour and microstructure of PAN based carbon fibres Naoyuki Oya", David J. Johnson Textile Physics Laboratory, School of Textile Industries, University of Leeds, Leeds, West Yorkshire LS2 91, UK 13 January 2000, accepted 30 May 200 Abstract The longitudinal compressive behaviour of HS and HM PAN-based carbon fibres has been examined by means of a direct ompression method which can be applied to single carbon fibres. Longitudinal compressive strength was found to be 30 to 50% of tensile strength depending on the modulus levels in carbon fibres used. Longitudinal compressive modulus was estimated making use of the Euler buckling formula applied to buckled samples; compressive modulus was about 50% of tensile modulus. SEM observations in-situ revealed distinct kink bands in HM fibres, which usually develop by splitting failure, this implies local crystallite buckling and the existence of needle-like pores in HM fibres. The compressive strength in HS fibres decreases with increased apparent porosity. HM fibres exhibit smaller amounts of disorder and larger crystallites, and have increased modulus; it may be deduced that large needle-like pores also exist. Therefore, a compressive failure mechanism which involves local buckling failure of large crystallites with insufficient lateral support of the large needle-like pores may be suggested in the case of HM fibres. 2001 Elsevier Science Ltd. All rights reserved Keywords: A. Carbon fibres, C. Scanning electron microscopy (SEM); X-ray diffraction; D. Mechanical properties; Microstructure 1. Introduction stress in tension and compression. However, there has been Carbon fibres are widely used as reinforcement in tudinal compressive properties and their relationships to carbon fibre reinforced plastics(CFRP), metals(CFRM), structural parameters. In particular, it is very difficult to ceramics(CFRC) and C-C composites. Recent develop- measure the compressive properties in carbon fibres at the ments have provided rapid growth in mechanical properties single filament level of polyacrylonitrile-(PAN)and mesophase pitch-(MP)- In our previous work [ 3], we endeavoured to develop a based carbon fibres; now tensile strength and modulus up direct compression technique which can be applied to to 7 GPa and 600 GPa are available in PAN-based carbon single carbon fibres. That method suggested the possibility fibres [1]. However, general concern still exists about their for longitudinal compressive properties to be successfully elatively poor compressive properties along the fibre axis, measured by means of a strain gauge and a minimotor since most materials reinforced by carbon fibres frequently based compression device. In this study, we further fail in compression during practical applications modified the previous technique to achieve a very slow fibres are most probably attributed to their unique micro- nation and direct observations for compressed fibrous structure which consists of carbon crystallite layers, crys- samples with light and scanning electron microscope tallite disorder regions and needle-like pores oriented along (SEM). Furthermore, longitudinal compressive properti the fibre axis [2]. It is understood that such a highly ave been correlated to structural parameters obtained by oriented structure is responsible for the relatively poor wide-angle X-ray scattering (WAXS) study in order to compressive properties, by showing different responses to discuss structure-compressive property relations in carbon fibres. It may be noted that the device shows an advantage Corresponding author for in-situ microscopic studies compared to the work of E-mail address: oya( @ipc. osaka-pct ac Jp(N. Oya Nakatani et al. [4], who recently reported the measurement 0008-6223/01/S-see front matter 2001 Elsevier Science Ltd. All rights reserved PII:S0008-6223(00)00147-0
PERGAMON Carbon 39 (2001) 635–645 Longitudinal compressive behaviour and microstructure of PANbased carbon fibres Naoyuki Oya , David J. Johnson * Textile Physics Laboratory, School of Textile Industries, University of Leeds, Leeds, West Yorkshire LS2 9JT, UK Received 13 January 2000; accepted 30 May 2000 Abstract The longitudinal compressive behaviour of HS and HM PAN-based carbon fibres has been examined by means of a direct compression method which can be applied to single carbon fibres. Longitudinal compressive strength was found to be |30 to 50% of tensile strength depending on the modulus levels in carbon fibres used. Longitudinal compressive modulus was estimated making use of the Euler buckling formula applied to buckled samples; compressive modulus was about 50% of tensile modulus. SEM observations in-situ revealed distinct kink bands in HM fibres, which usually develop by splitting failure; this implies local crystallite buckling and the existence of needle-like pores in HM fibres. The compressive strength in HS fibres decreases with increased apparent porosity. HM fibres exhibit smaller amounts of disorder and larger crystallites, and have increased modulus; it may be deduced that large needle-like pores also exist. Therefore, a compressive failure mechanism which involves local buckling failure of large crystallites with insufficient lateral support of the large needle-like pores may be suggested in the case of HM fibres. 2001 Elsevier Science Ltd. All rights reserved. Keywords: A. Carbon fibres; C. Scanning electron microscopy (SEM); X-ray diffraction; D. Mechanical properties; Microstructure 1. Introduction stress in tension and compression. However, there has been little support from quantitative measurement of longiCarbon fibres are widely used as reinforcement in tudinal compressive properties and their relationships to carbon fibre reinforced plastics (CFRP), metals (CFRM), structural parameters. In particular, it is very difficult to ceramics (CFRC) and C–C composites. Recent develop- measure the compressive properties in carbon fibres at the ments have provided rapid growth in mechanical properties single filament level. of polyacrylonitrile- (PAN) and mesophase pitch-(MP)- In our previous work [3], we endeavoured to develop a based carbon fibres; now tensile strength and modulus up direct compression technique which can be applied to to 7 GPa and 600 GPa are available in PAN-based carbon single carbon fibres. That method suggested the possibility fibres [1]. However, general concern still exists about their for longitudinal compressive properties to be successfully relatively poor compressive properties along the fibre axis, measured by means of a strain gauge and a minimotorsince most materials reinforced by carbon fibres frequently based compression device. In this study, we further fail in compression during practical applications. modified the previous technique to achieve a very slow Poor longitudinal compressive properties in carbon rate of loading, accurate sample gauge-length determi- fibres are most probably attributed to their unique micro- nation and direct observations for compressed fibrous structure which consists of carbon crystallite layers, crys- samples with light and scanning electron microscope tallite disorder regions and needle-like pores oriented along (SEM). Furthermore, longitudinal compressive properties the fibre axis [2]. It is understood that such a highly have been correlated to structural parameters obtained by oriented structure is responsible for the relatively poor wide-angle X-ray scattering (WAXS) study in order to compressive properties, by showing different responses to discuss structure–compressive property relations in carbon fibres. It may be noted that the device shows an advantage *Corresponding author. for in-situ microscopic studies compared to the work of E-mail address: oya@ipc.osaka-pct.ac.jp (N. Oya). Nakatani et al. [4], who recently reported the measurement 0008-6223/01/$ – see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S0008-6223(00)00147-0
