You are on page 1of 14

Engineering Fracture Mechanics 115 (2014) 241254

Contents lists available at ScienceDirect

Engineering Fracture Mechanics

journal homepage:

A fracture mechanics approach to interior fatigue crack growth

in the very high cycle regime
Takeshi Ogawa a,1, Stefanie E. Stanzl-Tschegg b,, Bernd M. Schnbauer b,2

Aoyama Gakuin Univ., Dept. of Mechanical Engineering, Fuchinobe 5-10-1, Chuou-ku, Sagamihara-shi, Kanagawa 252-5258, Japan
University of Natural Resources and Life Sciences (BOKU), Peter Jordan Strasse 82, 1190 Wien, Austria

a r t i c l e

i n f o

Article history:
Received 26 February 2013
Received in revised form 18 October 2013
Accepted 2 November 2013
Available online 14 November 2013
Ultrasonic fatigue
Cumulative damage
Repeated two-step tests
Fish-eye fracture
Fatigue crack growth

a b s t r a c t
Growth rates of optically dark areas (ODA) and sh-eyes (FE) were quantied in kHzultrasonic fatigue tests on SUJ2 and 17-4PH steels at constant and repeated two-step
amplitudes. Sizes of ODAs and FEs depended on the stress intensity factor (SIF) range,
and interior fatigue crack growth rates (FCGR) were slower than those of long cracks
in air, suggesting vacuum as ODA growth environment. Repeated two-step tests on SUJ2
steel served to form beach marks so that, a quantication of ODA sizes, interior FCGRs
and SIFs became possible. Additional FCGR measurements of long cracks in vacuum and
comparable fracture morphologies allowed estimating the growth rates of ODAs and FEs
in 17-4PH steel.
2013 Elsevier Ltd. All rights reserved.

1. Introduction
Due to the demand for longer fatigue lives in technical equipment and constructions, research on fatigue failure has attracted increased attention recently. Responding to this demand, material scientists continually develop new materials with
improved fatigue properties, therefore also relying on exact fatigue failure measurements.
As early as 1978, Schijve emphasized the importance of internally growing cracks in aluminum alloys and showed that the
resulting fatigue lives were determined by such cracks growing in a vacuum under realistic loading conditions [1]. In the
1980s, research on the very high cycle fatigue (VHCF) properties of high-strength steels received a most important input,
namely the detection of failures beyond the traditionally assumed fatigue limit in the regime of 106107 cycles [2,3]. Moreover, under special conditions (i.e. material properties, surface condition, amplitude of loading, number of cycles, etc.) such
VHCF failures were found to originate in the interior of the specimen, whereas fatigue cracks in the regime of up to approximately 107 cycles usually start from some kind of surface imperfections. Since that time, interior fatigue crack initiation has
been discovered in other materials than high-strength steels as well. In addition, cracks were found to originate not only
from inclusions, but also from interior pores or porous areas in cast aluminum alloys [4] or even from internal persistent
slip bands in polycrystalline high purity copper [5].
Numerous investigations were performed to determine how structural features such as grain size, morphology and size of
inclusions, surface treatment and pre-straining inuence the fatigue strength and fatigue life of different materials such as

Corresponding author. Tel.: +43 1 47654 5160; fax: +43 1 47654 5159.
E-mail addresses: (T.
(B.M. Schnbauer).
Tel.: +81 42 759 6203; fax: +81 42 759 6502.
Tel.: +43 1 47654 5169; fax: +43 1 47654 5159.


0013-7944/$ - see front matter 2013 Elsevier Ltd. All rights reserved.




T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254

SN curve

compact tension
fatigue crack growth rate in m
cumulative damage
number of cycles
optically dark area within sh-eye fracture surface
stress ratio
scanning electron microscope
plot of stress versus number of cycles to failure
stress intensity factor in MPa m (dened by tensile component, i.e. amplitude of cyclic stress at R = 1 in
this paper)

Austenitic stainless steels [6]. A comprehensive overview of the most important ndings by Li [7] was published recently.
The study investigates the effects of hydrogen, vacuum, the boundary segregation of ne grains and carbides. According
to Li, most of the fatigue life is spent on the formation of a characteristic fracture surface morphology. This characteristic
fracture surface is called ODA (optically dark area) [8] because it looks dark when observed with an optical microscope, while
the same region is also called FGA (ne granular area) [9] or GBF (granular bright facet) [10] based on the morphology. Thus,
the terms correctly indicate the morphological characteristics and the term ODA is used in this paper, because optical
microscopy was the main means of determining the sizes of the regions.
Detailed fractography and transmission electron micrographs both led to corresponding roughness values of ODAs [9,11]:
compared to the original material microstructure, the grain size of the ODAs subsurface structure is extremely reduced [12].
Another remarkable characteristic of ODAs is that a similar morphology can be observed for different material systems, i.e.
high strength steels [13], titanium alloys [1416] and aluminium alloys [17]. In addition, such a morphology can also be
found on the fracture surface of compact tension (CT) specimens made of these materials for standard FCG tests when tested
in a high vacuum environment [13,14,18,19].
Several theories try to explain the formation of ODAs: Murakami [20] emphasizes the inuence of hydrogen, e.g. hydrogen assisted fatigue crack growth. Oguma [12] proposes the segregation of extremely ne grains of material as inuencing
and accompanying ODA formation and Shiozawa et al. [21] emphasize the importance of the segregation of ne carbides and
subsequent formation and coalescence of microcracks. At variable amplitude loading, beach marks were formed on the fracture surface of the sh-eye region around the ODA. These were used to characterize the ODAs and sh-eyes growth rate of
an ODA using a fracture mechanics approach [22].
Only very few papers using fracture mechanical methods to quantify the areas of sh-eyes and especially ODAs around
inclusions manage to give quantitative results for fatigue crack propagation rates and stress intensity factors [9,17,2330]. In
the present study, ultrasonic fatigue tests were performed on high carbon chromium bearing steel, SUJ2, and a precipitation
hardened chromiumnickelcopper steel, 17-4PH, at constant amplitudes. Furthermore, two types of two-step amplitude
stressing tests were conducted on SUJ2 steel in order to investigate the fatigue damage accumulation and the growth behavior of interior cracks. Based on these results together with those from the literature, a formation mechanism of the ODA is

