You are on page 1of 14

The influence of modified intermetallics and Si particles on fatigue

crack paths in a cast A356 Al alloy


K . G A L L , N . YA N G , M . H O R S T E M E Y E R , D. L . M c D O W E L L 1 a n d J . FA N 1
Materials & Engineering Sciences Center, Sandia National Laboratories, Livermore, CA 94550, USA, 1 GWW School of Mechanical Engineering,
Georgia Institute of Technology, Atlanta, GA 30332, USA

Received in final form 16 July 1999

A B S T R A C T Mechanical fatigue tests were conducted on uniaxial specimens machined from a cast
A356-T6 aluminium alloy plate at total strain amplitudes ranging from 0.1 to 0.8%
(R=−1). The cast alloy contains strontium-modified silicon particles (vol. fract. ~6%)
within an Al–Si eutectic, dispersed a intermetallic particles, Al15 (Fe,Mn)3 Si2 (vol. fract.
~1%), and an extremely low overall volume fraction of porosity (0.01%). During the
initial stages of the fatigue process, we observed that a small semicircular fatigue crack
propagated almost exclusively through the Al–1% Si dendrite cells. The small crack
avoided the modified silicon particles in the Al–Si eutectic and only propagated along
the a intermetallics if they were directly in line with the crack plane. These growth
characteristics were observed up to a maximum stress intensity factor of ~K max tr
=
1/2
7.0 MPa m (maximum plastic zone size of 96 mm). When the fatigue crack propagated
with a maximum crack tip driving force above 7.0 MPa m1/2 the larger fatigue crack tip
process zone fractured an increased number of silicon particles and a intermetallics
ahead of the crack tip, and the crack subsequently propagated preferentially through
the damaged regions. As the crack tip driving force further increased, the area fraction
of damaged a intermetallics and silicon particles on the fatigue fracture surfaces also
increased. The final stage of failure (fast fracture) was observed to occur almost
exclusively through the Al–Si eutectic regions and the a intermetallics.
Keywords cast Al–Si alloy, fatigue crack propagation, silicon and intermetallic
particles.

NOMENCLATURE a=crack length


af =final crack length
aMSC =crack length of a microstructurally small crack
atrans =transition crack length
d=gauge section diameter
E=elastic modulus
f ( g)=geometric factor
Kmax =maximum applied stress intensity factor
tr
K max =transition maximum applied stress intensity factor
rp =plastic (process) zone size
R=load ratio, smin /smax
a=Al15 (Fe,Mn)3 Si2 intermetallic
DKeff =effective stress intensity range
DKeff, th =effective threshold stress intensity range

De=applied far-field strain range


smax =maximum applied far-field stress
smin =minimum applied far-field stress
sy =yield strength

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172 159
160 K. GALL et al.

containing well-dispersed and spherodized particles by


INTRODUCTION
alloying with trace amounts of strontium or sodium, or
Depending on their size, microstructural inhomogeneit- subjecting the casting to extended solutionizing heat
ies can have a strong influence on the fatigue properties treatments. Cast Al–Si alloys with modified silicon par-
of cast aluminium–silicon alloys. Previous research ticles generally exhibit a better resistance to long fatigue
efforts have demonstrated that if large casting pores crack growth compared to unmodified alloys.14–17
(&100 mm) are present, they control the fatigue life of Even though silicon particles play a key role in the
the cast samples.1–13 When fatigue cracks initiate from fatigue crack growth characteristics of cast Al–Si alloys,
the inner surfaces of large pores, the effective crack silicon particles are not the only second phase constituent
length (pore size+initial crack length) is long with common to this alloy system. Commercial Al–Si alloys
respect to other microstructural features, e.g. the den- often contain trace amounts of iron (%1%) as an
drite cell size and silicon particle diameter. Consequently, impurity or an alloying element added to prevent die
numerous fracture mechanics approaches based on aver- soldering for castings solidified under high pressures.18
age material properties have been successfully used to Although the iron circumvents the die soldering, it also
predict the fatigue life of cast aluminium alloys contain- readily combines with aluminium and silicon to form
ing relatively large pores.3,4,6,7,11–13 All of the referenced intermetallic flakes that are detrimental to the fatigue
models are based on semiempirical approaches that life of cast aluminium.19 To rectify the loss in mechanical
estimate fatigue life by integrating a Paris law crack properties due to the brittle iron intermetallic flakes,
growth rate equation. Some of the models also use elements, e.g. beryllium or manganese are further added
empirical equations to estimate the initiation life of the to the melt. It has been demonstrated that the cyclic20
fatigue cracks, thus accounting for incubation periods and monotonic21–23 mechanical properties of cast Al–Si
as cracks nucleate from defects.4,12 Other research- alloys are improved by the subsequent addition of
ers3,6,7,11,13 assumed the initial crack size to be approxi- manganese or beryllium. These elements promote the
mately equal to the initial defect size. Another distinction formation of compounded intermetallic phases, e.g. the
between the different fatigue life prediction models is a intermetallic Al15 (Fe,Mn)3 Si2 .20 The manganese-
related to the parameter used to characterize the crack modified a intermetallics have an intricate branched
tip driving force. Nearly all of the models use a transient- structure and thus are sometimes referred to as the
closure modified stress intensity factor,3,4,6,7,12 while one Chinese script phase.21 Alternatively, the beryllium-
uses a so-called strain intensity factor.11,13 Notably, the modified iron intermetallics possess a nearly spherical
latter parameter has more physical meaning when the shape.22
fatigue cracks are expected to violate small-scale yielding Improvements in the mechanical properties of cast
conditions. Al–Si alloys have been demonstrated by modifying the
The semiempirical models predict several experimen- intermetallics and silicon particles, but the explicit role
tally observed fatigue phenomena in cast alloys, e.g. the of the two coexisting phases on the local fatigue mechan-
effect of maximum pore size,3,6 mean stress effects,3 isms has not been elucidated. One reason that the
short crack and pore interactions,4 underload effects,11 influence of iron intermetallics has not received strong
and notch size effects.13 The models rely primarily upon attention is because some research-grade materials do
a Paris law crack growth exponent determined from not contain significant amounts of iron. However, indus-
fatigue crack growth tests in, e.g. compact tension speci- trial castings will invariably contain iron that was inten-
mens. However, continual advances in casting technology tionally added to avoid die soldering or inadvertently
are decreasing the maximum defect sizes found in indus- present from extracting the pure aluminium from the
trial cast components. In parallel with these materials ore. Therefore, it is important to understand the poten-
processing developments, it is imperative to further tial role of iron intermetallics in addition to modified
understand the micromechanisms of fatigue crack growth silicon particles on the local fatigue mechanisms of cast
over multiple length scales. This knowledge is required Al–Si alloys.
because fracture mechanics approaches may underesti- With this background, the primary aim of the present
mate fatigue lives as initial defect sizes reach the length work is to study the microstructural paths chosen by
scale of other microstructural features that are distributed fatigue cracks at different stages during the fatigue
throughout the material. In an initial effort to understand process. The observed changes in the local crack growth
fatigue crack growth mechanisms in cast Al–Si alloys, mechanisms will be correlated to changes in the crack
several studies have shown that the growth rate of long tip driving force by using fracture mechanics concepts.
cracks is dependent on the silicon particle morphology Such detailed information can be used to guide the
in an average sense.14–17 The silicon particle morphology development of multiscale fatigue models that incorpor-
can be modified from a coarse lamellar structure to one ate the spatial distribution of defects and local defor-