N. Oya, D.J. Johnson /Carbon 39(2001)635-645 Mechanical and physical properties of PAN-based carbon fibres Diameter Tensile strength Tensile modulus train-to-failure (g/cm 7.01 1.75 237 4.97 T1000 7.1 M40 5.2 M60J 194 Fibre diameter measured by SEM assuming a circular cross section. Tensile properties based on ASTM D-3379 Density supplied by the manufacturer. of compressive strength in carbon fibres by means of a showing the highest Youngs modulus. All fibres are similar direct technique commercially available from Toray Industries, Inc 2.2. Single fibre compression test 2. Experimenta In the previous study [3], a compression device was constructed for single carbon fibres using a cantilever 2. 1. Carbon fibres used beam with strain gauges attached and a minimotor-cam The cantilever beam was moved laterally towards a fixed Seven kinds of PAN-based carbon fibres including four fibre sample with a rotation of the elliptical cam, and the high strength(HS)and three high modulus(HM)fibres strain gauges detected the small force applied on the were investigated. The physical properties of these fibres sample. However, that device had problems concerning a are presented in Table 1. HS fibres have substantially fast rate of loading, possible misalignment of fibres, and improved tensile strength whereas their tensile modulus inaccurate gauge-length determination. As suggested then has not changed greatly. It should be noted that T700S and further modifications were needed to obtain a more cor T1000 were supplied as epoxy-sized. HM fibres show trolled loading motion and in-situ micro observa- higher tensile modulus in the order of M40J, M50J and tions in order to have accurate sample gauge-length and M60J, but their tensile strength decreases in the same better understanding of the compressive failure behaviour order. As seen from density values in HM fibres, M60J of carbon fibres presumably has the most crystallite layers packed as aA schematic diagram of the present compression result of the highest graphitization temperature, thus is shown in Fig. 1. Firstly, a single carbon fibre Brass stage train gauges (Top view of the sample stage) Cantilever beam Disk stage Reduction stag reduction gear Fig. 1. Compression device for single carbon fibre
636 N. Oya, D.J. Johnson / Carbon 39 (2001) 635 –645 Table 1 Mechanical and physical properties of PAN-based carbon fibres ab b c Diameter Tensile strength Tensile modulus Strain-to-failure Density3 (mm) (GPa) (GPa) (%) (g/cm ) T300 7.01 3.5 210 1.68 1.75 T700S 6.75 5.3 237 2.23 1.82 T800H 4.97 6.4 306 2.19 1.81 T1000 5.00 7.1 294 2.40 1.82 M40J 5.21 4.9 335 1.45 1.77 M50J 5.19 4.3 430 1.00 1.88 M60J 5.07 3.5 535 0.65 1.94 a Fibre diameter measured by SEM assuming a circular cross section. b Tensile properties based on ASTM D-3379. c Density supplied by the manufacturer. of compressive strength in carbon fibres by means of a showing the highest Young’s modulus. All fibres are similar direct technique. commercially available from Toray Industries, Inc. 2.2. Single fibre compression test 2. Experimental procedures In the previous study [3], a compression device was constructed for single carbon fibres using a cantilever 2.1. Carbon fibres used beam with strain gauges attached and a minimotor-cam. The cantilever beam was moved laterally towards a fixed Seven kinds of PAN-based carbon fibres including four fibre sample with a rotation of the elliptical cam, and the high strength (HS) and three high modulus (HM) fibres strain gauges detected the small force applied on the were investigated. The physical properties of these fibres sample. However, that device had problems concerning a are presented in Table 1. HS fibres have substantially fast rate of loading, possible misalignment of fibres, and improved tensile strength whereas their tensile modulus inaccurate gauge-length determination. As suggested then, has not changed greatly. It should be noted that T700S and further modifications were needed to obtain a more conT1000 were supplied as epoxy-sized. HM fibres show trolled loading motion and in-situ microscopic observahigher tensile modulus in the order of M40J, M50J and tions in order to have accurate sample gauge-length and M60J, but their tensile strength decreases in the same better understanding of the compressive failure behaviour order. As seen from density values in HM fibres, M60J of carbon fibres. presumably has the most crystallite layers packed as a A schematic diagram of the present compression device result of the highest graphitization temperature, thus is shown in Fig. 1. Firstly, a single carbon fibre sample Fig. 1. Compression device for single carbon fibre
N. Oya, D.J. Johnson /Carbon 39(2001)635-645 was randomly chosen from a tow and vertically cut by a sharp razor blade to obtain a fat end-surface. The sample X Pgra (2) was then placed on a metallic stage by means of yanoacrylate adhesive, with some portion sticking out. where per is the density of pure graphite (por=2.268 This portion was regarded as the sample gauge length, and /cm). The total pore volume pore was defined as was accurately adjustable from 20 to 550 um with the aid of an optical microscope. The sample stage was then connected to micrometer screw threads that rotate with a C minimotor and reduction gear(68, 200: 1). By slowly rotating the motor, the sample stage moved linearly where Pobs is the density of observed carbon fibre. The forward at 0. 