2. Material and experimental procedures

The tested materials were SUJ2, a high carbon chromium steel for bearings (Japan industrial standards JIS) and 17-4PH, a
precipitation hardened chromiumnickelcopper steel which is often used for steam turbine blades. Their chemical compositions and mechanical properties are presented in Tables 1 and 2, respectively. The SUJ2 steel was oil quenched from 850 C
and tempered at 180 C for 120 min. The 17-4PH steel was age hardened at 620 C (condition H1150) and stress relief annealed at 600 C. The specimen shapes for the SUJ2 steel and the 17-4PH steel are shown in Fig. 1(a) and (b), respectively.
The surfaces of the test specimens were polished with abrasive paper in axial direction (nal grade 2000 for SUJ2 and 4000
for 17-4PH).

Table 1
Chemical composition (mass%) of SUJ2 and 17-4PH steel.







Nb + Ta

O (ppm)














T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254

Table 2
Mechanical properties of SUJ2 and 17-4PH steel.



Tensile strength (MPa)

Yield strength (MPa)

Youngs modulus (GPa)

Poissons ratio

Vickers hardness

Density (kg/m3)








Quench: 850 C: 50 min., Temper: 180 C: 120 min.

Condition H1150, Stress relieved at 600 C: 60 min.

Fig. 1. Specimen shapes of SUJ2 specimen (a) and 17-4 PH steel (b).

For the ultrasonic fatigue tests on the SUJ2 steel, a machine consisting of an ultrasonic welder (BRANSON Co., CR-20), a
power supply (2000bcd) and software developed at Aoyama Gakuin University (where the experiments were performed)
were used. The cyclic frequency of testing was 20 kHz, and the displacement amplitude at the specimen end lay between
10 and 50 lm. Fully reversed loading fatigue tests (R = 1) were performed in laboratory air at 24 C with a relative humidity
of less than 40%. The Dr values were dened as the tensile component of the stress range for the experiments at R = 1 in
this paper. In order to reduce specimen heating by the high-frequency ultrasonic vibration, compressed cold air was blown
on the specimen surface using a vortex tube, the specimen temperature was monitored with a radiation thermometer and
intermittent loading was employed: typical loading conditions in the constant amplitude test were 120 ms loading with
3601500 ms pauses. The experiments on the 17-4PH steels were performed at the University of Natural Resources and Life
Sciences in Vienna using closed-loop controlled ultrasonic equipment with vibration gauges serving as feed-back for the
amplitudes [5,17]. The total strain ranges were, in addition, measured with small strain gauges (1 mm gauge length,
2.10 1.0% gauge factor) resulting in a strain measurement accuracy of 2% in all cases. The stress ranges were determined
from the strain ranges using Hookes law. Intermittent loading with pulse lengths of 200010000 cycles and pauses of 500
1000 ms, depending on the amplitude, served to cool the specimens. The experiments were performed at R = 0.05 by superimposing a static tension load with a servo-hydraulic machine. The test temperature was kept constant at 90 C.
Besides constant amplitude loading, two-step and repeated two-step loading tests were performed on the SUJ2 steel. In
the two-step tests (see Fig. 2(a)), the initial stress amplitude Dr1 was 1200 MPa and was applied up to the average number of
cycles to failure determined in the constant amplitude tests (N1 = 6.85  106). This was followed by a second stress amplitude, Dr2 with values between 1100 and 850 MPa, which was applied either until fracture occurred or until the shut-off
number of cycles, N1 = 1  1010, was reached (if no fracture occurred).
In the repeated two-step tests, a higher stress amplitude DrH = 1200 MPa and lower stress amplitude DrL = 1050 or
850 MPa were repeated alternately until fracture, see Fig. 2(b). Table 3 shows the different testing conditions used in the
repeated two-step tests. Most of the tests began with the application of DrL followed by DrH, which is dened as one block.
The number nH of cycles with DrH, was either nH = 1.0  103 or 1.0  104, and the number nL of cycles with DrL was varied
widely according to the nH/nL values indicated in Table 3. The symbols used to designate the different testing conditions in
the following gures are also listed in Table 3.
The cumulative damage, Df, is dened by Eq. (1), where NHf and NLf are the numbers of cycles to failure for DrH and DrL,
respectively, on the regression line of experimental data (Dr versus NF) as shown in Fig. 3. The cumulative damage for DrH is
dened by Eq. (2).


X nH X nL



T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254

Fig. 2. Schematic of two-step tests (a) and repeated two-step tests (b).

Table 3
Testing parameters of four kinds of repeated two-step tests with symbols as used in Figs. 7, 9 and 10.

DrL (MPa)

DrH (MPa)

Starting stress

nH (cycles)






1.0  103
1.0  104
1.0  103

2.0  103  1.0

5.0  102  2.0  101
1.0  101  2.5  101
1.0  103  1.0





X nH

3. Results and discussion

3.1. Constant amplitude and two-step amplitude tests
Fig. 3 displays the fatigue life diagram resulting from the constant amplitude and two-step amplitude tests on SUJ2. The
line represents a linear regression of eight fatigue lives under constant amplitude loading. In the two-step amplitude tests,

Fig. 3. SN curve under constant and two-step loading of SUJ2 steel.

