Professional Documents
Culture Documents
Chawla (2005)
Chawla (2005)
Journal
Monotonic and Cyclic Fatigue Behavior of High-Performance
Ceramic Fibers
N. Chawlaw and M. Kerr
Department of Chemical and Materials Engineering, Arizona State University, Tempe, AZ 85287-6006
K. K. Chawla*
Department of Materials Science & Engineering, University of Alabama at Birmingham, Birmingham, AL 35294
Monotonic and cyclic fatigue behavior of single fibers or fiber ing. In order to obtain a better understanding of the fatigue
fabrics are of significant interest, since fiber assemblies or fiber- behavior of various composites, it is of great interest to study the
reinforced composite materials in structural applications are behavior of reinforcement fibers subjected to cyclic fatigue
often subjected to cyclic loading. Studying the cyclic fatigue conditions.
behavior of fibers is particularly difficult because of their small Relatively few studies have been conducted on the cyclic
diameter (B10 lm) and high aspect ratio. In this paper, we fatigue behavior of high-performance fibers.8–12 Most of these
report results of monotonic tension and tension–tension fatigue studies have examined the stress versus cycles (S–N) behavior of
behavior of two sol–gel-derived ceramic fibers: Al2O3–SiO2– the fibers. Bunsell and Somer8 examined the fatigue behavior
B2O3 (Nextel 312) and Al2O3 (Nextel 610). Nextel 312 ex- of carbon fibers and reported no significant change in tensile
hibited a great deal of variability in tensile strength, reflected by strength after cyclic loading. In fact, they observed a slight
a Weibull modulus of 4.6, versus Nextel 610, which had a increase in fiber modulus with increasing cycles, which was
Weibull modulus of 10.5. Our experiments showed clearly that attributed to increasing orientation of the fibrils in the turbos-
cyclic loading was more damaging than static loading and, thus, tratic structure of graphite fibers. Minoshima et al.9 studied the
resulted in a lower cyclic fatigue life compared with static cyclic behavior of aramid fibers (Kevlar), and indicated that
loading. The fracture behavior under fatigue loading was dis- while inelastic deformation took place due to fatigue, the S–N
tinctly different from that under monotonic loading. It is behavior was inherently dominated by nodular flaws at the
believed that processing-induced flaws acted as crack initiation surface of the fibers. Other studies on the cyclic deformation
sites, and that the cyclic loading induced subcritical cracking, behavior of natural (polymeric) fibers indicate a cyclic fatigue
followed by coalescence of cracks immediately prior to failure. effect, with fracture taking place by microvoid nucleation and
coalescence.12 In glass fibers, static fatigue plays a role in time-
dependent failure, but this is attributed mainly to environmental
effects such as moisture absorption.13,14 In this paper, we
I. Introduction
examined the monotonic and fatigue behavior of two high-
101
102 Journal of the American Ceramic Society—Chawla et al. Vol. 88, No. 1
Nextel 62% Al2O3, 24% SiO2, and 14% B2O3 10–12 2.7 9Al2O3–2B2O3 grains (B0.1 mm) and globular SiO2 (20
312 nm)
Nextel 99% Al2O3, 0.2–0.3% SiO2, 0.4–0.7% 10–12 3.75 Single phase a-Al2O3 grains (B0.1 mm)
610 Fe2O3
and proper alignment of the fibers was ensured using an optical was measured using a scanning electron microscope (SEM). The
microscope. A gage length of 20 mm was used for all tensile and average fiber diameter and standard deviation were similar to
fatigue testing. A microforce testing system (MTS Systems that reported by Wilson and Visser,18 who made measurements
Corp., Eden Prairie, MN) was used to conduct tensile and of the fiber diameter for Nextel 610 and determined that the
fatigue testing of the ceramic fibers (Fig. 2). The load resolution coefficient of variance in fiber diameter was approximately
achieved with this system was on the order of 0.001 N and 1.8%. Thus, an average fiber diameter, measured at around 11
displacements on the order of 1 mm was achieved with this mm, was used to calculate the cross-sectional area and stress.
system. Tensile testing was conducted in displacement control at Recently, Lara-Curzio and Russ19 have shown that assuming a
a strain rate of 103 s1. Fatigue testing was conducted in stress constant cross-section for all fibers may introduce a significant
control, tension–tension with R 5 0.2 (smin/smax), at a fre- error in Weibull modulus, particularly when the variability in
quency of 2 Hz. Figure 3 shows a typical sinusoidal waveform fiber diameter is quite high. Using the model of Lara-Curzio and
of force versus time measured from the microforce testing Russ,19 Wilson and Visser18 predicted a reduction in Weibull
system. Note the high accuracy and control of the applied modulus of only 5%. Thus, since the variability in fiber dia-
loading. All testing was conducted at ambient temperature meters is quite small, using average fiber diameter in computing
(B281C). All fibers tested, including those that failed on load- Weibull modulus appears to be a reasonable assumption for the
ing, are included in the data set, i.e., no screening of fibers was fibers studied here.
conducted.
