International Journal of Fatigue 31 (2009) 880–888

Contents lists available at ScienceDirect

International Journal of Fatigue

journal homepage: www.elsevier.com/locate/ijfatigue

Very high cycle fatigue properties of bearing steel under axial loading condition
K. Shiozawa a,*, T. Hasegawa b, Y. Kashiwagi a, L. Lu c
a
Graduate School of Science and Engineering, University of Toyama, Toyama 930-8555, Japan
b
Okuma Co., Oguchi-cho, Niwa-gun, Aichi 480-0193, Japan
c
State Key Laboratory of Traction Power, Southwest Jiaotong University, Chengdu 610031, China

a r t i c l e i n f o a b s t r a c t

Article history: It has been observed to appear a step-wise or duplex S–N curve under the test of rotary bending fatigue
Received 29 October 2007 using high strength steel. This behavior was caused by the transition of fracture mode from surface-
Received in revised form 31 October 2008 induced fracture to subsurface inclusion-induced fracture. The aim of this study is to clarify the S–N char-
Accepted 2 November 2008
acteristic under an axial loading fatigue in the very high cycle fatigue regime. In order to investigate the
Available online 12 November 2008
mean-stress effects, fatigue tests were carried out in air at room temperature under three applied stress
ratios of 1, 0 and 0.5 using a hour-glass shaped specimen of high carbon–chromium bearing steel, JIS
Keywords:
SUJ2. From the results, three types of fracture mode were observed on the fracture surface, such as sur-
Very high cycle fatigue
Subsurface crack
face-induced fracture, subsurface inclusion-induced fracture without granular bright facet (GBF) area and
Fracture mode that with GBF area around an inclusion. Fatigue lifetime for transition in the fracture mode depended on
Axial loading the applied stress ratio. Shape of the S–N curve was a smooth and continuous under three testing condi-
Stress ratio tions in spite of the occurrence on the three types of fracture. Detail discussion for fatigue fracture behav-
Residual stress ior was made through the observation of fracture surface and from point in view of the fracture
mechanics. In addition, an effect of residual stress in the specimen surface layer on the transition of frac-
ture mode was discussed and compared with the experimental results.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction area revealed a very rough and granular morphology in compari-

In recent years, attention has been given to the fatigue behavior
of materials in the very high-cycle region, because of a require-
ment for high efficiency and reliability for machines and struc-
tures. An increasing interest of this topic is reflected by the fact
that the international conference is already held four times since
1998 as a series of ‘‘Very High Cycle Fatigue Conference” [1]. It
has been reported by some researchers that a step-wise or a duplex
S–N curves occurs for high-strength steels tested under a rotary
bending fatigue [1–12]. This distinctive behavior is appeared by
change in crack initiation site from a surface to a subsurface defect,
depending on applied stress amplitude. That is, one of two S–N
curves appears at high-stress amplitude level and low number of
cycles, and is governed by a surface fracture mode. The fatigue lim-
it appears at critical stress amplitude where the initiated surface
microcracks are arrested. Another S–N curve appears at low-stress
amplitude level and high number of cycles, and is governed by an
internal crack initiation and growth. On the fracture surface result-
ing from internal crack initiation and propagation, a white-bright
facet area was found in the vicinity of a non-metallic inclusion at
fracture origin inside the fish-eye zone by SEM observation. This
* Corresponding author. Tel./fax: +81 76 445 6775.
E-mail address: shiozawa@eng.u-toyama.ac.jp (K. Shiozawa).
son with the area inside the fish-eye [2–4,13], and the authors have
named this a ‘‘granular-bright-facet” (GBF) [2]. The formation of
the GBF area during the high cycle fatigue process controls the
internal fracture mode and is an important factor clarifying fatigue
behavior in the very high cycle regime, and ensuring the long dura-
bility of machine elements and structures [14–16]. Mechanism for
the GBF formation was discussed in detail through the experimen-
tal facts obtained by the rotary bending fatigue test and proposed
as ‘dispersive decohesion of spherical carbide’ model by the
authors [15,16]. It is important to confirm the justification of this
model under the other loading conditions.
Now, actual mechanical components of machine and structure
are operated under various type of service loading, such as rotary
bending, plane bending, axial loading, torsional loading and com-
bined these loadings. It is important for establishment of high
safety and reliability fatigue design method of machine elements
to obtain fundamental characteristic and data under each of the
above loading conditions. Experimental study for fatigue properties
under axial loading [17–19] is very few compared with that under
rotary bending. The aim of this study is to clarify the effect of ap-
plied stress ratio on fatigue behavior of high strength steel in very
high cycle regime. Axial loading fatigue tests were carried out un-
der three applied stress ratios using hourglass-shaped specimens of
high carbon–chromium bearing steel, and S–N characteristic and