T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254


the rst-step cyclic stress amplitude, Dr1 = 1200 MPa, was applied for N1 = 6.85  106, which corresponded to the average
number of cycles to failure at this stress range in the constant amplitude tests. In order to avoid confusion, broken specimens
under the rst-step amplitude tests were ignored. Then a second-step cyclic stress amplitude was applied until all specimens
which had survived the rst step (in Fig. 3, only those specimens are plotted) fractured. Therefore, the cumulative damage Df
of these specimens exceeded unity. However, the Nf values resulting from the secondary cyclic stress (added to the rst step
cyclic stress) were lower than the ones of the constant amplitude test, which suggests that the Df values lay below two.
Fish-eye fractures originated from interior inclusions in all tested specimens. Fig. 4 shows the optical micrographs,
depicting a specimen after the constant amplitude test at Dr = 1200 MPa (a) and after the two-step test at Dr1 = 1200 MPa
followed by Dr2 = 900 MPa (b). In both cases, ODAs, are visible, their sizes depending on the test conditions. An inclusion can
be seen at the center of each ODA.
The areas of the inclusion, ODAs and sh-eyes were measured and their square root values, area, were used to calculate
the stress intensity factor DK, using the equation for interior cracks [25]:

DK 0:5  Dr 

p  area;

Dr being the stress amplitude for R = 1 or the stress range for R = 0.05. DK thus always corresponds to the tensile component of the cyclic stress.
Fig. 5(a) presents the square root of the inclusion sizes, areaInc and their stress intensity value, DKInc, as a function of Dr
for the SUJ2 and the 17-4PH steel: both areaInc and DKInc were independent of Dr. The DKInc values varied between 2.3 and
4 MPa m. Fig. 5(b) shows the square root value of the areas and the stress intensity factors for the ODAs, denoted as
areaODA and DKODA, respectively. The values of DKODA were almost constant, ranging from 5 to 5.5 MPa m for both steels.
This almost constant DKODA value implies that this is the border line between the ODA and the sh-eye, i.e. a criterion for the
transition of the crack from an ODA to the neighboring sh-eye area. Equally, the stress intensity values of DKFE for the shp
eyes in SUJ2 steel (Fig. 6) were approximately constant, ranging from 16 to 21 MPa m, which agrees with the fracture
toughness measured in a standardized fracture mechanical test with a CT specimen. The asterisks in Fig. 6 mark the cases
in which the sh-eyes joined the specimen surface. In these cases, DKFE was calculated using the following equation [25]:

DK 0:65  Dr 

p  area

To summarize, the DKInc and DKODA values in the threshold regime were rather similar (assuming that only the tensile portion of cyclic loading contributed to the crack propagation) although two different steels were investigated: DKInc lay bep
tween 2.17 and 3.03 MPa m and DKODA was 5.025.79 MPa m (cf. Fig. 5). These similarities were found even though
different testing equipment was used and the R-ratio was different (R = 1 for SUJ2 and R = 0.05 for 17-4PH).

Fig. 4. Optical micrographs of SUJ2 showing ODAs after a constant amplitude loading test at Dr = 1200 MPa (a) and a two-step loading test at
DrH = 1200 MPa followed by DrL = 900 MPa (b).


T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254

Fig. 5. Size and stress intensity factors of SUJ2 for cyclic stress amplitudes at fracture and cyclic stress ranges at fracture of 17-4PH steel for inclusions (a)
and optically dark areas (b).

The run out specimen in Fig. 3 tested at Dr2 = 850 MPa (2nd step) was broken by subsequent fatigue loading at
Dr = 1200 MPa. On the fracture surface, a large ODA with DKODA = 6.3 MPa m was observed. This value is apparently larger
than that of the DKODA demonstrated in Fig. 5(b) suggesting that the ODA grew due to the cyclic stress at Dr = 850 MPa. In
order to conrm this hypothesis, two similar experiments were conducted by applying different run-out numbers of cycles
(N = 2.0  109 and 3.0  1010) at Dr2 = 850 MPa. In both tests, however, the specimens did not break at Dr2 = 850 MPa and
were subsequently broken at Dr = 1200 MPa. The calculated values of DKODA for these two experiments were 5.3 and
5.1 MPa m, respectively. These results indicate that in contrast to the above mentioned expectation, there is no growth
of ODA at a cyclic stress of Dr2 = 850 MPa and that in addition, the statistical scatter of DKODA values and the sizes of the
ODAs is rather large.

T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254


Fig. 6. Sizes and stress intensity factors of sh-eyes for stress amplitudes at fracture in SUJ2.

The resulting DKInc and DKODA values for different sizes of inclusions and ODAs in constant amplitude tests are summarized in Fig. 5(a) and (b) for SUJ2 and 17-4PH steel. Although two different steels were investigated, their DKInc and DKODA in
the threshold regime are rather similar assuming that only the tensile portion of cyclic loading contributes to crack propp
agation -, with DKinc between 2.17 and 3.03 MPa m and DKODA  5.025.79 MPa m, though different testing equipment
was used and the R-ratio was different (R = 1 for SUJ2 and 0.05 for 17-4PH).