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
M I C R O S TR U C T U R A L I N H O M O G E N E I TI E S I N A l A L LOY 161

mation mechanism.24 In addition, information on the fatigue tests, the machine feedback mode was switched
micromechanisms of fatigue crack growth is useful for to load control and the frequency was increased to
designing commercial cast alloys to be more resistant to 10 Hz. Local fatigue mechanisms were evaluated by
fatigue crack nucleation and propagation. The present examining fracture surfaces with a scanning electron
study will not consider alloys containing unmodified microscope (SEM) using either secondary electron imag-
silicon particles as it is well established that the fatigue ing (SEI) or backscatter electron imaging (BEI). When
properties of these alloys are relatively poor, particularly information containing the spatial arrangement and area
at high stress ratios.14–16 In addition, it is a widely fraction of a particular phase was required, BEI was
accepted industrial practice to modify cast Al–Si alloys.18 used. BEI differs from SEI in that contrast depends on
the atomic mass of the constituent elements of a particu-
lar phase. Because of this, phases comprised of higher
EXPERIMENTAL TECHNIQUES
atomic mass elements create more backscatter electrons
The nominal compositions of typical A356 aluminium and thus appear brighter on the SEM image. In the
alloys are presented in Table 1. To produce A356 cast current material, iron has a much higher atomic weight
plates for the present study, a high-grade A356.2 ingot versus aluminium or silicon, thus it appears brighter in
(Table 1) was melted in an induction furnace. The melt BEI. Because aluminium and silicon have comparable
was grain refined with titanium–boron, strontium modi- atomic weights they appear similar in BEI. Consequently,
fied, and degassed using a rotary degasser. The castings the chemical compositions on the samples were measured
were poured under the influence of gravity between 950 using an energy dispersive X-ray (EDX) analysis. The
and 977 K in room temperature air, and then fully cooled EDX analysis is a semiquantitative measurement of the
over a 16-h period. The A356 aluminium plates were relative composition of the area of interest. The EDX
cast in rectangular moulds with interior dimensions of measurements are not used to characterize the exact
25–14–5 cm. Iron chills were employed on the top, composition of the alloy, but rather to approximate the
bottom and end of the casting mould to simulate a relative differences in the area fraction of different phases
permanent mould casting. A no-bake silica sand was between fractured samples and undeformed cross-sec-
used to create the sides of the plate, the riser, and the tions. In addition, X-ray mapping was used to determine
down sprue. A ceramic foam filter was used between the the spatial arrangement of silicon particles on one par-
down sprue and the riser. The plates were removed from ticular fracture surface.
the mould and then heat treated with a T6 ageing
treatment (solutionized at 810 K for 16 h, quenched in
EXPERIMENTAL RESULTS
hot water at 344 K, and then aged for 4 h at 518 K).
Standard ASTM low-cycle fatigue specimens were
Mechanical fatigue tests
machined from the ends of the plates where porosity
levels were minimized (0.01% average porosity). The The fatigue life curve based on total strain amplitude
gauge length of the specimens was 25.4 mm and the for the cast A356 aluminium alloy is shown in Fig. 1.
gauge section diameter was 9.53 mm. The average den- Even in this premium cast alloy, specimen lives demon-
drite cell size of the cast material in the region with low strate significant variability at a given applied strain
porosity was ~40 mm. The fatigue specimens were all amplitude. This is caused by the fact that the largest
given an axial polish to minimize machining surface defect in the gauge section controls the fatigue process
roughness effects on fatigue. Constant amplitude fatigue in cast materials. Due to the statistical nature of pore
tests were conducted in strain control at total strain nucleation and growth during solidification,25,26 the
amplitudes ranging from 0.1 to 0.8% (R=−1, fre- maximum pore size will vary between samples.
quency=0.5 Hz). After cyclic saturation in high-cycle Specimens containing a relatively large pore in the gauge

Table 1 Typical composition of A356


alloys, including the A356.2 alloy employed Others
for the present study
Material Al Si Mg Fe Mn Zn Ti Each Total

A356.0 Balance 6.5–7.5 0.25–0.45 0.20 0.10 0.10 0.20 0.05 0.15
A356.1 Balance 6.5–7.5 0.30–0.45 0.15 0.10 0.10 0.20 0.05 0.15
A356.2 Balance 6.5–7.5 0.30–0.45 0.12 0.05 0.05 0.20 0.05 0.15

The concentrations for Fe, Mn, Zn, Ti and other elements are maximum allowed values of
the Aluminium Association.

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
162 K. GALL et al.

Fig. 1 Total strain life curve for cast A356 Al. All samples were
cycled until failure with exception of the sample cycled at a 0.1%
strain amplitude.