0142-1123/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijfatigue.2008.11.001
 K. Shiozawa et al. / International Journal of Fatigue 31 (2009) 880–888
fatigue fracture mode were discussed based on the fracture
mechanics.
2. Experimental procedures
2.1. Testing material and specimen
The material used in this study was a high carbon–chromium
bearing steel, JIS SUJ2. The chemical composition (mass percent-
age) of this steel is shown in Table 1. Hourglass-shaped specimens
with a minimum diameter of 3 mm and round notch radius
of 15 mm with a flange, as shown in Fig. 1, were machined. The
specimens were heated for 2400 s in a vacuum at 1008 K and
oil-cooled, tempered for 1200 s in a vacuum at 453 K and then
air-cooled. The round notch surface was polished by an emery
paper having a mesh of #2000 and buff polishing. The elastic stress
concentration factor of this specimen is 1.029.
Fig. 2 shows the microstructure of the heat-treated material ob-
served by an SEM, which was prepared with micro-etching by Nital
(a) and electro-etching (b). It can be seen from the photographs
that many spherical carbide particles are distributed in the matrix
which is a tempered martensite structure composing of an average
prior-austenitic grain size of 5.7 lm. Size of the carbide particle
distributed in the range of 0.2–2.0 lm and the average was
0.8 lm. The tensile strength and the Vickers hardness of the
heat-treated material are 2316 MPa and HV749, respectively.
Table 1
Chemical composition of tested materials (mass%).
C Si Mn P S Cr Cu Ni
1.01 0.23 0.36 0.012 0.007 1.45 0.06 0.04
Fig. 1. Shape and dimension of specimen used.
Fig. 2. Microstructure of tested materials, JIS SUJ2.
881
2.2. Fatigue testing method
Fatigue tests were performed using a newly developed multi-
type fatigue testing machine in axial loading [20,21]. This testing
machine can simultaneously perform fatigue tests for four differ-
ent specimens at the frequency of 80 Hz, using a special hydraulic
system with rotary valve to distribute high-pressure oil into the
respective actuators. The loading capacity for each specimen is
±10 kN and the applied load can be given independently to the
individual specimen. In this study, force-controlled fatigue tests
under axially-reversed loading of sinusoidal wave type were car-
ried out at three applied stress ratios, R, of 1, 0 and 0.5 in an open
environment at room temperature.
3. Experimental results
3.1. S–N Curve
The S–N curve obtained from the axial loading fatigue tests un-
der different applied stress ratio is shown in Fig. 3a. As observation
result of the fracture surface by SEM, fatigue fracture mode was
classified into two types, such as surface fracture mode resulting
from the surface crack initiation and growth (referred to as S-
mode) and subsurface non-metallic inclusion-induced fracture
mode with a fish-eye. Also, the subsurface fracture mode was di-
vided two types of fracture mode; One was formed the GBF area
in the vicinity of a non-metallic inclusion at the fracture origin in-
side the fish-eye zone for the specimens ruptured at the high cycles
of a lifetime of more than 106 cycles (IG-mode), plotted by solid
Mo [O] mark in Fig. 3a. The other was not formed the GBF area at short
0.02 8 ppm lifetime bellow 106 cycles plotted by half-solid marks (I-mode).
Fig. 3b shows the comparison between the S–N curve obtained
from the axial loading fatigue test under the stress ratio of 1 and
that from the cantilever-type rotary bending fatigue test using
hourglass-shaped specimens with a minimum diameter of 3 mm
[2]. It can be seen from the figure that the S–N curve depends on
the applied loading mode. Even though the results obtained from
the rotary bending fatigue test shows the step-wise S–N curve or
a duplex S–N curve, shape of S–N curve obtained from the axial
loading fatigue tests is a smooth and continuous in spite of the
occurrence of the three types of fracture mode. The fatigue life in
surface-induced fracture mode can be found to have no difference
between loading modes. However, the fatigue life in subsurface-in-
duced fracture mode obtained from the axial loading fatigue test is
shorter than that from the rotary bending fatigue test. It is sug-
gested that a reason of the difference in lifetime between them is
882 K. Shiozawa et al. / International Journal of Fatigue 31 (2009) 880–888

Surface Surface Subsurce Subsurface 108

a 2000 inclusion inclusion GBF Axial loading fatigue
R=-1

Number of cycles to failure

R=0 107 S IG mode
MPa

R=0.5
1500
I IG mode
a,

106

Nf, cycles
Stress amplitude

Rotating bending fatigue

1000 5
10
SUJ2
4
500 10
S I mode SKH51

103
0 -1.5 -1 -0.5 0 0.5 1
103 104 105 106 107 108 109 Stress ratio R
Number of cycles to failure Nf, cycles
Fig. 4. Experimental results for the effect of applied stress ratio on the transition
life of fracture mode.
b 2000
SUJ2
Cantilever-type Fig. 5 shows typical examples of SEM observation for the IG-mode
rotary bending fatigue
MPa

Surface fracture on the fracture surface of the specimen obtained under the
Subsurface testing condition of three applied stress ratios. Distinctive feature
Subsurface GBF
a,