13 m/s, and eventually, the fibre sample total porosity p was then calculated from made contact with a thin cantilever beam causing it to be p=pore X Pobs X 100 deflected. The cantilever had strain gauges attached in order to detect the small deflection. With a suitable calibration for the deflection and force applied to the 3. Experimental results cantilever established, longitudinal compressive force on the fibre sample was monitored during the course of 3.1. Longitudinal compressive behaviour testing. In addition, this device was able to operate not only under an optical microscope, but also in a scanning 3.1.1.Lon opressIve sireng electron-microscope chamber in order to observe longi Fig. 2 shows typical stress-time curves obtained with tudinal compressive behaviour of carbon fibres in detail different sample gauge lengths. Compressive stress linearly In this study, tensile recoil tests were also carried out to increased until the final failure with short gauge lengths measure longitudinal compressive strength in carbon fibres whereas a reduction of stress was evident with long gauge for comparative purposes. Details of the recoil technique ths for each carbon fibre such a difference was due to can be found elsewhere [ 3, 5-10 different compression modes of samples, either axial compression or buckling instability. Therefore, it was thought that the buckling effect can be avoided by making 2.3. Wide-angle x-ray diffraction the gauge length short enough, and longitudinal compres- sive strength must be estimated with such a gauge length X-ray diffraction patterns of aligned fibres Fig. 3 shows a typical plot of longitudinal compressive orded using a position-sensitive detector(PSD) with Cu strength as a function of sample gauge length with ka radiation (A=0.154 nm)operating at 40 kV and 16 mA. indications of compression modes. Compressive strength he diffraction pattern was normalised, corrected and generally declined because of the buckling effect with eaks resolved using standard computational methods [11] uge length i However. each carbon fibre showed recently adapted for PC analysis. The interlayer spacing the true axial compression mode at sufficiently short gauge looz and the crystallite size in the c-axis direction Le were lengths. As long as the axial compression mode appeared, estimated from the position and half-height width of the compressive strength was almost constant for any gauge esolved(002) refection, position being calibrated from length. Therefore, as a criterion for longitudinal co standard silver peak positions. The crystallite sizes in the sive strength, a gauge length of 20 um, the -axis directions, Lal and lai, were determined from the practical length with this method, was adopted (100) reflections on the meridian and equator, respectively parison purpose As a measure of disorder, the relative degree of in- It is noted that the buckling mode did not appear with tracrystallite imperfection D was calculated from the gauge lengths up to around 100 um with th simple relation compression device(see Fig 3 in Ref. 3)). The difference with the present device may be due to the rate of loading D 100 (1) appropriate compression mode was not detected with the former device due to the high strain rate whereas it is in a turbostratic carbon, perfect graphite, and the observed nd the obss oo clearly identified with the present device owing to the low where durb, dgra and dobs are the interlayer spacings door strain rate and direct microscope observations Table 2 lists the direct longitudinal compressive carbon fibre, respectively; it may be noted that durb was strengths together with the recoil and tensile strengths in taken as 0.350 nm. and d=0.335 nm each carbon fibre. In general, the direct compressive The microporosity p was estimated from doo? and strength was higher than the recoil strength. It is also seen density values in carbon fibres [12]. Firstly, the X- that the compressive strength was much lower than the density p was calculated from the interlayer spacing do tensile strength. The scatter of compressive strength was comparable to that of tensile strength. It is not possible to
N. Oya, D.J. Johnson / Carbon 39 (2001) 635 –645 637 was randomly chosen from a tow and vertically cut by a dgra sharp razor blade to obtain a flat end-surface. The sample r 5 3 ] r (2) x gra dobs was then placed on a metallic stage by means of cyanoacrylate adhesive, with some portion sticking out. where rgra gra is the density of pure graphite (r 52.268 3 This portion was regarded as the sample gauge length, and g/cm ). The total pore volume V was defined as pore was accurately adjustable from 20 to 550 mm with the aid 1 1 of an optical microscope. The sample stage was then V 5 2 ] ] (3) pore connected to micrometer screw threads that rotate with a r r obs x DC minimotor and reduction gear (68,200:1). By slowly where r is the density of observed carbon fibre. The obs rotating the motor, the sample stage moved linearly total % porosity p was then calculated from forward at 0.13 mm/s, and eventually, the fibre sample made contact with a thin cantilever beam causing it to be p 5V 3 r 3 100. (4) pore obs deflected. The cantilever had strain gauges attached in order to detect the small deflection. With a suitable calibration for the deflection and force applied to the 3. Experimental results cantilever established, longitudinal compressive force on the fibre sample was monitored during the course of 3.1. Longitudinal compressive behaviour testing. In addition, this device was able to operate not only under an optical microscope, but also in a scanning- 3.1.1. Longitudinal compressive strength electron-microscope chamber in order to observe longi- Fig. 2 shows typical stress–time curves obtained with tudinal compressive behaviour of carbon fibres in detail. different sample gauge lengths. Compressive stress linearly In this study, tensile recoil tests were also carried out to increased until the final failure with short gauge lengths, measure longitudinal compressive strength in carbon fibres whereas a reduction of stress was evident with long gauge for comparative purposes. Details of the recoil technique lengths for each carbon fibre. Such a difference was due to can be found elsewhere [3,5–10]. different compression modes of samples, either axial compression or buckling instability. Therefore, it was thought that the buckling effect can be avoided by making 2.3. Wide-angle X-ray diffraction the gauge length short enough, and longitudinal compressive strength must be estimated with such a gauge length. X-ray diffraction patterns of aligned fibres were re- Fig. 3 shows a typical plot of longitudinal compressive corded using a position-sensitive detector (PSD) with Cu strength as a function of sample gauge length with K radiation (l50.154 nm) operating at 40 kV and 16 mA. a indications of compression modes. Compressive strength The diffraction pattern was normalised, corrected and generally declined because of the buckling effect with peaks resolved using standard computational methods [11] gauge length increase. However, each carbon fibre showed recently adapted for PC analysis. The interlayer spacing the true axial compression mode at sufficiently short gauge d002 and the crystallite size in the c-axis direction Lc were lengths. As long as the axial compression mode appeared, estimated from the position and half-height width of the compressive strength was almost constant for any gauge resolved (002) reflection, position being calibrated from length. Therefore, as a criterion for longitudinal compresstandard silver peak positions. The crystallite sizes in the sive strength, a gauge length of 20 mm, the shortest a-axis directions, Lai and La', were determined from the practical length with this method, was adopted for com- (100) reflections on the meridian and equator, respectively. parison purposes. As a