section will demonstrate a severely reduced fatigue life Fig. 2 Backscatter SEM image of an arbitrary transverse cross-
as a nucleated fatigue crack will begin its growth as a section through an undeformed sample. The bright flakes are the a
very long effective crack. The pore size effect is partially Al15 (Fe,Mn)3 Si2 intermetallics, the small black circles are silicon
accounted for by fracture mechanics-based approaches3,6 particles, and the grey background is Al. The sample was etched
with a slightly acidic water solution to mark the silicon particles.
as a larger initial defect size dramatically decreases the
predicted crack propagation life. Furthermore, even in
specimens with negligible porosity, variability will occur the dark grey regions are Al–1% Si and the Al–Si
because of the statistical nature of crack interactions eutectic, the bright particles are a intermetallics
with silicon particles and other inclusions. Three failed (Al15 (Fe,Mn)3 Si2 ) and the black areas are valleys on the
specimens cycled at strain amplitudes of 0.2, 0.4 and fracture surface. We note that Al5 FeSi or so-called b
0.8% were used for fractographic analysis. intermetallics are also common to the present alloy
system.21 In nearly all cases, Mn was discovered in the
Fe inclusions using EDX analysis, thus we assert that
Fractography
the majority of the inclusions in the present alloy are a
A typical undeformed microstructure of the cast material intermetallics. Nevertheless, a small volume fraction of
is shown in Fig. 2. On this backscatter electron image, the b intermetallics surely exists.
the grey matrix is Al–1% Si, the small black circles are The crack initiation point was identified as a horse-
silicon particles, and the bright flakes are a intermetallic shoe-shaped aluminium oxide inclusion (confirmed with
particles.21 The compositions of these features were EDX) shown in Fig. 4. The fatigue crack initiating oxide
verified with local EDX measurements and found to be inclusion is relatively close to the specimen surface as
consistent with expectations. The large circular feature has been observed in previous fatigue studies on cast
on the lower right-hand side of the image is contami- Al–Si alloys.4,13 Aluminium oxide inclusions trapped
nation and not representative of the microstructure. during solidification often serve as nucleation sites for
Given the relative metallographic information for an shrinkage-assisted gas porosity.25,26 However, in this
arbitrary material cross-section, we will now examine particular specimen machined from the casting, a casting
the fracture surface features on the fatigued samples. Of pore surrounding the oxide inclusion was not present.
the samples fatigued at different strain amplitudes, the We note that even in the present alloy, which contained
fracture surface on the specimen cycled at 0.2% strain a low overall volume fraction of porosity (0.01%), fatigue
demonstrated the most obvious demarcation between a crack nucleation also occurred at relative casting pores
region that failed by progressive fatigue crack propa- in specimens with shorter fatigue lives (Fig. 1). However,
gation and one that failed by local tensile overload. the aim of the present study is to characterize the growth
Figure 3(a,b) shows low-magnification images of the behaviour of small cracks, thus only a specimen with a
overall fracture surface on the samples cycled at 0.2% relatively small oxide initiation site was carefully analysed
strain. Figure 3(a) was imaged using SEI while Fig. 3(b) in the SEM.
was imaged using BEI to reveal the area fraction and In the present study, the fatigue crack growth mechan-
spatial distribution of the a intermetallic. In Fig. 3(b), isms in two different stages will be elucidated. Although

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
M I C R O S TR U C T U R A L I N H O M O G E N E I TI E S I N A l A L LOY 163

(a)
Fig. 4 SEM image of the fatigue crack initiation point at an
aluminium oxide inclusion in the sample cycled at 0.2% strain.

will not change abruptly at the distinct crack fronts, but


rather undergo a gradual transition at a smaller length
scale. However, as there are significant changes in the
overall fracture surface appearance at low magnifications,
the two crack fronts drawn in Fig. 3(b) are meaningful
from a macroscopic point of view. The different growth
stages proposed within these regions are labelled in
Fig. 3(a), where FCGR stands for fatigue crack growth
rates. The justification for the labels in Fig. 3(b) will
now be discussed in detail using SEM and EDX results.
High-magnification images in the low FCGR region
reveals extended flat areas perpendicular to the loading
axis [Fig. 5(a)]. The corresponding backscatter image to
Fig. 5(a) shows that the area fraction of a intermetallics
in the low FCGR region [Fig. 5(b)] is only slightly
(b) higher than in the undeformed material cross-section
Fig. 3 Low-magnification SEM images of the overall fracture
(Fig. 2). At even higher magnifications, we found that
surface on the 0.2% sample taken in: (a) secondary electron mode; the flat surfaces in Fig. 5 are covered with fine fatigue
(b) backscatter electron mode. FCGR stands for fatigue crack striations [Fig. 6(a)]. In these regions, the very low
growth (specimen life=~150 000 cycles). volume fraction of silicon particles on the fracture surface
coupled with the overall planar appearance of the fracture
surface indicates that the fatigue cracks cut straight
it is difficult to partition the growth of the crack into through the Al–1% Si dendrite cells. Consequently, the
different stages using Fig. 3(a), the backscatter SEM cracks did encounter some a intermetallics, which are
picture in Fig. 3(b) more clearly identifies the transitions found randomly within the Al–1% Si dendrite cells
between different growth patterns. Black lines in [Fig. 5(b)]. When the crack encountered the a intermet-
Fig. 3(b) are used to indicate the crack positions corre- allics, it propagated preferentially along or through them
sponding to macroscopic transitions in the fracture sur- as evident from the slight increase in the area fraction
face morphology. The curvatures of the crack paths were of a in Fig. 5(b) versus Fig. 2. It should be noted that
drawn to match approximate contours of constant area wherever the area fraction of damaged a intermetallics
fraction of the a intermetallics. The positions of the on the fracture surface was low, the corresponding area
crack paths were chosen where the area fraction of a fraction of silicon particles was also relatively low. Thus,
intermetallics demonstrated the most obvious macro- the a intermetallics served as reliable indicators for the
scopic increases. Notably, the local fatigue mechanisms overall silicon particle distribution on the fracture sur-

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
164 K. GALL et al.

(a)

(a)

(b)
Fig. 5 Intermediate magnification SEM images of the fracture (b)
surface on the 0.2% sample near the fatigue crack nucleation site
taken in: (a) secondary electron mode; (b) backscatter electron Fig. 6 High-magnification SEM images of the fracture surface on
mode. the 0.2% sample near the fatigue crack nucleation site taken in:
(a) secondary electron mode; (b) backscatter electron mode.

face. The observations near the crack nucleation site in initiation site and becomes larger, the local crack tip
the 0.2% sample also hold true for the sample cycled at driving force increases as the maximum far field applied
0.4% strain as demonstrated in Fig. 7(a),(b). The strain (stress) amplitude is kept constant. An increase in
nucleation site in the 0.4% sample was also an oxide the local crack tip driving force results in higher (steady)
inclusion, however, the area of the fracture surface fatigue crack propagation rates. In the region of steady
covered by fatigue striations on the Al–1% Si was smaller FCGR, the area fraction of a intermetallics is noticeably
compared to the 0.2% sample. The sample cycled at higher compared to the regions where the cracks experi-
0.8% strain showed very little evidence of fatigue stri- enced low FCGR [Fig. 3(b)]. Furthermore, the gradual
ations on the specimen fracture surface. increase in local crack tip driving force is also
As the crack grows a significant distance from the accompanied by increases in the area fraction of silicon