1500 of the fracture surface (GBF) was observed in the vicinity of a

non-metallic inclusion at the fracture origin inside the fish-eye
Stress amplitude

zone. Rough and granular morphology of the GBF area depends

1000
Axial-loading fatigue
Surface
Subsurce inclusion
Subsurface inclusion with GBF
500
103 104 105 106 107 108
Number of cycles to failure Nf, cycles
Fig. 3. S–N curve obtained from axial loading fatigue tests (a), and comparison of S–
N curve between axial loading fatigue (R = 1) and rotary bending fatigue (b).
ascribed to the difference in applied stress distribution across the
section of a specimen. That is, damage area of the specimen in axial
loading is larger than that in rotary bending and the maximum
inclusion size in the damage area is different between them from
probabilistic point of view.
From the S–N curve and observation of fracture surface, three
types of fracture modes were observed, such as S-, I- and IG-mode.
Fig. 4 shows the experimental relationship between fatigue life-
time for transition of fracture mode and applied stress ratio, R. In
this figure, the results obtained from specimen of high-speed tool
steel, JIS SKH51, are also shown [19]. It can be seen from this figure
that transition lifetime from I-mode to IG-mode does not depend
on the applied stress ratio below R = 0 and is about 106 cycles.
On the other hand, transition lifetime from S-mode to I-mode in-
creases with increasing R. At the high stress ratio of 0.5, fracture
mode changes from S-mode to IG-mode in long lifetime over 107
cycles.
3.2. Fractography
3.2.1. SEM observation of fracture surface
From SEM observations, three types of fracture mode, such as
S-, I- and IG-mode, were observed on the fracture surface of the
specimens tested under R = 1 and 0. On the other hand, two types
of fracture mode, S- and IG-mode, were observed under R = 0.5.
on the applied stress ratio and tends to be obscurely at higher
stress ratio.
3.2.2. Quantitative evaluation of the crack initiation site
Various dimensions of the crack initiation site on the fracture
surface were measured by SEM. Fig. 6a shows the experimental
relationship between the size of the non-metallic inclusion at crack
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
109 initiation site, areainc , and the number of cycles to failure, Nf. The
size of the inclusion was scattered over the range between 9 and
43 lm, and was independent both of fatigue lifetime and stress ra-
tio, R. The dispersion in size of the inclusion was larger than that
obtained from the cantilever-type rotary bending fatigue test, 7–
13 lm, indicated in this figure [2]. Fig. 6b shows the relationship
between distance from specimen surface to crack initiation site,
dinc, and Nf. Value of dinc was distributed widely throughout 40–
1240 lm, and identified no correlation with Nf and R. The crack ini-
tiation site in the axial loading fatigue distributed widely on the
across the specimen section, compared with that in rotary bending
fatigue, 20–240 lm, limited on the surface layer of the specimen.
From these experimental facts, the difference of the S–N curve be-
tween axial loading fatigue and rotary bending one was caused by
difference in the stress distribution across the specimen section be-
tween them. That is, a uniform stress distribution was applied
across the section in the axial loading fatigue. On the other hand,
in case of the rotary bending samples, an applied stress gradient
across the section exists and the stresses gradually decrease from
the maximum stress at the specimen surface to zero at the neutral
axis. Therefore, the damage area for fatigue crack initiation in the
specimen applied axial loading is larger than that in rotary
bending.
3.2.3. Detail observation of the GBF area
GBF area was observed in the vicinity of the non-metallic inclu-
sion on the fracture surface obtained from the fatigue test more
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
than 106 cycles. Size of the GBF area, areaGBF , which includes
the size of the inclusion, was measured and results are shown in
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Fig. 7. In this figure, relationship between ratio of areaGBF with
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
areainc;i and Nf is also shown. The GBF area increases with in-
crease of fatigue lifetime and is independent on stress ratio. It
 K. Shiozawa et al. / International Journal of Fatigue 31 (2009) 880–888 883

SUJ2
GBF a 50

m
Surface Subsurface
inclusion inclusion
40 R=-1

area inc,s, area inc,i,

Size of inclusion
R=0
R=0.5
30

20

10
Rotary bending fatigue (Subsurface inclusion)
20µm 0
10 3 10 4 10 5 10 6 10 7 10 8 10 9
(a) R= 1, a=850MPa, Nf=1.32 108 cycles Number of cycles to failure Nf, cycles

SUJ2
b 1500
Rotary bending fatigue

m
R=-1
R=0
R=0.5

Depth of inclusion d inc,

1000

500

240 m
20µ
0 20 m
10 3 10 4 10 5 10 6 10 7 10 8 10 9
(b) R=0, a=525MPa, Nf=1.03 108 cycles Number of cycles to failure Nf, cycles
Fig. 6. Experimental relationship between size of inclusion (a) and depth of
inclusion from specimen surface (b), and number of cycles to failure.

ture process using the FRASTA (fracture surface topography analy-

sis) method, the authors have proposed the mechanism of GBF
formation in a very high cycle fatigue regime as the ‘‘dispersive
decohesion of spherical carbide” model [15,16]. Fig. 8 shows a car-
bon distribution in the vicinity of a non-metallic inclusion (TiC) on
the fracture surface tested under R = 0 and 0.5 by EPMA. It can be
seen from the figure that rich carbon distribution is detected in
the GBF area compared with outside the GBF. It is noted that the
rich carbon distribution is detected on an unclear GBF area of frac-
20µm ture surface tested under R = 0.5.