The compliance of the loading and gripping system was
quantified by obtaining the force versus displacement behavior 0.15
of the fiber at various gage lengths.1,17 The cross-head displace-
ment during fiber testing, dt, can be expressed by:
dt 1
¼ ‘þc ð1Þ
F EA 0.1
Force (N)
0
0 0.5 1 1.5 2
Time (s)
Fig. 3. Sinusoidal waveform used for fatigue testing. Note the high
resolution and control of the applied load with time.
5
Fig. 1. Schematic of the paper frame technique used for gripping and
testing of single fibers. Al2O3-SiO2-B2O3 (Nextel 312)
Displacement/Force (mm/N)
3 1
EA
1
Al2O3 (Nextel 610)
0
0 10 20 30 40 50 60
Gage length (mm)
Fig. 4. Normalized displacement versus gage length plot that yields a
linear relationship with slope 1/(EA). The compliance of the load train,
Fig. 2. Microforce testing system used for fiber testing. given by the intercept, can also be determined from this plot.
January 2005 Monotonic and Cyclic Fatigue of Ceramic Fibers 103
Table II. Tensile Properties of Nextel Fibers upon fracture. Other researchers have used similar procedures
with no noticeable adverse effects on fiber strength.18
Fiber Young’s modulus Weibull modulus Tensile strength
(GPa) (m) (MPa)
data
1500 larger displacement than the true displacement of the fiber.
Measurement of the modulus from the slope of the corrected
stress–strain curve yielded results similar to those obtained from
1000 the displacement/force versus gage length approach.
Nextel fibers exhibited variability in tensile strength charac-
teristic of ceramic fibers. Weibull statistics were used to rank the
500 relative fiber strength versus probability of failure of the fibers to
obtain a measure of the variability in fiber strength.1,23,24
According to the Weibull analysis, the probability of survival
0 of a fiber at a stress, s, is given by the following relation:
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Strain (%) m
s
Fig. 5. Stress–strain behavior of Nextel 312, showing the ‘‘as-mea- PðsÞ ¼ exp ð2aÞ
s0
sured’’ data and that corrected for load train compliance.
where s is the fiber strength for a given probability of survival,
10 and m is the Weibull modulus.
Al2O3-SiO2-B2O3 (Nextel 312)
s0 is defined as the characteristic strength, which corresponds
Relative Rank ln [(N+1)/(N+1-i)]
1 i
PðsÞi ¼ 1 ð2bÞ
Nþ1
Fig. 7. Fiber fracture surface of Nextel 312 after tensile loading showing mirror–mist–hackle fracture morphology. Nodules on the fiber surface are a
result of grease used to dissipate fracture energy and retrieve fibers for fractographic analysis.
104 Journal of the American Ceramic Society—Chawla et al. Vol. 88, No. 1
Fig. 8. Fiber fracture surface after tensile loading. The polycrystalline nature of the fiber results in a rougher fracture surface than that of Nextel 312.
Note the fine grain size (on the order of 100 nm).
tensile strength reported here is somewhat higher. This can be which corresponds to the approximate stress below which the
attributed to our use of the average fiber diameter, and the probability of failure is zero, is clearly much higher for Nextel
smaller gage length used in our study (20 vs 25.4 mm in Wilson 610 than for Nextel 312.