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
M I C R O S TR U C T U R A L I N H O M O G E N E I TI E S I N A l A L LOY 165

tip encounters them. The silicon particles in the eutectic


provide more resistance to fatigue crack growth due to
their disperse and spherodized nature within the ductile
Al–1% Si phase.
When the crack tip driving force becomes large
enough, the remainder of the specimen fails through
tensile overload. Figure 8(a)–(c) shows representative
images taken from an overloaded region of the fracture
surface on the specimen cycled at 0.8% strain. The
fracture surfaces in the overloaded regions on the 0.4
and 0.2% specimens appeared identical to Fig. 8. The
overloaded regions [Fig. 8(a)] consist of an even mixture
of faceted and dimpled areas. To identify the phases
corresponding to the different types of fracture, images
of the region in Fig. 8(a) were captured using backscatt-
ered electrons [Fig. 8(b)] and silicon Ka X-ray mapping
[Fig. 8(c)]. Figure 8(b) shows the distribution of a inter-
metallics (the bright areas) in a typical overloaded region.
(a) The area fraction of the a intermetallics on the over-
loaded fracture surface is significantly larger than in the
undeformed material cross-section (Fig. 2), the low
FCGR region [Fig. 5(b)], or the steady FCGR region
[Fig. 3(b)]. By comparing Fig. 8(a) and (b), one observes
that the faceted regions correspond directly to fractured
a intermetallic particles. Figure 8(c) highlights the distri-
bution of silicon particles (the white specks) on the
region of the fracture surface that failed through tensile
overload. The area fraction of silicon particles on the
overloaded fracture surface is also significantly larger
than was found on the undeformed material cross-section
(Fig. 2) or the low FCGR regions [Fig. 6(b)]. Higher
magnification images of the different fractured phases in
the overloaded regions are shown in Fig. 9(a),(b). The a
intermetallics demonstrate a very flat appearance on the
fracture surface, which indicates that they fractured
through cleavage. It will be noted that the a intermet-
allics demonstrated brittle fracture characteristics
[Fig. 9(a)] in both the fatigued and overloaded regions.
The Al–Si eutectic regions are covered with dimples
(b) containing fractured silicon particles. Thus, the eutectic
Fig. 7 High-magnification SEM images of the fracture surface on regions failed primarily through particle cleavage and
the 0.4% sample near the fatigue crack nucleation site taken in: extensive ductile tearing.
(a) secondary electron mode; (b) backscatter electron mode.

EDX analysis
particles during steady FCGR. The fatigue cracks experi- To confirm and better quantify the above observations,
ence instantaneous extension through the a intermet- measurements of element concentrations were made on
allics as the particles showed clear evidence of cleavage the fracture surfaces and the arbitrary undeformed mate-
failure, as will be shown in subsequent micrographs. rial cross-section. If a phase area fraction is relatively
Such a brittle phase is not capable of promoting crack high on the fracture surface, it can be asserted that the
tip blunting and resharpening, a mechanism required for particle provides a weak link for material failure, regard-
the steady growth of fatigue cracks.27,28 Consequently, less of the local fracture mechanism. The measurements
the irregular shaped a intermetallics provide little or no on the undeformed cross-section (Fig. 2) provide a base-
resistance to fatigue crack propagation when the crack line comparison tool for EDX measurements on the

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
166 K. GALL et al.

(a)

(a)

(b)
(b)
Fig. 9 High-magnification SEM images of the fracture surface on
the 0.8% sample in a region that experienced tensile overload taken
in: (a) secondary electron mode; (b) backscatter electron mode.

fracture surfaces. Row 1 of Table 2 presents the measured


elemental concentrations on the arbitrary undeformed
material cross-section for a sampling area of ~4 mm2 .
The measured iron and manganese concentrations on
the surface of the undeformed sample are <1.0%. The
difference between this result and the nominal composi-

Fig. 8 SEM images of the fracture surface on the 0.8% sample in a


region that experienced tensile overload taken in: (a) secondary
electron mode; (b) backscatter electron mode; (c) X-ray mapping
mode.

(c)
© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
M I C R O S TR U C T U R A L I N H O M O G E N E I TI E S I N A l A L LOY 167

Table 2 Average element concentrations on


an arbitrary transverse cross-section surface Measured elemental Calculated approximate phase
and the specimen fracture surfaces concentrations (%) area fractions (%)

Sample Al Si Fe Mn Al–1% Si Si Particles a

Undeformed arbitrary cross-section 85.0 14.3 0.2 0.4 84.0 14.2 1.3
0.2% low FCGR region 86.8 11.8 1.0 0.4 81.8 11.1 6.7
0.2% overloaded region 67.8 25.1 4.9 2.2 43.3 21.8 32.7

EDX measurements are cumulative values taken at a magnification of 50×, covering large
specimen areas. Area fraction calculations assume that the a particles have a Al15 (Fe,Mn)3 Si2
composition.

tion of the alloy (Table 1) is of the order of the exper- the crack tip directly encounters them. This is evident
imental detection limitations (<1%) in the EDX from the slight increase in the area fraction of a in the
measurements. The silicon concentration in the unde- EDX calculations (Table 2) and the fatigued fracture
formed sample surface is twice what is expected, and this surface images [Fig. 2 versus Fig. 5(b)].
is primarily due to the peak overlap between the alu- The region that failed through tensile overload
minium and silicon spectra. The peak overlap will affect [Fig. 3(a)] demonstrates a greater area fraction of a and
all semiquantitative EDX calculations similarly, thus the silicon on the 0.2% fracture surface compared to the
relative differences in compositions between samples will undeformed cross-section (Table 2). The area fractions
still have comparative significance. Because all of the found in row 3 of Table 2 are consistent with SEM obser-
iron in the melt will readily form an intermetallic,18 the vations in regions that failed through tensile overloaded
area fraction of a particles, Al and silicon particles on all specimens (0.2, 0.4 and 0.8%). The increase of
expected on an arbitrary cross-section can be calculated silicon in the fracture surface is undoubtedly due to prefe-
from the EDX data. The results of these calculations rential fracture through the Al–Si eutectic. The pro-
are shown in bold in row 1 in Table 2. gression of fracture through the silicon particles in the
For the sample cycled at 0.2% strain, EDX measure- eutectic under large strain monotonic loading (i.e. an
ments were taken over 1-mm2 areas in the ‘low FCGR’ overload) is consistent with previous observations in modi-
region and ‘overloaded region’ indicated in Fig. 3(a). fied alloys.20,21,29,30 However, the progression of fracture
The last two rows in Table 2 present the results of the through the a intermetallics during a tensile overload has
EDX measurements on the 0.2% sample’s fracture sur- not received considerable attention. It has been previously
face. For calculations of the area fraction of different observed that iron intermetallics fracture during tensile
phases on the fracture surfaces, we assumed that all of loading,21 however, it was not quantitatively demonstrated
the iron forms the a intermetallic. The calculated phase that they play such a dominant role as reported here.
area fractions on the respective fracture surface regions According to the results presented in Table 2, the a inter-
are shown in bold in Table 2. The area fractions found metallics cover ~one-third of the fracture surface in an
in row 2 of Table 2 are consistent with visual SEM overloaded region. Considering the fact that the area
observations in regions that failed through progressive fraction of a intermetallics in the undeformed material
fatigue failure on all specimens (0.2, 0.4 and 0.8%). The cross-section is ~1% (Table 2), this increase is remarkable.
EDX measurements confirm that the fatigue cracks The fact that the a intermetallics appear this frequently
initially propagate straight through the Al–1% Si den- on the overloaded fracture surface compared to a random
drite cells as the arbitrary material cross-section and low cross-section decisively identifies them as weak microstruc-
FCGR region have comparable phase area fractions. tural links for damage progression during an overload. In
Because the overall area fraction of the silicon phase unmodified cast Al–Si alloys, the large volume fraction of
decreases in the fatigued region, the fatigue cracks coarse silicon particles in the eutectic controls the ductility
growing at low rates are effectively evading the modified of the alloy.21 However, in the present alloy, which contains
silicon particles. Furthermore, the fatigue cracks growing an optimally modified silicon particle microstructure, the
at low rates must be avoiding the eutectic phases altog- a intermetallics limit the macroscopic ductility of the
ether as a larger volume fraction of silicon particles on material. If the a intermetallics were not present, the
the fracture surface would be observed if they cut Al–1% Si dendrite cells would be capable of plastically
through the tortuous network of silicon particles in the deforming to very large strain levels, and the modified
Al–Si eutectic. However, fatigue cracks are expected to silicon particles in the eutectic would once again limit
propagate preferentially along the a intermetallics when the ductility.