(c) R=0.5, a=390MPa, Nf=1.13 108 cycles

4. Discussions
Fig. 5. Typical SEM observation around a subsurface inclusion on fracture surface.
4.1. Stress intensity factor at the crack initiation site

Based on the inclusion size at the crack initiation site obtained

has been reported in previous papers by the authors [15,16], based
on the experimental results obtained from cantilever-type rotary
bending fatigue tests that the GBF area had three specific charac-
teristics: That is, (1) surface of the GBF area is very rough compared
with the surface outside it in the fish-eye zone, (2) the surface
roughness in the GBF area is related to a size of spherical carbide
particle in the microstructure of the tested materials, and (3) rich
carbon distribution in the GBF area is detected by EPMA (electron
probe micro analyzer), compared with that outside the GBF. From
these facts and the computational simulation of the interior frac-
from the fractography, the initial stress intensity factor range,
DK inc , was calculated using the following formula [22]:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
DK inc;s ¼ 0:65Dr p areainc;s for surface inclusion ð1Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
DK inc;i ¼ 0:5Dr p areainc;i for internal inclusion ð2Þ
where Drð¼ rmax  rmin Þ is the total stress range. Fig. 9a shows the
relationship between DK inc and Nf. In addition, DK GBF calculated
using the size of GBF area was shown in this figure. The values of
884 K. Shiozawa et al. / International Journal of Fatigue 31 (2009) 880–888

SUJ2
a 100
Axial loading fatigue SEM SEM
80
R=-1 Rotary bending fatigue
R=0
R=0.5
m

60
area GBF,

40
20µm 20µm
20
EPMA EPMA
0
106 107 108 109
Number of cycles to failure Nf, cycles

b 8
SUJ2
Axial loading fatigue
C C
m

R=-1 Rotary bending fatigue

6 R=0
area inc,i,

R=0.5

4
area GBF/

0
Ti Ti
106 107
10 8
10 9
(a) R= 0, a=550MPa (b) R= 0.5, a=380MPa
Number of cycles to failure Nf, cycles
Nf =5.11 107 cycles Nf =7.12 107 cycles
Fig. 7. Experimental relationship between size of GBF area (a) and the ratio of GBF
area size to inclusion size to inclusion size (b), and number of cycles to failure. Fig. 8. EPMA analyzed results around an inclusion on fracture surface.

DK inc;s and DK inc;i decrease for increasing Nf, and depends on the ap-
plied stress ratio. To evaluate the effect of the stress ratio on crack
initiation at an inclusion, modified stress intensity factor range,
DK inc and DK GBF , was calculated using a stress amplitude of tensile
range, Dr (for R = 1; rmax ¼ ra , for R = 0 and 0.5; rmax  rmin ¼
2ra ), instead of the value of Dr. The results are shown in Fig. 9b.
It can be see from this figure that DK inc;s and DK inc;i do not depend
on the applied stress ratio and decrease continuously with increas-
pffiffiffiffiffi
ing Nf. The values of DK inc;i are below 4 MPa m for the specimens
ruptured in the range of long fatigue life over 106 cycles and by the
IG-mode fracture. On the other hand, the DK inc;i are above
pffiffiffiffiffi
4 MPa m for the specimen ruptured by the I-mode. This value
may be the limit, DK th , to bring the I-mode cracking. Below the
DK th , the GBF area is formed in the vicinity of non-metallic inclu-
sion during fatigue process and the fracture is occurred by the
pffiffiffiffiffi
IG-mode. The values of DK GBF are in the range of 4–6 MPa m, not
depended on applied stress ratio, and above the value of DK th .
Therefore, it is expected that the internal fracture mode initiated
from an inclusion in very high cycle fatigue regime is controlled
by the formation of the GBF in the vicinity of the inclusion. This
mechanism is the same as the result of a rotary bending fatigue
and does not depend on the applied stress ratio. From above dis-
cussions, the mechanical factor to control the internal crack initia-
tion, propagation and fracture will be tensile stress amplitude, that
is, equivalent to the effective stress intensity factor range.
4.2. Effect of applied stress ratio on the roughness of GBF area
Fig. 10 shows a typical example for the birds-eye pattern of the
GBF area observed by scanning probe microscope (SPM). Some
granular convex particles can be seen on the fracture surface of
the each specimen ruptured under the stress ratio of 1, 0 and
0.5. It has been reported by the previous paper [15,16] that the for-
mation of convex and rough surface on the GBF area related to the
carbide particles in the microstructure of the specimen.
Surface roughness measurements of the fracture surface were
taken by three-dimensional SEM analysis. The roughness of the
GBF area was 2–2.5 times greater than that outside the GBF area
in fish-eye zone. Fig. 11a shows values of the surface roughness,
Rmax and Ra, on the GBF area for every specimen obtained from
each applied stress ratio. An average distance of roughness on sur-
face roughness curve, S [see Ref. [15] for detail definition], is also
shown in this figure. The Rmax and Ra decreased and the S increased
with increasing applied stress ratio. Fig. 11b shows the relation-
ship between the values of Rmax, Ra and S, and stress intensity fac-
tor range for subsurface inclusion, DK inc;i . These values depend
only on DK inc;i and do not depend on the applied stress ratio.
Roughness of the GBF area increases for increasing DK inc;i . It was
pointed out in previous papers by authors [15,16] that the rough-
ness on GBF area causes from decohesion of spherical carbide in
the matrix. Therefore, it is suggested that the decohesion of spher-
 K. Shiozawa et al. / International Journal of Fatigue 31 (2009) 880–888 885