and Visser18). Thus, a lower volume was tested and, thus, The tensile fracture surface of the fibers exhibited three
the probability of encountering a strength-limiting flaw also distinct regions. Figure 7 shows a typical tensile fracture surface
decreased. of Nextel 312. A smooth initiation region or fracture mirror is
In comparing Nextel 312 and Nextel 610, the latter fiber observed at the fiber surface. Surface flaws are more highly
exhibited a much lower variability in tensile strength. Variability stressed than defects in the interior, making a surface flaw a
in fiber strength may be attributed to the variability in fiber flaw more likely site for crack initiation.25 Small diameter fibers have
size, in the form of porosity or ‘‘weld-lines’’ where two fibers are a relatively large surface-to-volume ratio, which makes surface
bonded together during processing.18 Presumably, the smaller flaws on fibers more critical in controlling fiber strength. The
average flaw size and narrower distribution of flaws are respon- crack nucleation site is followed by a transition region or ‘‘mist,’’
sible for the higher strength and higher Weibull modulus and the rapid crack propagation region or ‘‘hackle.’’ The
observed in Nextel 610, than in Nextel 312. The threshold stress, fracture behavior illustrated here is quite characteristic of brittle
1800
Al2O3-SiO2-B2O3 1800
1700 Failed on
Nextel 312 1700 loading
Failed on
Maximum stress (MPa)
1600 1600
loading
1500
Stress (MPa)
1500 n = 82
1400
1400
1300
1300 1200
5000
Al2O3-Nextel 610 Failed on
Failed on 4000 loading (2)
4000 loading
Maximum stress (MPa)
3000 n = 55
Stress (MPa)
3000
2000
2000
1000
Al2O3-Nextel 610
1000
100 101 102 103 104 105 106 107 100 101 102 103 104 105 106
(b) Nf, Cycles to Failure (b) Time to failure (s)
Fig. 9. Stress versus cycles behavior of (a) Nextel 312 fibers and (b) Fig. 10. Stress versus time to failure of (a) Nextel 312 fibers and (b)
Nextel 610 fibers. The fibers show increasing fatigue life with decreasing Nextel 610 fibers. Static stress exponents of 73 and 55 were obtained for
stress. Nextel 312 and Nextel 610, respectively.
January 2005 Monotonic and Cyclic Fatigue of Ceramic Fibers 105
1800 should be similar. The static fatigue stress, ss, can be related to
Cyclic the time to failure, ts, by the static fatigue exponent, n:
Cyclic (predicted)
1700 Static
ts ¼ K ðssÞn ð3Þ
Maximum stress (MPa)
1600
The analysis of static fatigue data, shown in Fig. 10, indicated n
1500
values for Nextel 312 and Nextel 610 of 82 and 55, respectively.
These values compare well with n values for bulk Al2O3, which
can range between 44 and 78.28,29
1400
A comparison of cyclic and static fatigue lives is complicated
by the fact that under cyclic loading less time is spent at
1300 maximum stress per unit time. Thus, in order to compare
adequately the time to failure under static fatigue, ts, and cyclic
1200 Al2O3-SiO2-B2O3 fatigue, an effective time to failure under cyclic fatigue, tc, must
Nextel 312 be computed. Evans and Fuller30 proposed that the ratio of tc
1100 and ts can be written in terms of the cyclic waveform (sinusoidal,
100 101 102 103 104 105 106 107 trapezoidal, square, etc.) as a function of time. Thus, for a
(a) Time to failure (s) sinusoidal waveform:
5000 8 l 91
Z
Cyclic
Cyclic (predicted)
tc < 1 sðtÞ n =
¼ dt ð4aÞ
Static t s :l ss ;
4000 0
Maximum stress (MPa)
Fig. 12. Fiber fracture surfaces of Nextel 312 after fatigue loading. Note the rougher nature of fracture and pronounced serration-like features on the
fracture surface of fatigued fibers.
indeed taking place in the fibers. Fractographic analysis of the face roughness as a means to differentiate between monotonic
fracture surfaces of several samples after fatigue was conducted and cyclic loading are more difficult in Nextel 610.
to elucidate the fatigue damage mechanisms, and contrast these From the fractographic analysis it is possible to conclude that
to those observed under tensile loading. All the cyclic fatigue crack initiation in the fibers takes place at pre-existing flaws in
fracture surfaces of the Nextel 312 fibers were quite different the material. Thus, the size and distribution of flaws will affect
from those observed after tensile loading. Figure 12 shows crack the onset of crack initiation during cyclic loading. We believe
initiation at the fiber surface, followed by what appears to be a that during cyclic fatigue of the fibers, fibers with large flaws
stable crack, due to cyclic loading, that stemmed from the crack failed immediately upon loading in the first cycles, while fibers
initiating defect. The fast-fracture region of the fiber is also with smaller flaws had much longer fatigue lives. These smaller
rougher in character than that observed in tension (Fig. 7) with flaws likely served as crack-initiating defects for subcritical crack
more evidence of damage associated with fatigue microcracks. growth, which propagated until a critical crack size was formed
Figure 13 shows a representative fatigue fracture surface of leading to unstable fracture. The same behavior was observed in
Nextel 610. Crack initiation in this particular fiber appears to cyclic fatigue of monolithic yttria-stabilized zirconia.31 Thus,
have started at an internal flaw. As observed for the tensile our data indicate that cyclic damage has a significant role in
fracture surface of Nextel 610, the 0.1 mm-size grains of the fiber nucleating subcritical damage at fatigue strength-limiting flaws.
give it a rougher appearance than the nanometer-sized grains of Static fatigue life on the other hand, similar to monotonic tensile
Nextel 312. Thus, qualitative comparisons of the fracture sur- loading, appeared to be controlled by critical flaw size, above
which instantaneous failure took place, and below which the
fiber did not fail. A similar observation has been made by Hulm
and Evans31 in their study of yttria-stabilized zirconia.