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
168 K. GALL et al.

bulk material (Fig. 10). A crack is considered microstruc-


DISCUSSION
turally small if the crack length and crack tip process
Using the SEM and EDX results from the fatigued (plastic) zone exist at the same length scale as pertinent
fracture surfaces as a guide, a schematic of the pro- microstructural features. The crack tip process zone is a
gression of fatigue damage in cast Al–Si alloys is pre- region containing highly localized plasticity and high
sented in Fig. 10. As previously reported, the most triaxial stress levels. Research on low-porosity squeeze
dominant fatigue crack is envisioned to nucleate from cast Al–Si alloys has demonstrated that a crack experi-
the largest microstructural defect, e.g. an oxide inclusion ences microstructural small growth rate effects if its size
or casting pore near the specimen surface. The incu- is of the order of several dendrite cell sizes.31 Therefore,
bation period (crack nucleation life) of the formation of if the initial defect size is larger than several dendrite
such cracks in Al–Si alloys has only received approximate cell sizes, the crack will spend a negligible amount of
treatments in the literature. A local strain approach for time propagating as a microstructurally small crack. In
predicting fatigue crack nucleation life from large initial the specimen cycled at 0.2% strain, the crack-initiating
pores (average size of 500 mm) was utilized by Ting and oxide inclusion was nearly three times the dendrite cell
Lawrence on cast A319 aluminium.4 For smaller initial size. Consequently, the fraction of life spent while the
defects (Si particles and pores ranging from 10 to crack propagated in the microstructurally small regime
190 mm), Shiozawa et al.12 used a fracture mechanics- (a<aMSC ) is small or negligible.
based approach to predict the nucleation life of fatigue Even after the crack is no longer microstructurally
cracks. However, there is some uncertainty over what small, a>aMSC , it continues to propagate straight
defects will control fatigue crack nucleation when the through the Al–1% Si dendrite cells (Fig. 10). During
maximum pore size becomes less than ~100 mm. In the this growth period, the local fracture surface is rather
absence of large pores, it has been argued that fatigue flat and perpendicular to the tensile axis even at the
crack nucleation occurs from the largest silicon particles microscale. The advancing fatigue crack avoids the sili-
or clusters.12,31 Nevertheless, there exists conflicting con particle-rich eutectic and only propagates preferen-
evidence6 that fatigue cracks initiate from micropores tially through a intermetallic particles if they are in-line
on the specimen surface of the order of 12 mm despite with the crack plane. The fatigue crack propagation at
the fact that silicon particles as large as 15 mm were low rates through the Al–1% Si dendrite cells only
present. Consequently, more work is needed on the continues up to a definite crack length. At a transition
actual mechanisms of fatigue crack nucleation from crack length, atrans , the fatigued fracture surface begins
defects of varying size in cast Al–Si alloys. During high- to show increased evidence of damaged a intermetallics
cycle fatigue, the time spent forming cracks near defects and silicon particles (Fig. 10). The magnitude of atrans is
is critical as it will constitute a considerable fraction of dictated by the microstructure of the alloy and the crack
the total fatigue life in premium cast Al–Si alloys.6,12 tip driving force. The microstructural effect is charac-
After nucleation, the microstructurally small cracks terized by the propensity of the silicon and a particles
propagate towards the specimen surface and into the to crack/debond coupled with the nearest-neighbour

Fig. 10 Schematic illustration of the crack


paths and progression of fatigue damage in
cast Al–Si alloys containing defects of the
order of several dendrite cell sizes.

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
M I C R O S TR U C T U R A L I N H O M O G E N E I TI E S I N A l A L LOY 169

distance between the respective phases. The silicon mated in Eq. (1) are the appropriate measure of crack
particles are concentrated in a compact cell wall-type length and the geometry factor. Several works have
structure while the a intermetallics are evenly dispersed provided numerical solutions of the stress intensity fac-
over a much larger length scale. Thus, even if the brittle tors for semicircular or semielliptical flaws in a cylindrical
irregular shaped a intermetallics possess negligible resist- specimen subjected to tensile loading.36,37 Experimental
ance to cracking, they will not strongly affect the crack evidence38 suggests that a crack in a cylindrical specimen
propagation rates until the crack tip process zone is large grows with a semicircular crack front. As the crack
enough to sense a statistically significant number of grows, the front remains almost perfectly circular, how-
them. ever, the radius of curvature continually increases. The
It is useful to quantify the macroscopic transition approximate crack fronts marked by the constant area
point between dominant fatigue crack growth through fraction contours of a intermetallics [Fig. 3(b)] are
the dendrite cells versus the second phase particle extremely well defined using circular arcs. Consequently,
regions. The most widespread parameter used to charac- the numerical stress intensity solution for an expanding
terize the crack tip driving force during fatigue is the semicircular crack37 is used in this analysis:
effective stress intensity range, DKeff .34,35 However, in

AB AB
2 3
the present study the maximum stress intensity value, a a a
f (g)=0.67−1.24 +28.0 −162.4
Kmax , will be calculated and used as the parameter d d d
governing the fatigue crack path transition. The Kmax