3.2. Repeated two-step amplitude tests

Fig. 7 presents the cumulative damage at fracture Df, resulting from the repeated two-step amplitude tests, as a function
of nH/nL. Furthermore, Fig. 7 also shows the value of the damage ratio by higher stress cycles, DH/Df, of the tests with
DrL = 1050 MPa and 850 MPa. In the tests with DrL = 850 MPa (black dots), DH/Df was higher than 0.95 which indicates that
Df was mainly caused by the higher stress cycles. When DrL was chosen to be 1050 MPa (open symbols), the scatter of the Df
values was much larger, lying between 0.40 and 14.29, and no correlation with nH/nL was obvious. Since, however, in ve of
33 experiments, Df values lower than one resulted, it is assumed that a two-step variation of the stress amplitude increases
the fatigue life when increasing the ratio nL/nH.

Fig. 7. Relationship between cycle ratios and cumulative damage values (symbols explained in Table 3).


T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254

A sh-eye fracture surface originating from an interior inclusion was observed in all he tested two-step specimens.
Fig. 8(a)(c) show the optical micrographs of a specimen tested with the starting stress range of DrL = 850 MPa
(nL = 1.0  104, nH = 1.0  103 and Nf = 2.87  108) using different magnications in order to show sh-eye, beach marks
and ODA, respectively. ODAs can be seen around the inclusions in all the fractured specimens after two-step tests. Beach
marks were identied in the sh-eye regions but not in the ODAs. The variation of the growth increment of the beach marks
depended on the nH/nL values (see next section), and the relationships revealed that the dark and bright fracture surfaces
corresponded to the crack growth increments caused by DrH and DrL, respectively. Fig. 8(d) presents a scanning electron
microscope (SEM) image of an ODA obtained by backscattering electron imaging; the morphology of the ODA region around
the inclusion is granular. The stress variations were not visible in the high magnication SEM micrograph, which is in agreement with the optical result. The SEM observation indicated that the dark beach mark regions in the sh-eye region were
slightly rougher than the bright one. The number of beach marks in the sh-eye region outside the ODA area indicated that
the number of cycles to form this fracture surface was less than 1% of the total fatigue life, and most of the fatigue cycles
were spent for the formation of the ODA. Therefore, the number of cycles to form an ODA is concluded to be approximately
the same as the total fatigue life.
Fig. 9 shows the relationship between areaInc and nH/nL, where, as expected, no correlation could be found. This result
and the mentioned tendency of Df to decrease with a decreasing ratio of nH/nL in Fig. 7 provides the relationship between Df
and areaInc shown in Fig. 10. Ignoring the two exceptions of very small inclusion sizes, Df tended to decrease with an
increasing areaInc, and the values were smaller than the one for the three largest inclusions. This result implies that Df depends on the inclusion size at the fracture origin in the repeated two-step tests.
Fig. 11 presents the variations of areaODA and DKODA with nH/nL. Some of the ODAs had different shades of darkness, as
shown in Fig. 12(a). As can be seen, the ODA is the darker interior region, and it is concluded that the widths of outer region
represent different nH/nL conditions: this assumption is based on a comparison of the image with the SEM micrograph shown
in Fig. 12(b). The outer less dark region in Fig. 12(a) grew due to a high stress range, similar to the beach mark region presented in Fig. 8(b). This fact suggests that the transition from ODA to sh-eye region occurred as soon as the high stress
amplitude was applied, and thus DKODA should be calculated for DrH = 1200 MPa. When nH/nL was larger than 0.1, the values
of DKODA were approximately the same as those presented in Fig. 5(b) for constant amplitude and two-step tests. However,
the ODAs tended to increase with decreasing nH/nL, i.e. increasing nL. This means that a high number of lower cyclic stresses
augmented the size of the ODA. Such behavior can be explained by the following: when an ODA reaches the size at which its
DK exceeds the constant value of DKODA shown in Fig. 5 at DrH = 1200 MPa, a transition of the growth mechanism from the
ODA to the sh-eye region mechanism may occur. If, due to the limited number of cycles, DK at DrL becomes lower than the
critical DKODA at the cyclic stress of DrL = 850 MPa, the growth mechanism returns to the ODA mechanism. This sequence is
repeated until transition from ODA to sh-eye mechanisms occurs in the following way: when the high cyclic stress is

Fig. 8. Optical and scanning electron micrographs showing a sh-eye (a), beach marks (b), an ODA (c) and an inclusion (d) for DrL = 850 MPa, nH = 1  103
and nL = 1  104. Note that the red and blue arrows in (b) indicate the crack growth increment under high and low stress ranges, respectively.

T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254


Fig. 9. Relationship between cycle ratios and inclusion sizes under repeated two-step amplitude stressing (symbols explained in Table 3).

Fig. 10. Relationship between inclusion sizes and cumulative damage values Df under repeated two-step amplitude stressing (symbols explained in
Table 3).

reapplied, DK becomes higher than DKODA and the crack grows. When the stress is reduced, i.e. at the beginning of low cyclic
stressing at a period of the high cyclic stress, where DK was higher than DKODA, the crack ceases to grow.
If the value of DK is calculated using the size of the outer contour of the dark area shown in Fig. 12(a) together with DrL, a
value of 8.8 MPa m results, which exceeds the transition DKODA by far. Oguma et al. [19] observed an ODA-like fracture surface morphology on fatigue fracture surfaces which had repeated contact in a high vacuum environment. This observation
supports our theory of the ODA growth sequence: sh-eye growth increments change to ODA growth by the repeated surface
contact under DrL conditions. This will be discussed in more detail in 3.4.