measure of disorder, the relative degree of in- It is noted that the buckling mode did not appear with tracrystallite imperfection Dc was calculated from the gauge lengths up to around 100 mm with the previous simple relation compression device (see Fig. 3 in Ref. [3]). The difference with the present device may be due to the rate of loading; dturb obs 2 d ]]] the appropriate compression mode was not detected with D 5F G 1 2 3 100 (1) c dturb gra 2 d the former device due to the high strain rate whereas it is clearly identified with the present device owing to the low where d , d and d are the interlayer spacings d strain rate and direct microscope observations. turb gra obs 002 in a turbostratic carbon, perfect graphite, and the observed Table 2 lists the direct longitudinal compressive carbon fibre, respectively; it may be noted that d was strengths together with the recoil and tensile strengths in turb taken as 0.350 nm, and d 50.335 nm. each carbon fibre. In general, the direct compressive gra The microporosity p was estimated from d002 and strength was higher than the recoil strength. It is also seen density values in carbon fibres [12]. Firstly, the X-ray that the compressive strength was much lower than the density r was calculated from the interlayer spacing d tensile strength. The scatter of compressive strength was x 002 from comparable to that of tensile strength. It is not possible to
N. Oya, D.J. Johnson /Carbon 39(2001)635-645 confirmed that the com proved with sFF 20mic increase of the tensile Stress (GPa ver, the improve- 7.01m ment of tensile strens nore remarkable 20 especially for HS fibres, than that of compressive strength This may be reflected by the fact that the highly oriented 1.8 structure of carbon fibre is more convenient to improve the tensile properties 12 3.1.2. Longitudinal compressive modulus 0.8 The direct evaluation of modulus vas not very successful in the present device of large 0.4 1.92GPa experimental errors which were mainly difficulty of measuring the microscopic machine com- pliance at very small force level However, the longitudinal compressive modulus may be Time (sec) evaluated by making use of the Euler buckling formula pplied for buckled long samples. This (a)20um derived to account for the critical buckling load of an axially compressed straight column just like a carbon filament under longitudinal compression. If the straight Sample: T300 column has a length L. cross-sectional moment of inertia I Stress (GPa) FL= 100mic and elastic modulus E, and is subjected to a centrally FD= 7. 01mic applied compressive load P, the critical buckling load P 2.0 for the column to show elastic buckling can be expressed P=o tet The factor a depends on the boundary condition of the 0.8 column. For the particular case seen in this work, the clamped-simple support condition which gives a=2.04 0. 68GPa may be used. The elastic modulus E can be obtained if the 47 89sec critical buckling load Per is experimentally measured from 0.0 kled long samples with any particular gauge lengt 020406080100 However, it must be noted that the formula is derived by Time sec) applying the equilibrium equations to the isotropic column in a slightly deformed state. Assuming that the neutral plane of the fiexure exists in the middle of the cross- Fig. 2. Stress-time curves with different sample gauge lengths; section, the apparent elastic modulus E must be considered (b)100 as the average of tensile and compressive moduli for anisotropic carbon fibres obtain the scatter of data in the recoil method. in which e E+E only one representative value is available as a critical rength from several samples. where Et and ec are tensile and compressive modul Fig. 4 presents a comparison of the direct and recoil respectively. As the tensile modulus e, is already ki compressive strengths for each carbon fibre. It is apparent the compressive modulus E can be readily calculated if that both strengths showed different values but in some- the average modulus E is obtained experimentally what similar trends. The reason why the tensile recoil Fig. 6 shows a typical plot of average modulus as a method resulted in lower strengths is probably associated gth for T300 fibre. The with the dynamic loading [6 and the buckling problem modulus generally showed a constant value with long [7, 8] of long samples gauge lengths, but declined with short lengths. The reduc- Fig 5 compares the longitudinal compressive and tensile tion of modulus may be attributed to a change of buckling ngths The compressive strengths were definitely lower; mode from a clamped-simple support(a=2.04)to a from about 30 to 50% of the tensile strengths. It was also 'clamped-free(a=0.25)condition. By reducing the gauge
638 N. Oya, D.J. Johnson / Carbon 39 (2001) 635 –645 confirmed that the compressive strengths improved with increase of the tensile strengths. However, the improvement of tensile strength was much more remarkable, especially for HS fibres, than that of compressive strength. This may be reflected by the fact that the highly oriented structure of carbon fibre is more convenient to improve the tensile properties. 3.1.2. Longitudinal compressive modulus The direct evaluation of modulus and strain was not very successful in the present device because of large experimental errors which were mainly associated with the difficulty of measuring the microscopic machine compliance at very small force level. However, the longitudinal compressive modulus may be evaluated by making use of the Euler buckling formula applied for buckled long samples. This formula was derived to account for the critical buckling load of an axially compressed straight column just like a carbon filament under longitudinal compression. If the straight column has a length L, cross-sectional moment of inertia I, and elastic modulus E, and is subjected to a centrally applied compressive load P, the critical buckling load Pcr for the column to show elastic buckling can be expressed by 2 p EI P 5 a]]. (5) cr 2 L The factor a depends on the boundary condition of the column. For the particular case seen in this work, the ‘clamped-simple support’ condition which gives a 52.04 may be used. The elastic modulus E can be obtained if the critical buckling load Pcr is experimentally measured from buckled long samples with any particular gauge length. However, it must be noted that the formula is derived by applying the equilibrium equations to the isotropic column in a slightly deformed state. Assuming that the neutral plane of the flexure exists in the middle of the crosssection, the apparent elastic modulus E must be considered Fig. 2. Stress–time curves with different sample gauge lengths; as the average of tensile and compressive moduli for (a) 20 mm and (b) 100 mm. anisotropic carbon fibres as follows; Et c 1 E obtain the scatter of data in the recoil method, in which E 5 ]] (6) 2 only one representative value is available as a critical strength from several samples. where E and E are tensile and compressive moduli, t c Fig. 4 presents a comparison of the direct and recoil respectively. As the tensile modulus E is already known, t compressive strengths for each carbon fibre. It is apparent the compressive modulus E can be readily calculated if c that both strengths showed different values but in some- the average modulus E is obtained experimentally. what similar trends. The reason why the tensile recoil Fig. 6 shows a typical plot of average modulus as a method resulted in lower strengths is probably associated function of sample length for T300 fibre. The average with the dynamic loading [6] and the buckling problem modulus generally showed a constant value with long [7,8] of long samples. gauge lengths, but declined with short lengths. The reducFig. 5 compares the longitudinal compressive and tensile tion of modulus may be attributed to a change of buckling strengths. The compressive strengths were definitely lower; mode from a ‘clamped-simple support’ (a 52.04) to a from about 30 to 50% of the tensile strengths. It was also ‘clamped-free’ (a 50.25) condition. By reducing the gauge