AB AB AB
4 5 6
a a a
and DKeff parameters both have potential advantages for +472.2 −629.6 +326.1 (2)
analysing the transition behaviour. In general, DKeff may d d d
better characterize the fatigue crack driving force, how-
In Eq. (2), d is the specimen diameter and a is the
ever, Kmax may better characterize the propensity for the
maximum crack depth measured towards the centre of
a intermetallics and silicon particles to crack during
the specimen. It will be noted that Eq. (2) only deter-
fatigue. Consequently, Kmax will be used to quantify the
mines the stress intensity shape factor at the deepest
transition point between growth patterns through the
point in the crack. For an elliptical crack in a flat plate
Al–Si eutectic and a intermetallics versus through the
it is usually assumed that the stress intensity factor
Al–1% Si dendrite cells. Using the fracture surface
remains constant along the periphery of the crack, thus
observations, it is possible to estimate the value of the
requiring the crack aspect ratio to remain the same.
maximum stress intensity factor when the fatigue cracks
However, for a curved free surface, the crack aspect ratio
first begin to propagate more preferentially through the
has been observed to change as the crack moves towards
a intermetallics and silicon particles. The calculations
the centre of the specimen. Thus, a numerical model for
will be performed using the fracture surface markings
the stress intensity factor must allow the crack radius to
on the sample cycled at 0.2% strain. By definition, the
change. Furthermore, the stress intensity factor must be
transition stress intensity value that is determined from
correlated as a function of a/d as in Eq. (2). The details
the 0.2% sample is valid for other conditions (geometries
of the approach and determination of Eq. (2) are dis-
or applied stress) provided the cracks do not violate the
cussed more thoroughly by Liu.37 The maximum depth
small-scale yielding criteria. In Fig. 3(b) the smaller black
of the small crack in Fig. 3(b) is atrans =1.39 mm and
semicircle indicates the approximate crack shape and size
the specimen diameter is d=9.53 mm. Using these
corresponding to the point where the area fraction of
dimensions, Eq. (3) yields f ( g)=0.756 for the smaller
the a intermetallics begins to noticeably increase on the
crack in Fig. 3(b). It will be noted that an average value
fracture surface. At this transition point the crack also
of f ( g)=0.767 was obtained using the solution for an
begins to propagate more often through fractured silicon
elliptical crack36 in a round tension specimen [crack size
particles in the eutectic region. With this, the crack
of a=1.39, 2b=2.78 from Fig. 3(b)]. The numbers
position in Fig. 3(b) is defined a priori as the transition
differ by ~1%, and thus either approach is appropriate
length, atrans , for the 0.2% specimen. The generalized
for the analysis here. Substituting the calculated values
solution for the maximum stress factor is written as:
into Eq. (1) yields:
Kmax = f (g)smax 앀pa (1) tr
K max =7.0 MPa m1/2 (3)

where smax is the maximum applied cyclic stress during The previous calculations indicate that when the
the fatigue test, a is the crack length, and f ( g) is a maximum fatigue crack tip driving force is ~K max tr
=
1/2
constant dependent on the specimen geometry. 7.0 MPa m the crack will begin to propagate more
The smax value is 140 MPa as determined by the far- preferentially through a intermetallics and silicon par-
field boundary conditions. The other values to be esti- ticles as seen in Fig. 3(b). Preferential propagation

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
170 K. GALL et al.

occurs because the fraction of damaged particles at a a higher applied stress will decrease the length of atrans
significant distance from the crack tip increases due to through fracture mechanics arguments. In other words,
the larger crack tip process zone. The maximum size of the area fraction covered by steady fatigue crack growth
the crack tip process (plastic) zone at the transition through the Al–1% Si dendrites will decrease as the
point between local fatigue mechanisms can be esti- applied stress amplitude increases. This is consistent
mated using an Irwin-based approach.39 For the crack with observations as the area fraction of fatigue stri-
geometry and loading conditions given above, the ations decreased dramatically as the applied strain
maximum process zone size under plane strain con- amplitude was increased from 0.2% strain to 0.8%
ditions is estimated as: strain.
It is insightful to compare the relation between the

A B
tr
K max maximum transition stress intensity factor and a tra-
rp =0.11 #96 mm (4)
sy ditional fatigue crack growth curve for similar Al–Si–Mg
alloys. The effective threshold stress intensity factor
In Eq. (4), sy =240 MPa is the yield strength of the is ~DKeff,th =1.5 MPa m1/2 for A356 alloys.40 In modi-
A356 aluminium alloy. The calculation using Eq. (4) fied alloys with much higher silicon particle concentra-
reveals that at the critical transition point, the crack tip tions, DKeff, th is consistently larger (as high as
process zone is ~two–three times the average dendrite 4.0 MPa m1/2 ).16,17 The increase is due to the fact that
cell size. Within this process zone, intense local plasticity cracks propagating with a maximum applied stress intensity
and high triaxial stress fields will lead to the fracture and factor below the transition stress intensity value proposed
debonding of a intermetallics and silicon particles.32,33 here (K maxtr
=7.0 MPa m1/2 ), are slowed by growth
Subsequently, the crack links up with isolated second through eutectic regions as the forward crack tip driving
phase regions that have experienced extensive damage. force is not large enough to damage the silicon particles.
The local coalescence occurs because it is energetically Consequently, an alloy with a higher percentage of silicon
favourable for the crack to change its growth direction particles will present a more torturous route for cracks
and travel through weakened material paths. propagating with a maximum stress intensity below K max tr
Although the aforementioned process can be envi- =7.0 MPa m . For da/dN–DKeff crack growth curves
1/2
sioned when a fatigue crack propagates with a maximum for various Al–Si–Mg alloys, the values at ~5–7 MPa m1/2
stress intensity factor lower than 7.0 MPa m1/2 it is not correspond to the transition point between near threshold
as likely to occur for two critical reasons. growth and Paris law crack growth.2,7,16 Thus, in the
1 The process zone in which the local crack tip stress regime where the crack growth is represented by a Paris
and plasticity are capable of damaging particles does law, the a intermetallics and silicon particles are expected
not sample a statistically significant number of to control the resistance to long crack propagation in a
particles. sense quantifiable by overall average material properties.
2 The deformation band between any damaged particle However, below the K maxtr
=7.0 MPa m1/2 transition value,
regions and the existing crack tip is not intense enough the crack growth characteristics will also depend on the
to promote crack tip meandering. properties of the Al–1% Si dendrite cells and the blockage
induced by the intermittent Si particle-rich Al–Si eutectic
The transition stress intensity factor is expected to be regions. During high-cycle fatigue, alloys with smaller
strongly dependent on the material, particularly the a maximum defect sizes will have lives primarily spent during
and silicon particle distribution and morphology. It can the growth of cracks below the Paris law regime. Thus,
also be asserted that the transition value should change explicitly understanding the micromechanics of crack
if the material ahead of the crack tip has been subjected growth through the dendrite cells is pertinent to these
to a random fatigue overload or been excessively fatigue situations.
fatigued. Either situation could lead to the premature The final stage of fatigue occurs when the crack length
fracture of silicon and a particles, and the subsequent (stress intensity value) reaches a critical value denoted as
growth of isolated microcracks (Fig. 10). However, even af . At this point the maximum crack tip driving force is
if the microcracks are present, the crack tip driving extremely large and the fatigue crack is propagating almost
force still must be large enough to envelope these exclusively through the fractured silicon particles in the
fractured particles and link them to the main crack. In eutectic and a intermetallics (Fig. 10). Once the maximum
any case, given the applied stress level it is now possible applied stress intensity factor is greater than the fracture
to estimate the approximate transition crack length, toughness, KIc , the remaining material fails through a
atrans , where the cracks will start to significantly follow tensile overload process involving rapid crack growth. As
the a intermetallic and silicon particles. Because the in the final stages of the fatigue process, the overloaded
transition stress intensity value is a material parameter, regions fail almost exclusively through the silicon particles