3.3. Growth characteristics of interior crack

The beach marks shown in Fig. 8 provide growth rates for the interior crack in the sh-eye region [22]. Fig. 13 represents
the relationships of the crack growth rate, da/dN, and the stress intensity factor range, DK.
The crack growth increments for DrL and DrH are denoted DaL and DaH, respectively. The crack growth rate da/dN was
determined by DaL/nL or DaH/nH; the corresponding DKwas obtained from the area inside of the beach mark and the applied
Dr according to Eq. (3). The error bars for DK in Fig. 13 correspond to the widths of beach marks, i.e. DaL or DaH. For comparison, Fig. 13 also presents the growth characteristics of a long crack, which grew in a CT specimen in air at a stress ratio of
0.1 in a standard fracture mechanics test [31]: clearly, the interior crack growth rates obtained by the beach marks are lower
than those observed on the CT specimen. A possible reason for this deviation is the environment: while it is assumed that
interior cracks grow in a vacuum (the vacuum quality, however, has not been determined up to now), the crack in the CT
specimen grew in air.
To verify this assumption, crack growth experiments in vacuum are necessary, which, however, could not be performed
with this material. Instead of that, fracture surfaces of crack propagation tests on 12% Cr steel in a vacuum of approx. 103 Pa


T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254

Fig. 11. Sizes and stress intensity factors of ODAs under repeated two-step amplitude stressing.

Fig. 12. Micrographs showing a double shaded ODA after tests with DrL = 850 MPa, nH = 1  103 and nL = 1  106 observed with an optical microscope (a)
and a scanning electron microscope (b).

T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254


Fig. 13. Relationship between da/dN and DK. Closed symbols: determined from measurements on CT specimens in air. Open symbols with DK in the range
of 7 to 16 MPa: determined from measurements of sh-eye beach marks; open symbols with DK around 5 MPa: determined from radii of ODAs.

at 19 kHz and very high numbers of cycles are used and compared with the fracture morphology of ODAs [13]. It was concluded that within the ODAs, the growth rates were in the range of approximately 1012 m/cycle. The growth rates in the
smooth area adjacent to the ODAs were found to be below approximately 1011 m/cycle [13]. Based on the similarities
of the fracture morphologies of sh-eye regions and long-crack fracture surfaces formed in a vacuum, it was further concluded that fracturing occurs according to fracture mechanical laws for long cracks with rates depending on the prevailing
stress intensity factor. As soon as the maximum stress intensity factor exceeds the threshold stress intensity factor for fatigue crack propagation of long cracks, higher crack growth rates ranging from around 109 m/cycle up to about 107 m/cycle will occur outside the sh eye, thus forming the contour of the sh-eye. As stated earlier, the cycle ratio of this sh-eye
growth phase to the total fatigue life is believed to be very small in comparison with the ODA growth and crack initiation
stage. Concerning the environmental question, these studies conrm that vacuum is the crucial environment for fatigue
crack propagation in the ODA, and the authors assume that it is justied to apply this result to the SUJ2 steel.
Concerning fatigue crack growth within the whole sh-eye area (not only ODA), numerous interesting studies are reported in the literature. Since this paper, however, is focused on the ODA growth, only few important studies are mentioned.
Petit and Sarrazin-Baudoux [31] developed a model based on plastic deformation accumulation controlling intrinsic crack
propagation in vacuum. They estimated the life of high strength steels with cracks that initiated at a sh-eye and could show
that the portion of life due to crack propagation in vacuum was two to three orders of magnitude greater than in air. This
result conrms the measurements of this study. In [32], ultra-slow fatigue crack propagation at ultrasonic frequency is
Murakami [20] discovered the inuence of hydrogen content trapped around inclusions to play an important role during
internal fatigue crack growth. The extremely low crack growth rates of the ODAs measured in the present study, however,
rather exclude the inuence of hydrogen. In a more recent study [33], Murakami pointed out that most of life cycles Nf are
spent in the ODA area. He distinguishes between material containing 0.8 ppm and, respectively, 0.1 ppm hydrogen (CrMo
steel SCM 435 after different heat treatment) and suggested a master curve for the number of cycles to failure Nf versus
( areaODA/ areaInclusion). This curve showed material independence at a hydrogen content of 0.20.9 ppm. For material containing only 0.20.3 ppm hydrogen, a modied model for the fatigue limit was suggested considering that the ODA grows as
a crack. Further work is planned to apply this model to the results of the present study.
The ODA growth rates in the two-step tests of the present study could be estimated making some assumptions, as explained in the following: in the present study, the rst step of Dr1 = 1200 MPa was applied up to a number of cycles that
was somewhat smaller than the number of cycles to failure obtained in the constant amplitude test given in Fig. 3. At this
point, we assumed, an ODA of similar size as those observed in the constant amplitude tests at Dr = 1200 MPa had been
formed. In the subsequent second step, after applying a second dened stress Dr2, this ODA then grew up to the size observed at fracture. In other words, we considered the ODA at Dr1 = 1200 MPa as a pre-crack, which grew at the cyclic stress
amplitude during the second step without any effects of stress history. Fig. 13 shows the da/dN values calculated by the ODA
growth increment divided by the number of cycles to failure during the second step, in which the number of cycles to form a
sh-eye was ignored owing to their much smaller values. The values of DK were determined by the Dr of the second step
together with the initial and nal sizes of the ODA, whose corresponding values are characterized by the error bars. As can be
seen, the growth rate of a sh-eye is ten to hundred thousand times larger than that of an ODA. This great difference in the
growth rates seems to be the most probable reason for the apparent change in the fracture surface morphology.