N. Oya, D.J. Johnson /Carbon 39(2001)635-645 639 1.0 04 200 ge length(um) Fig 3. Change of longitudinal compressive strength against sample gauge length(L, axially compressed; L, buckled samples he sample did not easily show typical buckling Comparison of longitudinal strength values in PAN-based carbon slipped due to the low frictional for fibres he fibre and the cantilever before the buckling load was reached. In other words, the sample showed a strength slippage before the clamped-simple support buckling (GPa) (GPa) Pa) occurred and the modulus had been underestimated with 1.8(16.9 5.3(21.4) Table 3 lists average, compressive and tensile moduli of T800H 23(146 each fibre. Average modulus was calculated using the 28(18.8) 7. 1(193) critical buckling load obtained with sufficiently long gauge 8(13.0) 49(125) lengths to eliminate any effect of different buckling modes 4.3(13.0) Fig. 7 shows a comparison of longitudinal compressive M60J 35(14.1) and tensile moduli in carbon fibres Compressive modulus Coefficient of variation(%)in parentheses. was generally lower than tensile modulus; <50% of the 3.0 8 Direct method ■ Recoil method g2.0 15 T300T700sT800HT1000M40M50JM60J Fig. 4. Comparison of direct and recoil compressive strengths
N. Oya, D.J. Johnson / Carbon 39 (2001) 635 –645 639 Fig. 3. Change of longitudinal compressive strength against sample gauge length (h, axially compressed; j, buckled samples). Table 2 length, the sample did not easily show typical buckling Comparison of longitudinal strength values in PAN-based carbon because its end slipped due to the low frictional force fibres between the fibre and the cantilever before the buckling Compressive Recoil Tensile load was reached. In other words, the sample showed a strength strength strength slippage before the ‘clamped-simple support’ buckling (GPa) (GPa) (GPa) occurred and the modulus had been underestimated with a short gauge lengths. T300 1.8 (16.9) 1.0 3.5 (12.6) T700S 2.4 (15.6) 1.6 5.3 (21.4) Table 3 lists average, compressive and tensile moduli of T800H 2.3 (14.6) 1.6 6.4 (14.8) each fibre. Average modulus was calculated using the T1000 2.8 (18.8) 2.2 7.1 (19.3) critical buckling load obtained with sufficiently long gauge M40J 1.8 (13.0) 1.0 4.9 (12.5) lengths to eliminate any effect of different buckling modes. M50J 1.3 (15.5) 0.7 4.3 (13.0) Fig. 7 shows a comparison of longitudinal compressive M60J 1.0 (13.9) 0.5 3.5 (14.1) and tensile moduli in carbon fibres. Compressive modulus a Coefficient of variation (%) in parentheses. was generally lower than tensile modulus; |50% of the Fig. 4. Comparison of direct and recoil compressive strengths
N. Oya, D.J. Johnson /Carbon 39(2001)635-645 B Compressive strength ■ Tensile strength 40 045 0.29 T300T70sT800HT100M40J M50J M60J Ratio of compressive strength to tensile strength Fig. 5. Comparison of longitudinal compressive and tensile strengths. tensile modulus except for T700S and T1000 fibres. These to form the smooth surface. Therefore, it is not appropriate two fibres exhibited higher compressive moduli presumab- to directly compare the compressive moduli of T700S and ly due to the epoxy-sizing effect Since the modulus was T1000 here with other non-sized fibres based on the flexure of fibres, the results were extremely ensitive to the surface state. It is known that non-sized 3.1.3. Longitudinal compressive failure aspect carbon fibres generally show a rugosity including distinct In-situ SEM observations were also attempted for flaws on the lateral surface. In such a case, the estimated compressed single carbon fibres For HS fibres, there was modulus from the buckling test could be lowered because no significant surface cha lange during compression until the fibre is likely to be bent at parts with reduced diameter. catastrophic failure. On the contrary, for HM fibres(pai On the other hand, the sized fibres may give higher ticularly for M60d), kink bands appeared on the surface modulus since the defects are covered with the sizing ager shown in Fig. 8a and b. Subsequently, as shown in Fig. g 0 100 00 600 Fig. 6. Change of average modulus against sample gauge length(T300)
640 N. Oya, D.J. Johnson / Carbon 39 (2001) 635 –645 Fig. 5. Comparison of longitudinal compressive and tensile strengths. tensile modulus except for T700S and T1000 fibres. These to form the smooth surface. Therefore, it is not appropriate two fibres exhibited higher compressive moduli presumab- to directly compare the compressive moduli of T700S and ly due to the epoxy-sizing effect. Since the modulus was T1000 here with other non-sized fibres. based on the flexure of fibres, the results were extremely sensitive to the surface state. It is known that non-sized 3.1.3. Longitudinal compressive failure aspect carbon fibres generally show a rugosity including distinct In-situ SEM observations were also attempted for flaws on the lateral surface. In such a case, the estimated compressed single carbon fibres. For HS fibres, there was modulus from the buckling test could be lowered because no significant surface change during compression until the fibre is likely to be bent at parts with reduced diameter. catastrophic failure. On the contrary, for HM fibres (parOn the other hand, the sized fibres may give higher ticularly for M60J), kink bands appeared on the surface as modulus since the defects are covered with the sizing agent shown in Fig. 8a and b. Subsequently, as shown in Fig. 8c Fig. 6. Change of average modulus against sample gauge length (T300)