a 20
SUJ2 a 2
SUJ2
Kinc,s, Kinc,i, KGBF, MPa m

Surface roughness on GBF area

S
Stress intensity factor range

1
10

Ra, Rmax and S, m

Rmax
0.5
5 Rotary bending fatigue

Surface Subsurface GBF

inclusion inclusion Ra
R=-1 0.1
R=0
R=0.5 Axial-loading fatigue
0.05
1 -1.5 -1 -0.5 0 0.5 1
103 104 105 106 107 108 109
Stress ratio R
Number of cycles to failure Nf, cycles
b 2
SUJ2

Surface roughness on GBF area

SUJ2 S
b 10
m

Rotary bending fatigue 1

Ra, Rmax and S, m

K*inc,s, K*inc,i, K*GBF, MPa

Subsurface inclusion
Stress intensity factor range

GBF 0.5 R max

5
4
Ra
Axial loading fatigue R=-1
0.1
2 Surface Subsurface GBF R=0
inclusion inclusion R=0.5
R=-1
R=0 0.05
R=0.5 1.5 2 3 4 5 6 7 8 9
Stress intensity factor range Kinc,i, MPa m
1
103 4
10 10 5
10 6
10 7
10 8
10 9
Fig. 11. Effect of applied stress ratio (a) and stress intensity factor range for
subsurface inclusion (b) on surface roughness on the GBF area.
Number of cycles to failure Nf, cycles
Fig. 9. Stress intensity factor range at inclusion and GBF area relating to number of
cycles to failure. 4.3. Simulation of fracture process by the FRASTA method

The fracture surface topography analysis (FRASTA) method is a

ical carbide occurs in the cyclic plastic zone formed around an procedure for computationally reconstructing the process of crack
inclusion depending on stress intensity factor range, and with growth in microscopic detail by comparing topographic features of
inhomogeneous deformation between an inclusion and surround- a conjugate area of opposing fracture surfaces [23]. Fig. 12 shows a
ing matrix. series of fracture-area projection plots (FAPPs) generated by the

µm µm µm µm µm µm

(a) R= -1, a =850 MPa (b) R= 0, a =525 MPa (c) R= 0.5, a =390 MPa
8 8 8
Nf = 1.32 10 cycles Nf = 1.03 10 cycles Nf = 1.13 10 cycles
Fig. 10. SPM observation of the GBF area on fracture surface of specimen obtained from fatigue tests under stress ratio of 1, 0 and 0.5.
886 K. Shiozawa et al. / International Journal of Fatigue 31 (2009) 880–888

SEM (1) 0.531 µ m (2) 0.594 µ m ( 3 ) 0.656 µ m

10 µm

(a) R=0, IG-mode

SEM (1) 0.739 µ m ( 2 ) 0.933 µ m ( 3 ) 1.283 µ m

5 µm

(b) R=0.5, IG-mode

Fig. 12. A series of fracture-area projection plots (FAPPs) generated by the FRASTA method.