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
M I C R O S TR U C T U R A L I N H O M O G E N E I TI E S I N A l A L LOY 171

in the Al–Si eutectic and a intermetallics. The approximate


CONCLUSIONS
location of the crack front at the point of unstable crack
propagation is indicated in Fig. 3(b) by the larger circular By studying the fatigue behaviour of a premium cast
arc. Therefore, the crack size corresponding to the chosen Al–Si alloy containing a low volume fraction of porosity
crack front in Fig. 3(b) is a priori defined as af (Fig. 10). (0.01% on average), modified silicon particles and modi-
The crack front position was chosen as an approximate fied iron intermetallics (a-Al15 (Fe,Mn)3 Si2 ), the follow-
contour where the increasing area fraction of a intermet- ing significant conclusions can be drawn.
allics saturated. To validate the fracture mechanics
approach and the selection procedure for the final crack 1 During the initial stages of fatigue crack growth
size, the KIc value for the cast alloy can be calculated and from a defect of the order of 100 mm, a crack
compared to other experimental values. propagates almost exclusively through the Al–1% Si
The KIc value is calculated using Eqs (1) and (2), and dendrite cells. The crack avoids the Al–Si eutectic
the maximum depth of the large crack in Fig. 3(b), a= phase and only follows the a intermetallic particles
3.51 mm. For a specimen diameter of d=9.53 mm, Eq. (2) if they are in line with the crack plane. During the
yields f (g)=1.14 for the larger crack in Fig. 3(b). later stages of fatigue, the fatigue cracks propagate
Substituting these values into Eq. (1) using smax = preferentially through damaged a intermetallics and
140 MPa yields KIc =16.8 MPa m1/2 . Previous studies silicon particles. The final overload fracture occurs
have reported KIc values of ~17–18 MPa m1/2 for a similar almost exclusively through a ductile tearing in the
composition and dendrite cell size to the current material.2 Al–Si eutectic phase and a cleavage-like fracture in
Thus, the fracture mechanics analysis and crack locating the a intermetallics.
procedure yield a reasonable KIc value for this alloy, thus 2 The transition between the two fatigue crack growth
the overall technique is a viable tool for analysing the mechanisms occurs when the forward crack tip driving
transition behaviour during fatigue. force and process zone are large enough to damage a
In summary, we assert that the progression of fatigue statistically significant number of silicon and a inter-
damage presented in Fig. 10 is applicable to most cast metallic particles ahead of the crack tip. Then it
Al–Si alloys containing modified silicon particles and dis- becomes energetically favourable for the crack to
perse a intermetallics. However, careful consideration of meander and propagate through these weakened mate-
the size of the defect that initiates the fatigue process must rial zones.
be taken. Fatigue crack paths through the Al–1% Si 3 Using a fracture mechanics solution for a semicircular-
dendrites have been previously reported3 for similar size shaped crack in a cylindrical specimen, coupled with
initiation sites. Nevertheless, conflicting evidence that fracture surface markings, the stress intensity value at
fatigue cracks propagate through the eutectic rather than the transition point between the fatigue crack growth
the Al–1% Si dendrites exists.31 The aforementioned con- mechanisms in point (2) is determined to be ~K max tr

clusion was based on observations in a squeeze casting =7.0 MPa m . For a traditional da/dN–DK curve
1/2

with lower porosity and oxide inclusion contents than the for similar Al–Si cast materials, this value falls close
present alloy. They demonstrated that crack nucleation to the transition point between near threshold and
occurred at eutectic silicon particles and that the small Paris law fatigue crack growth. Furthermore, the
cracks never meandered significantly past the eutectic maximum crack tip process (plastic) zone size corre-
region. The discrepancy between the present work and sponding to this driving force was estimated to be
the work on squeeze cast Al–Si is primarily caused by the rp =96 mm (two–three times the average dendrite
dispersed nature of fatigue damage in squeeze cast alloys cell size).
with small maximum defect sizes. The fatigue process in 4 As the applied strain amplitude is increased from 0.2%
such squeeze cast alloys is more dominated by isolated to 0.8% the fraction of material which fails through
nucleation events and microcracking. By the time several fatigue crack growth at low rates through the Al–1%
of these microcracks link together, the growth phase at Si dendrite cells decreases. This occurs because in the
low rates through the dendrite cells has been bypassed. samples cycled at higher strain (stress) the forward
On the other hand, when nucleation occurs from a defect crack tip driving force exceeds K maxtr
=7.0 MPa m1/2
of 100 mm or so the growth of this dominant crack will at much smaller crack lengths.
experience a growth stage through the Al–1% Si dendrite
cells as observed here. Conversely, if the crack nucleating
Acknowledgements
defect is several hundred mm, the driving force at initiation
may be higher than the transition value derived here, and This work has been sponsored by the US Department
the low crack growth rates through the Al–1% Si dendrite of Energy, Sandia National Laboratories under contract
cells are also bypassed. DE-AC04-94AL85000. This work was performed under

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172
172 K. GALL et al.