T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254

3.4. Mechanism of ODA formation

For further investigation of the mechanism of ODA formation, three-dimensional fracture surface observations were performed using an atomic force microscope (AFM, Keyence Co. VN-8000) for two different conditions of repeated two-step
tests, i.e. nH/nL = 0.20 and 0.02 for nH = 1.0  103. Fig. 14 presents surface roughness proles resulting from these two different testing conditions: AFM1 and AFM2 are images of the two opposite fracture surfaces of a specimen after testing with nH/
nL = 0.20 (a) and nH/nL = 0.02 (b). The graphs on the right show the sectional appearances along the dashed lines in the photos
on the left, which are considered to be corresponding locations. The sectional appearances revealed that the pair of the ODA
surfaces match nicely, as reported earlier by Shiozawa et al. [9,21].
In the past, signicant suggestions for the mechanism of ODA formation were made. Some came from Oguma et al. [12] as
mentioned above, from Furuya et al. [27,28] and Nakajima et al. [29], who studied the development of ODAs, crack initiation
and fatigue limit of SUJ2 steel with two-step loading tests with a rotating bending machine. Other important inputs came
from Murakami et al. [20], Murakami and Matsunaga [33] and Petit and Sarrazin [31,32] as mentioned above.
The present study revealed that the growth of ODAs is ten to hundred thousand times slower than that in the outer sheye regions, though both may be assumed to occur in a vacuum. Also, in the repeated two-step tests, the growth increment
under high stress amplitudes obviously changed to ODA growth, as was concluded from the experiments with higher runout numbers of cycles at the highstress amplitudes (cf. last paragraph of 3.1). This was probably due to the repeated contact under the low stress amplitude. Higher growth rates after the beginning of ODA formation owing to the initial absence of
crack closure effects could not be observed in this study, but cannot be excluded.
The very slow growth of the ODAs may be explained by the following mechanism based on Ogumas study [19], which
applies to ODA formation: it is assumed that the ODA morphology is formed by repeated fracture surface contact. In a vacuum, fracture surface roughness induces local contact, so that such locations may re-bond (cold-weld) because the surface is
active in a vacuum. During the following loading, the welded location may undergo plastic deformation and local cold work
may occur. These considerations are based on numerous pioneer studies on fatigue crack initiation and propagation being
inuenced by contact of rough fracture surfaces, rewelding of fresh fracture surfaces and enhanced plasticity thus leading
to longer fatigue lifes [3437]. If fatigue loading is repeated for a long time, the local cold work may produce ne grains
as observed by Oguma et al. [12] in the TEM. The results strongly suggest that the repeated contact in a vacuum produces
the ODA morphology. Since ODAs grow extremely slowly and discontinuously, the repeated fracture surface contact in a vacuum takes place very often, so that a very ne grained microstructure is formed. Thus, the darkness of the ODA seems now to
have the consensus interpretation of being caused by an extreme surface roughness associated with a ne granularity. Further investigations, for example, introducing Murakamis ideas [33] are needed to clarify the mechanisms being relevant for
the ODA formation.
The diffusive nature of the process of small grain formation could be explained as follows: low stress amplitudes, as those
discussed above, lead to shear crack propagation, which produces multiple crack plane deviations. Thus, the local strains at
the asperity interactions and the crack tip are probably very high and would therefore induce very intense point defect creation. This is one possible cause of a diffusive transformation in the near crack surface grain structure. Another possible cause
can be excluded, namely a signicant global temperature rise in the specimen owing to the high frequency of stressing. Such
a global heating was refuted by comprehensive experimental measurements [5]. Very localized temperature increases, however, cannot be excluded, so that both welding and recrystallization would be possible. An important argument against a

Fig. 14. Surface roughness prole of a fracture surface after repeated two-step amplitude stressing; DrL = 1050 MPa, nH = 1  103 and nL = 5  103 (a),
DrL = 1050 MPa, nH = 1  103 and nL = 5  104 (b).

T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254


Fig. 15. SEM images of ODAs in AISI 420 Cr-steel (a), in AISI 316L-PM/HP steel (b) and spray formed hypereutectic Al-Si alloy (c).

high-frequency inuence is offered by the results of Furuya et al. [27]. They found ODAs at testing frequencies of 100 Hz,
600 Hz and 20 kHz around inclusions of Al2O3 or Al2O3.CaO and, in addition, a frequency independence of the SN diagrams
for low-temperature-tempered JIS SNCM430 steel. As can be concluded from the above mentioned considerations, ODAs are
probably formed by repeated fracture surface contact in a vacuum without the inuence of the original material microstructure such as grain size. This interpretation also implies that ODA formation is a crack propagation mechanism and cannot be
considered as crack initiation, as assumed by a few authors (e.g. [30]).
After hydrogen and ne carbides were found to affect the growth rates of ODAs [12,20,21]. the nding of the present study
that the ODAs morphology was independent of these inuences seems interesting. The other characteristic of ODAs their
presence in different materials (in high-strength steels [7,912,22,2729], titanium alloys [15,16,18,19], medium strength
alloy steel [13,38,39] and Al alloys [17]) is corroborated by examples of ODAs in AISI 420 Cr steel, AISI 316L-PM/HIP steel
and spray formed hypereutectic AlSi alloy in Fig. 15.
4. Conclusion
In the present study, ultrasonic fatigue tests at constant amplitudes were carried out on SUJ2, a high-strength carbon
chromium bearing steel, and 17-4PH, a precipitation hardened chromium-nickel-copper steel at constant amplitudes. In
addition, two-step and repeated two-step amplitude tests were performed on the SUJ2 steel in order to investigate fatigue
damage accumulation and growth behavior of interior cracks. The results obtained are summarized in the following:
(1). In all steels tested in the present study, the fracture origin was from an interior inclusion, around which an optically
dark area (ODA) and sh-eye regions were observed. Their sizes decrease with increasing stress amplitude, whereby
the stress intensity factors (SIF) remain approximately constant.
(2). In the repeated two-step amplitude stress tests, the cumulative damage Df scattered between 0.40 and 14.29. Five of
33 cases of testing conditions resulted in a Df value below unity. It was thus concluded that variations of the stress
amplitude increase the fatigue life. The Df values decrease with increasing inclusion size, while their dependence
on the cycle ratio of high and low stress amplitudes remained unclear.
(3). Fatigue crack growth characteristics expressed by the relationships between the growth rate da/dN and the stress
intensity factor range DK were determined by beach marks in the sh-eye region which resulted from the repeated
two-step tests and the differences in ODA sizes which resulted from the two-step tests. The growth rates of the
sh-eyes were smaller than those for long cracks obtained by fracture mechanics experiments in air, and ten or hundred thousand times higher than those of the ODAs.
(4). It is estimated that the ODA morphology is formed by repeated fracture surface contact in a vacuum, i.e. in the interior
of material, and is not determined by its growing mechanism which, however, causes a disintegration the original
microstructural features.