FRASTA method for specimens tested under R = 0 and 0.5, and
ruptured by the IG-mode. In these plots of the figure, the dark areas
indicate the unfractured areas and white areas are the fractured
ones. The number in each plot of the figure gives the topographic
displacement relative to the reference state, which corresponds
to fatigue process. At the stage of a small displacement, small
spotty white area can be seen in the vicinity of an inclusion. These
areas are microcracks initiated during fatigue cycling. As the sepa-
ration between the conjugate maps increased, the multiple micro-
cracks increased in size and number, and then coalesce with each
other into the GBF area. The phenomena were also observed on
the specimen tested under R = 1 and same as the results of rotary
bending fatigue reported previous paper [15,16].
Fig. 13 shows cross-sectional plots (XSPs) of crack growth ob-
tained by displacing topographic contours along the white line in
Fig. 12. The darkly shaded areas are areas where the two topo-
graphic surfaces overlap, and indicate that plastic deformation
formed at a crack front during crack propagation. In these views,
the white area is the fractured area. From this figure, microcracks
initiate at a short distance from an inclusion and different points.
Based on the experimental results and computational simula-
tion of the fracture process by the FRASTA method, a mechanism
for formation of the GBF area is explained by the ‘‘dispersive dec-
ohesion of spherical carbide” model not only under the condition
of the stress ratio R = 1 but also R = 0 and 0.5. Outline of the
mechanism is as follows; initially, multiple microcracks are disper-
sively initiated by decohesion of spherical carbide from the matrix
in the vicinity of a non-metallic inclusion. The decohesion occurs
with dislocations pilled-up to around spherical carbide particles
in the zone of stress concentration due to an inclusion. These
microcracks grow and coalesce with each other during fatigue cy-
cles. At this time, the microcracks propagate along a boundary be-
tween spherical carbide and the matrix, and therefore the
roughness of the size of the carbide grade is generated on the frac-
ture surface. After formation of an appropriate size for the GBF area
to enable propagation as an ordinary crack, the crack propagates
without depending on the microstructure of the materials and
forms a fish-eye.
4.4. Effect of residual stress on transition of fracture mode
Three types of fracture mode, such as S-, I- and IG-mode, were
observed on the S–N curve. In this section, transition of the fracture
mode was discussed from a viewpoint of the fracture mechanics.
Initial stress intensity factor range of surface inclusion, DK inc;s ,
and subsurface inclusion, DK inc;i , are calculated as Eqs. (1) and
(2), respectively, based on the discussion mentioned above section.
The value of DK inc;s is usually larger than that of DK inc;i , if both size
of inclusion are same. Fig. 14 shows a schematic relation between
the DK inc;s indicated as dot-dashed line and DK inc;i as dashed line,
and applied stress amplitude. Above the threshold stress intensity
factor range of surface crack growth, DKth, crack will initiate at a

5µm 5µm

0.581µm 1.219µm
(a) R=0, IG-mode fracture (b) R=0.5, IG-mode fracture
Fig. 13. Analytical results for a series of cross-sectional plots (XSPs) generated by the FRASTA method.
 K. Shiozawa et al. / International Journal of Fatigue 31 (2009) 880–888 887