N. Oya, D.J. Johnson /Carbon 39(2001)635-645 Comparison of longitudinal modulus values in PAN-based carbon fibres Average modulus Compressive modul Tensile modulus (GPa) 210 T1000 M40 M60J and d, the kink bands usually developed into splitting had difference in of disorder, crystallite failure along the fibre axis. Such surface changes strongly porosity suggest local buckling of crystallites due to large needle- like pores in this type of fibre 3.3. Structure-compressive property relationships 3.3.1. 3. 2. Microstructure Fig. 9 shows the relationship between longitudinal compressive strength and intracrystalline disorder D. for Table 4 lists the structural parameters obtained from carbon fibres used in this work. HS fibres showed different /AXS study. Interlayer spacing d tracrystallite compressive strength with the same amount of disorder disorder D showed the same values for all HS fibres On Therefore, the differences of compressive strength in HS the other hand, both values declined with increase of fibre variants were probably due to another structural modulus level for HM fibres. Crystallite sizes in three difference. HM fibres showed higher compressive strength no size changes for HS fibres and significant increase for propagation with more disorder. HM fibres. As regards total porosity p, T300 showed the highest value among HS fibres. For HM fibres, the 3.3.2. Compressive strength vs. crystallite size porosity decreased with increase of modulus. From these Longitudinal compressive strength was compared with measurements, it was found that Hs fibres had only crystallite thickness Le in Fig. 10. HS fibres showed structural difference in the porosity, whereas HM fibres different compressive strength wit a Tensile modulus s Ratio of compressive modulus to tensile modulus Fig. 7. Comparison of longitudinal compressive and tensile moduli
N. Oya, D.J. Johnson / Carbon 39 (2001) 635 –645 641 Table 3 Comparison of longitudinal modulus values in PAN-based carbon fibres Average modulus Compressive modulus Tensile modulus (GPa) (GPa) (GPa) T300 158 106 210 T700S 218 199 237 T800H 215 124 306 T1000 270 246 294 M40J 277 219 335 M50J 343 256 430 M60J 409 283 535 and d, the kink bands usually developed into splitting had difference in respect of disorder, crystallite size, and failure along the fibre axis. Such surface changes strongly porosity. suggest local buckling of crystallites due to large needlelike pores in this type of fibre. 3.3. Structure–compressive property relationships 3.3.1. Compressive strength vs. disorder region 3.2. Microstructure Fig. 9 shows the relationship between longitudinal compressive strength and intracrystallite disorder D for c Table 4 lists the structural parameters obtained from carbon fibres used in this work. HS fibres showed different WAXS study. Interlayer spacing d and intracrystallite compressive strength with the same amount of disorder. 002 disorder Dc showed the same values for all HS fibres. On Therefore, the differences of compressive strength in HS the other hand, both values declined with increase of fibre variants were probably due to another structural modulus level for HM fibres. Crystallite sizes in three difference. HM fibres showed higher compressive strength directions (L , L and L ) also exhibited similar trends; with increase of disorder amount, probably hindering crack c a' ai no size changes for HS fibres and significant increase for propagation with more disorder. HM fibres. As regards total % porosity p, T300 showed the highest value among HS fibres. For HM fibres, the 3.3.2. Compressive strength vs. crystallite size porosity decreased with increase of modulus. From these Longitudinal compressive strength was compared with measurements, it was found that HS fibres had only crystallite thickness L in Fig. 10. HS fibres showed c structural difference in the porosity, whereas HM fibres different compressive strength with almost the same Fig. 7. Comparison of longitudinal compressive and tensile moduli
N. Oya, D.J. Johnson /Carbon 39(2001)635-645 the 101..t m Cantilever beam with strain gauges Kink band Fibre Failure 20um 20um Drving stage Kink band formation Kink band ① Splitting 20m Post-kink splitting Fig. 8. In-situ SEM observations for HM fibre (M60J) crystallite size. Therefore, there should be another structur width Lai and length Lal also showed a similar al reason for different compressive strength in these fibres on compressive HM fibres showed lower compressive strength with in- crease of crystallite size, because large crystallites are very 3.3.3. Compressive strength vs. porosity brittle and sensitive in compressive deformation. Crys- Fig. 11 shows the comparison between longitudina Structural parameters obtained from wide-angle X-ray scatterin Lai (nm) 3.14 T1000 M40J 199 7
642 N. Oya, D.J. Johnson / Carbon 39 (2001) 635 –645 Fig. 8. In-situ SEM observations for HM fibre (M60J). crystallite size. Therefore, there should be another structur- tallite width L and length L also showed a similar a' ai al reason for different compressive strength in these fibres. effect on compressive strength. HM fibres showed lower compressive strength with increase of crystallite size, because large crystallites are very 3.3.3. Compressive strength vs. porosity brittle and sensitive in compressive deformation. Crys- Fig. 11 shows the comparison between longitudinal Table 4 Structural parameters obtained from wide-angle X-ray scattering d (nm) D (%) L (nm) L (nm) L (nm) p (%) 002 cc a' ai T300 0.347 80.0 2.60 4.98 2.98 19.7 T700S 0.347 80.0 2.66 5.37 3.14 16.5 T800H 0.347 80.0 2.70 5.67 3.05 17.0 T1000 0.347 80.0 2.61 5.34 3.16 16.5 M40J 0.342 46.7 3.47 6.41 5.11 19.9 M50J 0.340 33.3 4.82 7.49 7.02 15.3 M60J 0.339 26.7 6.84 8.38 8.06 13.0
N. Oya, D.J. Johnson /Carbon 39(2001)635-645 compressive strength and total P. HS fibres T1000 showed lower compressive strength increased po- a25 rosity. However, HM fibres showed compressive strength with more porosity. The inverse relationships seen 三2 in HS and HM fibres are probably due to different types of pores in both categories M60J Y 4. Discussion O0.5 It was revealed that longitudinal compressive strength in PAN-based carbon fibres can only be estimated with sufficiently short gauge lengths of around 20 um. Accord- Intracrystalline disorder DC (% ng to the previous paper, it should be noted that the clar Fig. 9. Relationship between longitudinal compressive streng effect on samples must be taken into account in the case and intracrystalline disorder in PAN-based carbon fibres anisotropic carbon fibres [3, 13-16]. The results obtained this work may well include the effect of adhesive used fix the samples to some extent. It is not certain how the clamp effect infuences the compressive properties in the carbon fibres measured; however, it should be enough to discuss compressive properties of different fibres with the same condition as a comparative stud The structure-compressive property relations in PAN- based carbon fibres suggested that crystallite disorder and T crystallite stacking size appreciably affected longitudina compressive strength only when the fibres were produced with higher graphitization temperature to attain higher modulus as seen in HM fibres. HS fibres had almost the same disorder amount and crystallite size. however. thei 605 compressive strength decreased with increase of porosity t is thought that increased pores became stress concen- trators to induce compressive failure at lower stress level in Hs fibres Crystallite thickness Lc(nm) HM fibres exhibited lower compressive strength with Fig. 10. Relationship between longitudinal