We can gain a better understanding of cyclic damage in the
fibers by examining the work on cyclic fatigue of bulk ceramics.
Several mechanisms have been proposed for fatigue crack
growth in monolithic ceramics, and these have been summarized
by Horibe.15 Cyclic crack growth in ceramics is known to occur
predominantly in an intergranular fashion. One mechanism for
the intergranular crack growth is dependent on frictional wear
mechanisms between grains.32–35 ‘‘Grain bridging’’ contributes
to crack growth resistance, but continued cycling results in
debonding between the grains. The contribution from grain
boundary bridging is very small at very small grains sizes,
such as those in the fibers tested here.34 Crack growth may
also be affected by residual glassy phase between grains. Ewart
and Suresh32 showed that pure Al2O3 with very little glassy
phase did not exhibit subcritical crack growth. Al2O3 with some
amount of glass phase, however, did exhibit stable crack growth.
The Nextel 312 fiber has a substantial amount of glassy phase
Fig. 13. Fiber fracture surface after fatigue loading of Nextel 610. The between grains, while Nextel 610 has a negligible amount of
polycrystalline nature of the fiber results in a rougher fracture surface glassy phase. Both fibers were susceptible to cyclic fatigue, so
than that of Nextel 312. Thus, qualitative comparisons of differences in that the presence of a glassy phase does not appear to be
the roughness of fracture between tensile and fatigue are not as clear. a necessary condition for cyclic crack propagation. This is
January 2005 Monotonic and Cyclic Fatigue of Ceramic Fibers 107
supported by the work of Horibe and Hirahara36 on reaction- (3) It is believed that fatigue damage initiated at pores or
bonded silicon nitride, a polycrystalline material with no grain defects in the fiber. Fatigue crack growth likely occurred by
boundary glassy phase between the grains, where cyclic crack crack arrest and branching mechanisms. This is supported by
propagation occurred in intergranular fashion. fractographic analysis of fatigued fibers, which indicated a
Another mechanism for cyclic crack growth in monolithic significantly tortuous fracture surface under cyclic loading,
ceramics, such as Si3N4, involves crack deflection and branching compared with monotonic loading. This can be attributed to
between grains. The crack propagates between the grains and subcritical damage, in the form of microcracks formed during
may encounter a grain that arrests the crack until a sufficient the cyclic process, which are linked in the final stages of fatigue.
driving force is available to grow the crack. In the fibers studied (4) Design of fiber reinforced composites or other applica-
here, we do not have definitive experimental evidence for the tions where high-performance fibers are incorporated must take
precise mechanism of fatigue crack nucleation and growth. into consideration the cyclic fatigue behavior of the fibers.
Nevertheless, we postulate that cyclic crack growth in the
ceramic fibers studied here is most likely occurring by a mechan-
ism of crack deflection and branching between grains. It is Appendix A
envisioned that the cracks initiate at pores or defects at the
In this appendix, we describe the solution of the relation for the
surface or in the interior of the fiber. Because of the extremely
ratio of cyclic time to failure to static time to failure, tc =ts given
fine grain size, the cracks must branch and deflect through the
in Eq. (4f ), in terms of the phase angle y:
grain boundaries. The fracture surfaces of the fibers after fatigue
support this hypothesis. The much rougher fracture surface 8 l 9
Z n =1
observed after fatigue, compared with monotonic tensile load- tc < 1 1þR 1R 2pt
ing, indicates several nucleation sites for subcritical crack ¼ þ sin dt
t s :l 2 2 l ;
growth and coalescence of these cracks appears to have occurred 0
prior to fatigue fracture. When a large fatigue crack is formed at 2pt
the surface and propagates through the fiber, interlinking setting y ¼ ; ðA-1Þ
l
between the dominant fatigue crack and the smaller fatigue 8 91
cracks within the volume of the fiber may occur. This interlink- <1 þ R Z2p =
tc 1R 1
ing may result in a fracture surface that is quite tortuous, as ¼ þ ½sinn ydy
shown in Fig. 12. The crack growth may be accentuated by the ts : 2 2 2p ;
0
accumulation of debris and localized wear between the crack
faces, which has been shown to decrease the resistance to cyclic We can designate the integral term in the above equation by h:
crack growth.36 The debris may also cause wedging and loca-
lized mismatch between crack faces, which may result in rough- Z2p
ness-induced closure during fatigue.32,37 It should be noted that 1
h¼ ½sinn ydy ðA-2Þ
stress corrosion cracking (SCC) can be accelerated during cyclic 2p
loading of ceramics due to moisture, but cyclic crack growth has 0