JIS SUJ2-Axial loading

IG-mode S-mode 1600
A areainc =17.0 m
Stress intensity factor range

Surface crack, r =0 1400 * I IG

MPa
K*inc,s r=-373MPa
m

B 1200 *
K th= K'th C

a,
**
S I
Kinc, MPa

GBF formation Surface crack, r<0 1000

KTinc,s
IG-mode I-mode S-mode
Internal crack S : Surface fracture mode
*
K inc,i I : Subsurface fracture mode without GBF
IG : Subsurface fracture mode with GBF
Stress amplitude a, MPa
Fig. 14. Schematic diagram to explain the transition of fracture mode.
surface inclusion and propagate to final fracture. Below the DKth,
subsurface crack can start to grow after forming a GBF area in
the vicinity of subsurface inclusion. Therefore, surface fracture
mode, S-mode, occurs in the region of high stress amplitude level
and changes into an internal fracture mode with GBF area, IG-
mode, at the low stress amplitude level (point C in Fig. 14). In this
case, only two fracture modes appear on a S–N curve and an inter-
nal fracture mode without GBF area, I-mode, does not occur at any
stress amplitude level against the experimental facts.
In order to explain an occurrence of three fracture modes on a
S–N curve, an effect of residual stress on surface layer of specimen
will be considered for the above discussion. It is assumed that the
residual stress, rr, acts on the fatigue behavior as a mean stress ef-
fect, and stress range in tension side affects the crack initiation and
propagation, according to the experimental facts mentioned above
section. Based on the assumption, initial stress intensity factor
range of the surface inclusion, DK inc;s , is rewritten to DK Tinc;s using
DrT defined by Eq. (3) instead of Dr*.
 that there is not the residual stresses on the specimen surface, the
rmax  rmin for rmin P jrr j
DrT ¼
rmax þ rr for rmin < jrr j
In addition, it is considered that the residual stress in surface
layer affects only surface inclusion and does not affect on subsur-
face inclusion. Therefore, initial stress intensity factor range of
subsurface inclusion, DK inc;i , does not change without regard to
existence of residual stress on surface layer.
As schematically shown in Fig. 14, the value of DK Tinc;s indicated
as solid line becomes to small by the effect of compressive residual
stress comparing with the DK inc;s , relating with a constant applied
stress amplitude. The value of DK Tinc;s against a stress amplitude
intersects the DK inc;i at a point A in Fig. 14, and the fracture mode
changes from S-mode to I-mode at the stress amplitude level. Be-
cause DK inc;i at the point A is above the DK 0th , a subsurface crack
can initiate and propagate without forming the GBF area around
the inclusion. Where, DK 0th is a threshold stress intensity factor
range for subsurface crack growth without formation of a GBF area
and same as the DK th from the experimental fact. The line of DK inc;i
is correspondent to DK 0th ð DK th Þ at a point B with decreasing
stress amplitude. Below this corresponding stress amplitude, a
subsurface crack cannot directly initiate and propagate from an
inclusion, and the formation of GBF area around an inclusion is re-
quired for initiation and propagation of subsurface crack, that is,
IG-mode fracture appears. Therefore, due to the existence of the
compressive residual stress on the specimen surface layer, three
types of fracture modes will be occurred depending on the applied
stress amplitude.
Stress amplitude
800 * S IG
r =0
600 S IG *
Experimental results
400 S I *
I IG
200 * T. Sakai(Ritsumeikan Univ.)
S IG ** M. Nakajima(TCT)
0
-1.5 -1 -0.5 0 0.5 1
Stress ratio R
Fig. 15. Diagram for stress amplitude on the transition of fracture mode relating to
applied stress ratio, the comparison of calculated and experimental results.
Stress amplitude occurred the transition of fracture mode, such
as from S-mode to I-mode and from I-mode to IG-mode, was calcu-
lated for each stress ratio using the compressive residual stress on
the specimen surface of 373 MPa measured by X-ray diffraction
method, an average inclusion size of 17.0 lm and DK th ¼ DK 0th ¼
pffiffiffiffiffi
4 MPa m. The classical sin2 w method was applied to the determi-
nation of residual stress, using the diffraction pattern of the (2 1 1)
crystal plane by Cr Ka. Solid lines in Fig. 15 are the calculated re-
sults of stress amplitude for transition of fracture mode relating
to the stress ratio. From this figure, stress amplitude of the transi-
tion from S-mode to I-mode and I-mode to IG-mode are clearly
separated each other, and decreased with increasing the stress ra-
tio below about zero. On the other hand, at the range of higher
stress ratio above zero, fracture mode changes from the S-mode
to the IG-mode and I-mode fracture does not occur, because of
rmin P jrr j and no effect of residual stress on the fracture. In case
ð3Þ transition of fracture mode from S-mode to IG-mode occurs at low
stress amplitude level, indicated as fine dot-dashed line in Fig. 15.
The experimental results obtained from this study and reported by
other researchers [18] are shown in this figure. It can be seen from
this figure that agreement of predicted results and the experimen-
tal results is fairly good.
5. Conclusions
1. Three types of fracture modes, such as surface-induced fracture
mode, subsurface inclusion-induced fracture mode without for-
mation of GBF area and subsurface inclusion-induced fracture
mode with formation of GBF area, were observed on fracture
surfaces obtained under an axial loading fatigue. The fatigue
lifetime at which different fracture modes appear depends on
the applied stress ratio.
2. S–N curve under the axial loading fatigue tests showed a
smooth and continuous shape under three testing conditions
in spite of the fact the fracture mode changes with depending