compressive streng increase of fibre modulus. because of reduced disorder and and crystallite thickness in PAN-based carbon fibres enlarged crystallites. The compressive strength also de- creased with reduced porosity; it was a completely differ ent trend compared with HS fibres. This result implies that 30 different types of pores must be considered in HS and HM fibres PAN-based carbon fibre basically contains a turbostratic carbon structure in which interlinking crystallites exist between crystallite ribbons enclosing pores [17] However. the structure of carbon fibres moves from a system consisting of many small pores to one with fe larger needle-like pores as the modulus increases [12, 18- 23]. HM PAN-based carbon fibres used in this study may also have such a structural change, these fibres are produced with higher graphitization temperature, and the internal pores are much larger and longer as if small pores are coalesced and oriented along the fibre direction 15 Consequently, the number of pores is smaller than in HS Porosity p(%) fibres, even though the pore size is much larger. It is Fig.I1. Relationship between longitudinal compressive strength thought that such few and large needle-like pores cause the and porosity in PAN-based carbon fibres low porosity and significantly lower the compressi
N. Oya, D.J. Johnson / Carbon 39 (2001) 635 –645 643 compressive strength and total % porosity p. HS fibres showed lower compressive strength with increased porosity. However, HM fibres showed higher compressive strength with more porosity. The inverse relationships seen in HS and HM fibres are probably due to different types of pores in both categories. 4. Discussion It was revealed that longitudinal compressive strength in PAN-based carbon fibres can only be estimated with sufficiently short gauge lengths of around 20 mm. According to the previous paper, it should be noted that the clamp effect on samples must be taken into account in the case of Fig. 9. Relationship between longitudinal compressive strength anisotropic carbon fibres [3,13–16]. The results obtained in and intracrystallite disorder in PAN-based carbon fibres. this work may well include the effect of adhesive used to fix the samples to some extent. It is not certain how the clamp effect influences the compressive properties in the carbon fibres measured; however, it should be enough to discuss compressive properties of different fibres with the same condition as a comparative study. The structure–compressive property relations in PANbased carbon fibres suggested that crystallite disorder and crystallite stacking size appreciably affected longitudinal compressive strength only when the fibres were produced with higher graphitization temperature to attain higher modulus as seen in HM fibres. HS fibres had almost the same disorder amount and crystallite size, however, their compressive strength decreased with increase of porosity. It is thought that increased pores became stress concentrators to induce compressive failure at lower stress level in HS fibres. HM fibres exhibited lower compressive strength with Fig. 10. Relationship between longitudinal compressive strength increase of fibre modulus, because of reduced disorder and and crystallite thickness in PAN-based carbon fibres. enlarged crystallites. The compressive strength also decreased with reduced porosity; it was a completely different trend compared with HS fibres. This result implies that different types of pores must be considered in HS and HM fibres. PAN-based carbon fibre basically contains a turbostratic carbon structure in which many interlinking crystallites exist between crystallite ribbons enclosing pores [17]. However, the structure of carbon fibres moves from a system consisting of many small pores to one with few larger needle-like pores as the modulus increases [12,18– 23]. HM PAN-based carbon fibres used in this study may also have such a structural change; these fibres are produced with higher graphitization temperature, and the internal pores are much larger and longer as if small pores are coalesced and oriented along the fibre direction. Consequently, the number of pores is smaller than in HS fibres, even though the pore size is much larger. It is Fig. 11. Relationship between longitudinal compressive strength thought that such few and large needle-like pores cause the and porosity in PAN-based carbon fibres. low porosity and significantly lower the compressive
N. Oya, D.J. Johnson /Carbon 39(2001)635-645 strength in HM fibres. Therefore, the apparent porosity is energy. Also, because the small crystallite higher not well related to the compressive strength in the case of capacity to deform, these fibres are comp highly graphitized fibres Nakatani et al. [4] also reported degree of deformation until the final cat failure that longitudinal compressive strength is determined by the without any indication of crystallite buckling on the ore size rather than the number of pores per unit fibre sur volume Considering the failure mechanism which involves The effect of large needle-like pores in HM fibres can be crystallite buckling due to insufficient lateral support of he figure del pores, it might be possible to say that compressive mplified skin-core model of longitudinal cross-section in properties in carbon fibres can be improved by infiltrating HM fibre structure, highly oriented large crystallites a substance into the pores by means of a post-treatment to enclose large needle-like pores in the outer layers, carbon fibres. Dobb et al. [24] attempted to deposit silver relatively random structure with many small pores exists in sulphide into aramid fibres to infiltrate the internal pores the core region [17. There may be misoriented crystallites using a conventional method, which was originally de- surrounding the pores, and such crystallites suffer from veloped to utilise the silver sulphide as an indicator to concentrated shear stress. Moreover, crystallite layer reveal the pore distributions when sections were examined planes oriented along the fibre axis also contribute to the the transmission electron microscope (TEM)[25]. Very failure when they have insufficient lateral supports nearby interestingly, such treated fibres exhibited a remarkable the elongated pores. Therefore, the initiation of failure inprovement in longitudinal compressive strengths by mechanism is a crystallite buckling near the localised 0% compared to untreated fibres when tested by the pores, and consequently, shear between basal planes. With tensile recoil measurement [26]. It was concluded that such increase of the modulus in HM fibres, the local crystallite an improvement may be due to either an increase in the buckling is more likely to occur at lower stress level effective cross-sectional area caused by infiltration or an because of the longer and larger pores, and the failure actual reinforcing effect of the epi ially grown silver epeatedly occurs and propagates through the large struc- sulphide crystals. If such a technique is applied for PAN- tural continuity as multiple