The present research was conducted as a cooperative work of GCF3 committee of The Japan Welding Engineering Society
supported by Kyusyu Electric Power Co., Inc., Hokkaido Electric Power Co. Inc., Tohoku Electric Power Co. Inc., Tokyo Electric
Power Company, Chubu Electric Power Co., Inc., Hokuriku Electric Power Company, The Kansai Electric Power Co., Inc., The
Chugoku Electric Power Co., Inc., Shikoku Electric Power Co. Inc., The Japan Atomic Power Company and Electric Power
Development Co., Ltd. The SUJ2 specimens were provided by Yukio Matsubara of NTN Corporation. The experiments on
the 17-4PH steel were performed within a research project of the Electric Power Research Institute (EPRI), USA, and the University of Natural Resources and Life Sciences, Austria. Language support by Dipl.-Ing Veronika Doblhoff-Dier is gratefully
acknowledged. We also thank two of the reviewers for their constructive criticism, which we implemented in Sections
3.3 and 3.4.


T. Ogawa et al. / Engineering Fracture Mechanics 115 (2014) 241254

[1] Schijve J. Internal fatigue cracks are growing in vacuum. Engng Fract Mech 1978;8(4):57581.
[2] Naito T, Ueda H, Kikuchi M. Fatigue behavior of carburized steel with internal oxides and nonmartensitic microstructure near the surface. Metall Mater
Trans A 1984;15A(7):14316.
[3] Asami K, Sugiyama Y. Fatigue strength of various surface hardened steels. J Heat Treatment Technol Assoc 1985;25(3):14750.
[4] Caton MJ, Jones JW, Mayer H, Stanzl-Tschegg SE. Demonstration of a fatigue limit in cast 319 aluminium. Metall Mater Trans A 2003;34A:3341.
[5] StanzlTschegg S, Eichinger K, Weidner A, Tschegg E, Bernardi J, Schnbauer B. Role of impurities and PSBs on microcracking of polycrystalline copper
at very high numbers of cycles. Key Engng Mater 2011;465:2934.
[6] Ogawa T, Nakane N, Masaki K, Hashimoto S, Ochi Y, Asano K. Investigation of effect of pre-strain and very high-cycle fatigue strength of Austenitic
stainless steels. J Power Energy Syst 2009;3(1):3850.
[7] Li SX. Effects of inclusions on very high cycle fatigue properties of high strength steels. Int Mater Rev 2012;57(2):92114.
[8] Murakami Y, Nomoto T, Ueda T, Murakami Y. On the mechanism of fatigue failure in the superlong life regime (N > 107 cycles). Part II: a fractographic
investigation. Fatigue Fract Engng Mater Struct 2000;23(11):90310.
[9] Shiozawa K, Morii Y, Nishino S, Lu L. Subsurface crack initiation and propagation mechanism in high-strength steels in a very high cycle regime. Int J
Fatigue 2006;28:152132.
[10] Sakai T, Sakai Y, Oguma N. Characteristic SN property of high carbon Chromium bearing steel under axial loading in long life fatigue. In: StanzlTschegg S, Mayer H, editors. Proceedings of international conference on fatigue in the Very High Cycle Regime (later named VHCF-2), Vienna, Austria;
2001. p. 517.
[11] Murakami Y, Nomoto T, Ueda T, Murakami Y, Ohori M. Analysis of the Mechanism of Superlong Fatigue Failure by Optical Microscope and SEM/AFM
Observations. J Jpn Soc Mater Sci 1999;48(10):11127 (in Japanese).
[12] Oguma N, Harada H, Sakai T. Mechanism of long life fatigue fracture induced by interior inclusion for bearing steel in rotating bending. J Jpn Soc Mater
Sci 2003;52(11):11127 [in Japanese].
[13] Stanzl-Tschegg SE, Schnbauer B. Near-threshold crack propagation and internal cracks in steel. Procedia Eng 2010;2:154755.
[14] Nakamura T, Oguma H, Shinohara Y. The effect of vacuum-like environment inside subsurface fatigue crack on the formation of ODA fracture surface in
high strength steel. Procedia Eng 2010;2:21219.
[15] Shiina T, Nakamura T, Oguma H. Effects of high vacuum environment on high cycle fatigue properties of Ti6Al4V alloy. In: Tanaka K, editor.
International conference on advanced technique. in experimental mechanics, Nagoya, Japan (ATEM 2003); 2003 [CD-ROM].
[16] Shiozawa K, Kuroda Y, Nishino S. Effect of stress ratio on subsurface fatigue crack initiation behavior of Beta-type Titanium alloy. Trans Jpn Soc Mech
Eng A 1998;64(626):252835 [in Japanese].
[17] Stanzl-Tschegg SE, Mayer H, Schuller R, Przeorski T, Krug P. Fatigue properties of spray formed hypereutectic Aluminium silicon alloy DISPAL S232 at
high and very high numbers of cycles. Mater Sci Engng A 2012;538:32734.
[18] Oguma H, Nakamura T. Inuence of environment on the formation of unique morphology on fracture surface in sub-surface fractures. In: Berger C,