on applied stress. This was different from the clear duplex S–
N curve obtained from a rotary bending fatigue test.
3. Fatigue fracture in very high cycle fatigue regime was con-
trolled by formation of the GBF in the vicinity of non-metallic
inclusion without regard to applied stress ratio. The GBF area
couldn’t be observed clearly at higher applied stress ratio by
SEM, but large roughness and rich carbon distribution was
detected in the vicinity of subsurface inclusion.
888 K. Shiozawa et al. / International Journal of Fatigue 31 (2009) 880–888
4. Transition of the fracture mode was discussed with considering
the effect of residual stress on the specimen surface layer. The
relation between the predicted stress amplitude for the transi-
tion of fracture mode and the experimental results showed
good agreement.
Acknowledgement
This work was supported by Grant-in-Aid for Scientific Re-
search (C), No. 19560080, Japan Society for the Promotion Science.
References
[1] First international conference was held as ‘‘Fatigue life in the gigacycle
regime”, 1998 in Paris. The second: ‘‘Fatigue in the very high cycle regime”,
2001 in Vienna, the third: ‘‘Very high cycle fatigue (VHCF-3)”, 2004 in Kyoto/
Kusatsu, and the fourth: ‘‘Very high cycle fatigue (VHCF-4)”, 2007 in Ann
Arbor, Michigan.
[2] Shiozawa K, Lu L, Ishihara S. S–N curve characteristics and subsurface crack
initiation behavior in ultra-long life fatigue of a high carbon–chromium
bearing steel. Fatigue Fract Eng Mater Struct 2001;24(12):781–90.
[3] Sakai T, Takeda M, Tanaka N, Kanemitsu M, Oguma N, Shiozawa K. S–N
property and fractography of high carbon chromium bearing steel over ultra
wide life region under rotating bending. Trans Jpn Soc Mech Eng
2001;67A(663):1805–12.
[4] Ochi Y, Matsumura T, Masaki K, Yoshida S. High-cycle rotating bending fatigue
property in very long-life regime of high-strength steels. Fatigue Fract Eng
Mater Struct 2002;25(8/9):823–30.
[5] Nishijima S, Kanazawa K. Stepwise S–N curve and fish-eye failure in gigacycle
fatigue. Fatigue Fract Eng Mater Struct 1999;22:601–7.
[6] Sakai T, Takeda M, Shiozawa K, Ochi Y, Nakajima M, Nakamura T, et al.
Experimental reconfirmation of characteristic S–N property for high carbon-
chromium bearing steel in wide life region in rotating bending. J Soc Mater Sci
Jpn 2000;49(7):779–85.
[7] Shiozawa K, Nishino S, Ohtani T, Mizuno S. Subsurface fatigue crack initiation
and growth of plasma-assisted duplex surface-treated tool steel. In:
Ravichandran KS, Ritchie RO, Murakami Y, editors. Small fatigue cracks.
Mechanics, mechanisms and applications. Amsterdam: Elsevier Science; 1999.
p. 39–47.
[8] Shiozawa K. Characteristics of fatigue behavior of surface modified materials.
In: Nishida S-I, editor. Macro and microscopic approach to fracture. UK: WIT
Press; 2003. p. 117–70.
[9] Lu L, Shiozawa K, Morii Y. Gigacycle fatigue behavior and fracture morphology
of high speed tool steel, JIS SKH 51. Trans Jpn Soc Mech Eng
2003;69A(679):662–70.
[10] Shiozawa K, Lu L. Very high-cycle fatigue behavior of shot-peened high
carbon–chromium bearing steel. Fatigue Fract Eng Mater Struct 2002;25(8/
9):813–22.
[11] Lu L, Shiozawa K. Effect of two-step load variation on super-long life fatigue
and internal crack growth behavior of high carbon–chromium bearing steel.
Trans Jpn Soc Mech Eng 2002;68A(675):1066–73.
[12] Lu L, Shiozawa K. Effect of surface treatment on super-long-life fatigue
behavior in high speed tool steel, JIS SKH51. Trans Jpn Soc Mech Eng
2003;69A(684):1195–202.
[13] Murakami Y, Nomoto T, Ueda T, Ohtori M. Analysis of the mechanism of super
long fatigue failure by optical microscope and SEM/AFM observations. J Soc
Mater Sci Jpn 1999;48(10):1112–7.
[14] Murakami Y, Nomoto T, Ueda T. Factors influencing the mechanism of
superlong fatigue failure in steels. Fatigue Fract Eng Mater Struct
1999;22:581–90.
[15] Shiozawa K, Morii Y, Nishino S, Lu L. Subsurface crack initiation and
propagation mechanism in high-strength steel in a very high cycle fatigue
regime. Int J Fatigue 2006;28(11):1520–32.
[16] Shiozawa K, Morii Y, Nishino S. Subsurface crack initiation and propagation
mechanism under the super-long fatigue regime for high speed tool steel (JIS
SKH51) by fracture surface topographic analysis. JSME Int J 2006;49A(1):1–10.
[17] Sakai T, Sato Y, Nagano Y, Takeda M, Oguma N. Effect of stress ratio on long life
fatigue behavior of high carbon chromium bearing steel under axial loading.
Int J Fatigue 2006;28(11):1547–54.
[18] Sakai T, Takizawa R, Furusawa T, Nakajima M, Shiozawa K, Oguma N.
Reliability evaluation of SUJ2 in long life regime by the multi-type axial
fatigue testing machine. In: Proceedings of the 21st National Symposium on
Reliab Mater Struct; 2005. p. 75–80.
[19] Shiozawa K, Nishino S, Yachi Y. Effect of stress ratio on subsurface fatigue
fracture of high-speed tool steel, SKH51. Trans Jpn Soc Mech Eng
2006;72(720):1153–60.
[20] Sakai T, Furusawa T, Takizawa R, Nakajima M, Shiozawa K, Oguma N et al.
Development of multi-type testing machine in axial loading and some fatigue
test results on a bearing steel. In: Proceedings of the 3rd International
Conference on Very High Cycle Fatigue; 2004. p. 484–91.
[21] Furusawa T, Sakai T, Takizawa R, Nakajima M, Shiozawa K, Oguma N, et al.
Development of multi-type testing machine in axial loading and verification of
its fundamental performance. Trans Jpn Soc Mech Eng 2006;70(718):886–94.
[22] Murakami Y, Kodama S, Konuma S. Quantitative evaluation of effect of
nonmetallic inclusions on fatigue strength of high strength steel. Trans Jpn Soc
Mech Eng 1988;54(500):688–95.
[23] Kobayashi T, Schockey DA. FRASTA: a new way to analyze fracture surfaces.
Adv Mater Process 1991;140(5):28–34.