kinks, they eventually become based carbon fibres, especially for HM fibres, the large ible on the lateral surface at final failure as seen in the needle-like pore may be effectively filled to prevent any in-situ microscopic studies local buckling of adjacent crystallites; which hopefully A Also in HS fibres, compressive failure initiates when a produces high performance carbon fibres with high com- lure occurs from crystallites near small pores. However, pressive strength and modulus these fibres have almost the same level of modulus. and there is no significant difference in pore size between fibre variants. Therefore, the mean porosity well refects the 5. Conclusions difference of longitudinal compressive strength in these fibres. HS fibres generally show higher resistibility to A direct compression device was developed to be compressive deformation owing to the high amount of applied to single carbon fibres in this study. The measured disorder which effectively dissipates the crack propagation longitudinal compressive strength was generally higher than that from the tensile recoil technique. The compres- sive strength was <30 to 50% of the tensile strength mpression Longitudinal compressive modulus was also estimated using the Euler buckling formula, the compressive modulus was -50% of the tensile modulus. The in-situ Fibre arge crystallites SEM observation was useful to observe compressive behaviour of carbon fibres HM fibres indicated distinct kink bands that usually developed into splitting failure along the fibre Structure-compressive property relationships revealed that compressive strength of Hs PAN-based carbon fibre is mainly influenced by the mean porosity. In HM fibres. 00 reduced disorder and enlarged crystallites appeared with Pores 000 Kink bands increase of the fibre modulus. Consequently, the pores tended to be large and needle -like which increases the possibility of localised crystallite buckling with insufficient lateral support. The result suggests that the effect of pore size Fig.12.compressivefailuremechanisminHmPaN-bAsedparticularlyimportantforcompressivestrengthinPan- carbon fibre based carbon fibres. It will be necessary to investigate the
644 N. Oya, D.J. Johnson / Carbon 39 (2001) 635 –645 strength in HM fibres. Therefore, the apparent porosity is energy. Also, because the small crystallites have a higher not well related to the compressive strength in the case of capacity to deform, these fibres are compressed to a higher highly graphitized fibres. Nakatani et al. [4] also reported degree of deformation until the final catastrophic failure that longitudinal compressive strength is determined by the without any indication of crystallite buckling on the pore size rather than the number of pores per unit fibre surface. volume. Considering the failure mechanism which involves The effect of large needle-like pores in HM fibres can be crystallite buckling due to insufficient lateral support of envisaged as illustrated in Fig. 12. The figure depicts a pores, it might be possible to say that compressive simplified skin-core model of longitudinal cross-section in properties in carbon fibres can be improved by infiltrating HM fibre structure; highly oriented large crystallites a substance into the pores by means of a post-treatment to enclose large needle-like pores in the outer layers, a carbon fibres. Dobb et al. [24] attempted to deposit silver relatively random structure with many small pores exists in sulphide into aramid fibres to infiltrate the internal pores the core region [17]. There may be misoriented crystallites using a conventional method, which was originally desurrounding the pores, and such crystallites suffer from veloped to utilise the silver sulphide as an indicator to concentrated shear stress. Moreover, crystallite layer reveal the pore distributions when sections were examined planes oriented along the fibre axis also contribute to the in the transmission electron microscope (TEM) [25]. Very failure when they have insufficient lateral supports nearby interestingly, such treated fibres exhibited a remarkable the elongated pores. Therefore, the initiation of failure improvement in longitudinal compressive strengths by mechanism is a crystallite buckling near the localised 50% compared to untreated fibres when tested by the pores, and consequently, shear between basal planes. With tensile recoil measurement [26]. It was concluded that such increase of the modulus in HM fibres, the local crystallite an improvement may be due to either an increase in the buckling is more likely to occur at lower stress level effective cross-sectional area caused by infiltration or an because of the longer and larger pores, and the failure actual reinforcing effect of the epitaxially grown silver repeatedly occurs and propagates through the large struc- sulphide crystals. If such a technique is applied for PANtural continuity as multiple kinks; they eventually become based carbon fibres, especially for HM fibres, the large visible on the lateral surface at final failure as seen in the needle-like pore may be effectively filled to prevent any in-situ microscopic studies. local buckling of adjacent crystallites; which hopefully Also in HS fibres, compressive failure initiates when a produces high performance carbon fibres with high comfailure occurs from crystallites near small pores. However, pressive strength and modulus. these fibres have almost the same level of modulus, and there is no significant difference in pore size between fibre variants. Therefore, the mean porosity well reflects the 5. Conclusions difference of longitudinal compressive strength in these fibres. HS fibres generally show higher resistibility to A direct compression device was developed to be compressive deformation owing to the high amount of applied to single carbon fibres in this study. The measured disorder which effectively dissipates the crack propagation longitudinal compressive strength was generally higher than that from the tensile recoil technique. The compressive strength was |30 to 50% of the tensile strength. Longitudinal compressive modulus was also estimated using the Euler buckling formula; the compressive modulus was |50% of the tensile modulus. The in-situ SEM observation was useful to observe compressive behaviour of carbon fibres; HM fibres indicated distinct kink bands that usually developed into splitting failure along the fibre axis. Structure–compressive property relationships revealed that compressive strength of HS PAN-based carbon fibre is mainly influenced by the mean porosity. In HM fibres, reduced disorder and enlarged crystallites appeared with increase of the fibre modulus. Consequently, the pores tended to be large and needle-like which increases the possibility of localised crystallite buckling with insufficient lateral support. The result suggests that the effect of pore size is Fig. 12. Compressive failure mechanism in HM PAN-based particularly important for compressive strength in PANcarbon fibre. based carbon fibres. It will be necessary to investigate the