Christ H-J, editors. Proceedings of 5th international conference on Very High Cycle Fatigue (VHCF-5), Berlin; 2011. p. 25762.
[19] Oguma H, Nakamura T. Fatigue crack propagation properties of Ti6Al4V alloy in vacuum environments. Int J Fatigue 2012. Dot:10, 1016/
[20] Murakami Y, Nomoto T, Ueda T, Murakami Y. On the mechanism of fatigue failure in the superlong life regime (N > 107 cycles). Part I: inuence of
hydrogen trapped by inclusions. Fatigue Fract Engng Mater Struct 2000;23:893902.
[21] Shiozawa K, Shimatani Y, Nakada T, Yoshimoto T, Koshi M. Very high cycle fatigue properties of high speed steel developed by basis on internal failure
mechanism. Trans Jpn Soc Mech Eng A 2010;76(772):168997 [in Japanese].
[22] Sugeta J, Sugiyama Y, Minoshima K. Ultra high-cycle fatigue characteristics and interior crack growth behavior under repeated two-step loading on
high strength steel. In: Proceedings of JSME annual meeting 2007, No. 07-1; 2007. p. 3456 [in Japanese].
[23] Schijve J. Fatigue of structures and materials in the 20th century and the state of the art. Int J Fatigue 2003;25:679702.
[24] Tanaka K, Mura T. A theory of fatigue crack initiation at inclusions. Metall Mater Trans A. 1982;13A:11723.
[25] Murakami Y. Metal fatigue: effects of small defects and nonmetallic inclusions. 1st ed. Oxford: Elsevier; 2002.
[26] Kumar A, Torbet CJ, Pollock TM, Jones JW. In situ characterization of fatigue damage in a cast Al alloy via nonlinear ultrasonic measurements. Acta
Mater 2010;58:214354.
[27] Furuya Y, Matsuoka S, Abe T, Yamaguchi K. Gigacycle fatigue properties for high-strength low-alloy steel at 100 Hz, 600 Hz and 20 kHz. Scripta Terialia
[28] Furuya Y, Hiukawa H, Kimura T, Hayaishi M. Gigacycle fatigue properties of high-strength steels according to inclusion and ODA sizes. Metall Mater
Trans A 2007;38A:172230.
[29] Nakajima M, Kamiya N, Itoga H, Tokaji K, Ko H-N. Experimental estimation of crack initaion lives and fatigue limit in subsurface fracture of a high
carbon chromium steel. Int J Fatigue 2006;28:15406.
[30] Paris PC, Marines-Garcia I, Hertzberg RW, Donald JK. The relationship of effective stress intensity factor, elastic modulus and Burger s vector on fatigue
crack growth as associated with Fish Eye gigacycle fatigue phenomena, In: Sakai T, Ochi Y, editors. Proceedings of 3rd international conference on
Very High Fatigue (VHCF-3), Kusatsu, Japan, September 1619, 2004, JSMS, p. 113.
[31] Petit J, Sarrazin-Baudoux C. An overview on the inuence of the atmosphere, environment on ultra-high-cycle fatigue and ultra-slow fatigue crack
propagation. In: Sakai T, Ochi Y, editors. Proceedings of 3rd internatinal conference on Very High Fatigue (VHCF-3), Kusatsu, Japan, September 1619,
2004, JSMS, pp. 514527.
[32] Petit J, Sarrazin-Baudoux C. Some critical aspects of low rate fatigue crack propagation in metallic materials. Int J Fatigue 2010;32:96270.
[33] Murakami Y, Matsunaga H. Factors affecting ultralong life fatigue and design method for components. In: Lecture at Chinese Academy of Science, June
21, 2013.
[34] McEvily AJ. On the quantitative analysis of fatigue crack propagation, In: Lankford J, Davidson DL, Morris WL, Wei RP, editors. Fatigue mechanisms:
advances in quantitative measurement of physical damage, ASTM STP 811, American Society for Testing and Materials; 1983. p. 283312.
[35] Neumann P, Vehoff H, Fuhlrott H. In: Fracture 1977. Proceedings of 4th International Conference on Fracture, ICF4, vol. 2. University of Waterloo press,
Ont., Canada; 1977. p. 1313.
[36] Petit J, Hnaff G, Sarrazin-Baudoux C. Environmentally-assisted fatigue in gaseous atmosphere. In: Milne I, Ritchie RO, Karihaloo BL, editors.
Comprehensive strucural integrity 6: environmentally-assisted fatigue, Elsevier; 2003. p. 21180.
[37] Ritchie RO, Davidson DL, Boyce BL, Campbell JP, Roder O. High-cycle fatigue of Ti6Al4V. Fatigue Fract Eng Mater Struct 1999;22:62131.
[38] Nishizawa H, Ogawa T. JSMS4 Mode II fatigue crack growth characteristics and experimental evaluation of the crack driving force. J Jpn Soc Mater Sci
2005;54(12):1295300 [in Japanese].
[39] Sato M, Kaneda S, Ogawa T. Giga-cycle fatigue strength properties of low-alloy steel SFVQ1A evaluated by ultrasonic fatigue test. Trans Jpn Soc Mech
Eng A 2012;78(789):6048 [in Japanese].