Copyright © 1996 ASM International®

ASM Handbook, Volume 19: Fatigue and Fracture All rights reserved.
ASM Handbook Committee, p 331-336 www.asminternational.org

Contact Fatigue
W.A. Glaeser a n d S.J. S h a f f e r , B a t t e l l e Laboratories

CONTACT FATIGUE is a surface-pitting-type the direction of ball travel. Not all spalls in ball- automobile engine tests (Ref 3). Lifters were
failure commonly found in ball or roller bearings. bearing races are of the shape shown in Fig. 2. nodular iron, and cams were flake graphite cast
This type of failure can also be found in gears, Figure 3 shows a fatigue spall near the race iron. Fatigue cracks were associated with cracked
cams, valves, rails, and gear couplings. Contact shoulder of a deep-groove ball bearing. The spall carbides, graphite flakes, and hard inclusions.
fatigue has been identified in metal alloys (both appears to have been formed by the joining of Contact fatigue occurs in gears along the pitch
ferrous and nonferrous) and in ceramics and cer- several pits. The fact that the spall occurred close line. The geometry of tooth mesh is such that
mets. to the race shoulder may have distorted the con- rolling occurs at the pitch line while sliding oc-
Contact fatigue differs from classic structural tact state of stress, causing a multiple origin. curs at the addendum as the gears come out of
fatigue (bending or torsional) in that it results Fatigue in roller bearings may differ from ball- mesh. An example of pitch line contact fatigue is
from a contact or Hertzian stress state. This local- beating contact fatigue. Quite often the pitting shown in Fig. 5 (Ref4). The pits seen on the teeth
ized stress state results when curved surfaces are occurs in the inner race at the contact zone of the will grow in size and depth, ultimately resulting
in contact under a normal load. Generally, one roller ends. In some cases, contact stress peaks at in tooth fracture.
the roller ends and pitting originates in these Another form of contact fatigue, known as mi-
surface moves over the other in a rolling motion
locations. Roller-end pitting can be a sign of cropitting, occurs in bearings. An example is
as in a ball rolling over a race in a ball bearing.
misalignment. shown in Fig. 6. This feature can show up over
The contact geometry and the motion of the roll-
C a m s and Gears. Valve lifter cams and rollers the entire raceway surface. It is often the result of
ing elements produces an alternating subsurface
are subject to contact fatigue. An example is too thin a lubricant film or excessive surface
shear stress. Subsurface plastic strain builds up shown in Fig. 4 (Ref 3). The character of the roughness and sometimes heavy loading.
with increasing cycles until a crack is generated. damage is very similar to that found in rolling In gears, micropitting is termed frosting and in
The crack then propagates until a pit is formed. contact bearings. The example shown in Fig. 4 the present ANSIdAGNA standard it is consid-
Once surface pitting has initiated, the bearing was found in both cam nose and lifters during ered a form of contact fatigue. For bearings,
becomes noisy and rough running. If allowed to
continue, fracture of the rolling element and cata-
strophic failure occurs. Fractured races can result
from fatigue spalling and high hoop stresses.
Rolling contact components have a fatigue life
(number of cycles to develop a noticeable fatigue
spall). However, unlike structural fatigue, contact
fatigue has no endurance limit. If one compares
the fatigue lives of cyclic torsion with rolling
contact, the latter are seven orders of magnitude
greater (Ref 1). Rolling contact life involves ten
to hundreds of millions of cycles.

Examples of Contact Fatigue

Contact fatigue produces a surface damage that
is unique and well recognized. Familiar examples
are found in fatigue of ball and roller bearings. A
typical failure in a roller bearing is shown in Fig.
1. Although this spall is small, it would grow in
size until roller fracture would occur, as bearing
operation continues.
One classic shape of a fatigue spall in a ball
bearing is a delta shape, as shown in Fig. 2 with
a diagram of the pit (Ref 2). The apex of the pit is
the initiation point, usually the location of a sur-
face defect like a dent. The pit grows in a fan
shape, becoming wider and deeper as it grows in
Fig. 1 Scanning electron micrograph of a fatigue spall on a roller from a roller bearing after 630,000 cycles. Roller is AISI
] 060 steel, hardened to 600 HV. Spall is 400 ltm wide by 700 p.m long.
332 / Fatigue Strength Prediction and Analysis

X
; ¸¸¸¸¸¸¸¸¸%¸~ / : ..... ~ ::~ ~:~
i
J

c ~

//•
i
i

Spall depth in
o
i

AB C D E F G 0.001 in.
. . . .