the leadership of Dick Osborne and Don Penrod for the 18 A. Kearney and E. L. Rooy (1990) Aluminum foundry products.
USCAR Lightweight Metals Group. We would like to ASM Handbook 2, 123–151.
19 A. Wickberg, G. Gustafsson and L.-E. Larsson (1984)
thank Gerry Shulke for providing the material and
Microstructural effects on the fatigue properties of a cast
Westmoreland Mechanical Testing and Research for A17SiMg alloy. SAE Trans. 840 (121), 1.728–1.735.
conducting the fatigue tests. 20 S. Murali, T. S. Arvind, K. S. Raman and K. S. S. Murthy
(1997) Fatigue properties of sand cast, stircast and extruded
Al–7Si–0.3Mg alloy with trace additions of Be and Mn. Mater.
REFERENCES Trans. JIM 38, 28–36.
21 A. M. Samuel and F. H. Samuel (1995) A metallographic study
1 P. C. Inguanti (1985) Cast aluminum fatigue property/micro-
of porosity and fracture behavior in relation to the tensile
structure relationships. In: 17th SAMPE Conference, 22–24
properties in 319.2 end chill castings. Metall. Mater. Trans. 26A,
October, pp. 61–72.
2359–2372.
2 C. C. Wigant and R. I. Stephens (1987) In: Fatigue and Fracture
22 S. Murali, A. Trivedi, K. S. Shamanna and K. S. S. Murthy
Toughness of A356-T6 Cast Aluminum Alloy-SAE SP 760,
(1996) Effect of iron and combined iron and beryllium additions
Warrendale, PA, pp. 1–100.
on the fracture toughness and microstructures of squeeze cast
3 M. J. Couper, A. E. Neeson and J. R. Griffiths (1990) Casting Al–7Si-0.3Mg alloy. J. Mater. Engng Perform. 5, 462–468.
defects and the fatigue behavior of an aluminum casting alloy. 23 Y. H. Tan, S. L. Lee and Y. L. Lin (1995) Effects of Be and
Fatigue Fract. Engng Mater. Struct. 13, 213–227. iron content on plane strain fracture toughness in A357 alloys.
4 J. C. Ting and F. V. Lawrence (1993) Modeling the long-life Metall. Mater. Trans. A 26A, 2937–2945.
fatigue behavior of a cast aluminum alloy. Fatigue Fract. Engng 24 D. L. McDowell, J. Fan and M. F. Horstemeyer (in press)
Mater. Struct. 16, 631–647. Multi-length scale analyses of cyclically loaded A356 cast
5 S. E. Stanzl-Tschegg, H. R. Mayer, E. K. Tschegg and A. Beste aluminium alloy. AFS Trans.
(1993) In-service loading of AlSi11 aluminium cast alloy in the 25 J. Campbell (1991) Castings, Butterworth Heinemann.
very high cycle regime. Int. J. Fatigue 15, 311–316. 26 K. E. Tynelius, J. F. Major and D. Apelian (1993) A parametric
6 S. Gungor and L. Edwards (1993) Effect of surface texture on study of microporosity in the A356 casting alloy system. AFS
fatigue life in a squeeze cast 6082 aluminum alloy. Fatigue Fract. Trans. 101, 401–413.
Engng Mater. Struct. 16, 391–403. 27 R. M. N. Pelloux (1969) Mechanisms of formation of ductile
7 B. Skallerud, T. Iveland and G. Harkegard (1993) Fatigue life fatigue striations. Trans ASME 62, 281–285.
assessment of aluminum alloys with casting defects. Engng 28 P. Neumann (1974) New experiments concerning the slip
Fracture Mech. 44, 857–874. processes at propagating fatigue cracks—I. Acta. Metall. 22,
8 C. M. Sonsino and J. Ziese (1993) Fatigue strength and 1155–1165.
applications of cast aluminum alloys with different degrees of 29 R. C. Voigt and D. R. Bye (1991) Microstructural aspects of
porosity. Int. J. Fatigue 15, 75–84. fracture in A356. AFS Trans. 99, 33–50.
9 R. B. Gundlach, B. Ross, A. Hetke, S. Valtierra and J. F. Mojica 30 M. D. Dighe and A. M. Gokhale (1997) Relationship between
(1994) Thermal fatigue resistance of hypoeutectic aluminum– microstructural extremium and fracture path in a cast Al–Si-Mg
silicon alloys. AFS Trans. 102, 205–223. alloy. Scr. Mett. 37, 1435–1440.
10 J. F. Major (1994) Porosity control and the fatigue behavior in 31 A. Plumtree and S. Schafer (1986) Initiation and short crack
A356-T61 aluminum alloy. AFS Trans. 102, 901–906. behavior in aluminum alloy castings. In: The Mechanical Behavior
11 A. A. Dabayeh, R. X. Xu, B. P. Du and T. H. Topper (1996) of Short Fatigue Cracks (Edited by K. J. Miller and E. R. de los
Fatigue of cast aluminum alloys under constant and variable Rios), EGF Publication 1, Suffolk, UK, pp. 215–227.
amplitude loading. Int. J. Fatigue 18, 95–104. 32 J.-W. Yeh and W.-P. Liu (1996) The cracking mechanism of
12 K. Shiozawa, Y. Tohda and S.-M. Sun (1997) Crack initiation silicon particles in an A357 aluminum alloy. Metall. Mater.
and small fatigue crack growth behavior of squeeze-cast Al–Si Trans. 27A, 3558–3568.
33 A. Needleman (1987) A continuum model for void nucleation
aluminum alloys. Fatigue Fract. Engng Mater. Struct. 20,
by inclusion debonding. J. Appl. Mech. 109, 525–531.
237–247.
34 P. C. Paris and F. Erdogan (1963) A critical analysis of crack
13 A. A. Dabayeh, A. J. Berube and T. H. Topper (1998) An
propagation laws. Trans. ASME, J. Basic Engng D85, 528–534.
experimental study of the effect of a flaw at a notch root on
35 W. Elber (1970) Fatigue crack closure under cyclic tension.
the fatigue life of cast Al319. Int. J. Fatigue 20, 517–530.
Engng Fracture Mech. 2, 37–45.
14 G. A. Hoskin, J. W. Provan and J. E. Gruzleski (1988) The
36 Y. Murakami (1987) Stress Intensity Factors Handbook, Vols 1, 2,
in-situ fatigue testing of a cast aluminum–silicon alloy. Theor. Pergamon Press, Oxford.
Appl. Fract. Mech. 10, 27–41.~. 37 A. F. Liu (1995) Behavior of fatigue cracks in a tension bolt.
15 F. T. Lee, J. F. Major and F. H. Samuel (1995) Fracture Behavior Structural Integrity Fasteners, ASTM STP 1236, pp. 124–140.
of Al12wt%Si0.35wt%Mg(0–0.02)wt%Sr castings under fatigue 38 T. L. Mackay and B. J. Alperin (1995) Stress-intensity factors
testing. Fatigue Fract. Engng Mater. Struct. 18, 385–396. for fatigue cracking in high strength bolts. Engng Fracture Mech.
16 F. T. Lee, J. F. Major and F. H. Samuel (1995) Effect of silicon 21, 391–397.
particles on the fatigue crack growth characteristics of Al–12 39 H. L. Ewalds and R. J. H. Wanhill (1989) Fracture Mechanics,
Wt Pct Si-0.35 Wt Pct Mg-(0–0.02) Wt Pct Sr casting alloys. Chapman and Hall, New York, NY, USA.
Metall. Mater. Trans. 26A, 1553–1570. 40 M. J. Couper and J. R. Griffiths (1990) Effects of crack closure
17 M. Schaefer and R. A. Fournelle (1996) Effect of strontium and mean stress on the threshold stress intensity factor for
modification on near-threshold fatigue crack growth in an fatigue of an aluminum casting alloy. Fatigue Fract. Engng Mater.
Al–Si–Cu die cast alloy. Metall. Mater. Trans. 27A, 1293–1302. Struct. 13, 615–624.

© 2000 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 23, 159–172

You might also like