i I I I i I I
0 20 40 60 80 100 120 = ....
Spall length in 0.001 in. ',
F =

la)
i
i
Fig, 3 Multiple spall near a race shoulder

i
i

×
i
maximum shear stress depth is about the same as
the dimension of B. However, with increased
(hi traction or tangential force, the maximum shear
F i g . 2 Anatomy of a race spall in a ball bearing. (a) lypical delta shape with the apex at the origin. (b) Profiles of the spalL stress moves closer to the surface. In hardened
Source: Ref 2 martensitic steels used in ball and roller beatings,
the subsurface shear stresses produce plastic de-
formation in the martensitic structure. The resid-
ual strain increases with increase in rolling cy-
gears, and any contact, micropitting can be re- Analysis of the subsurface stress state indicates
cles. This has been shown by x-ray measurements
duced by improved surface finish, reduced tem- that a maximum shear stress exists at a given
during rolling contact experiments (Ref 7).
perature or loads, and providing sufficient elasto- depth below the surface. The stress distribution is
shown in Fig. 8. The maximum shear stress is Many researchers have studied the microstruc-
hydrodynamic film.
shown increasing with depth below the surface as tural changes that occur as a result of the buildup
Rails. Spalling and "shelly"-type failures occur
of subsurface strain (Ref 8-10). In AISI 52100
on track rails from wheel-track rolling contacts. discussed by Kloos and Schmidt (Ref 6).
steel, a common rolling-contact-bearingmaterial,
An example of shelly failure is shown in Fig. 7 The curves shown are based on two different
the accumulation of strain initially is associated
from Kilburn (Ref 5). The name comes from the mathematical approaches to the estimation of
contact state of stress. Both approaches produce with the formation of a dark etching Zone below
morphology of the fracture surface in the bottom the surface. Further swain causes the formation of
of the spall. Shelly failures are serious because a shear stress distribution quite close to each
light etching bands caused by the formation of a
they lead to rail fracture and derailments. Rail other. The z axis scale is Z/B, where B is the minor
new ferrite phase. Then carbides in the high stress
spalling has been reduced in recent years by the axis length of the contact ellipse and the usual
region begin to show decay and break up. Other
use of higher carbon steels for rails. direction of rolling motion. Note in Fig. 8 that the
microstructural features include "butterflies" or

Mechanisms of Contact Fatigue

The state of stress produced by rolling contact
is concentrated in a small volume of material and
produces intense plastic strain. The strain accu-
mulates as the same volume is stressed with each
rolling cycle until a crack is initiated and forms a
spall. In the real world of contact fatigue, the
mechanisms involved can be quite complex.
Most models assume a condition of ideal geomet-
ric surfaces and little input by heat generation,
environmental conditions, and inhomogeneities
of materials. Hertz stress analysis assumes a cir-
cular, elliptical, or line contact surface area be-
tween curved surfaces (depending on the geome-
try of the contacts) and a parabolic pressure
distribution with the maximum pressure at the
center of the contact.
The subsurface state of stress involves a hydro-
static component that inhibits tensile fracture. Fig. 4 Contact fatigue spalling of cam lifter surface, Source: Ref 3
 Contact Fatigue / 333

Fig. 5 Pitch line spalling of medium-hardenedgears.Source: Ref4

(a) (b)

Fig, 6 Micr•pitting•fr•••erbearingouterrace•Scanninge•ectr•nmicr•graph•(a)57xand(b)athighermagnificati•n

Large shell evidenced

by black spot Shell with
piece broken out
Gage

Fig. 7 Shelly rail spall from wheel-rail contact fatigue. Source:Ref 5

334 / Fatigue Strength Prediction and Analysis

z
3D contact pressure distribution for curved contacts

Oo

Surface 0.2XTM. 0.6 0.8 1.0

1.0 - 2~Max shear

stress Deve oping span.(a)Top view of developingspall at racesurfacedent. (b) Sectionthrough developing spall show-
Fig. 9
2.0- ing subsurfacecracking Source:Ref 4

Fig. 8 Stressdistributionsat a contact subsurface

contact fatigue. With shear stresses higher and
closer to the surface, surface defects (dents,
scratches, etc.) all contribute to higher incidence
of surface-originating fatigue. Figure 2 shows a
race spall that started at a dent in the race. This
c
=
"g"
.oL /
95

60 . . . . . . . . . . . . . . . . . . . . . .

white etching wings radiating out from large hard produced a delta-shaped spall as the cracking 401" / o
inclusions. Some features depend on the level of progressed from the origin. Sections through the
spall show it to be shallow at the origin and
contact stress. For example, butterflies are often
associated with high contact stress (2000 to 4000 deeper at the other end. Photomicrographs of a "6 20 , 7 ,
MPa), as discussed in more detail in the article developing spall (Fig. 9, Ref 4) caused by a dent ~
"Contact Fatigue of Hardened Steel" in this Vol- shows a ridge between the dent and the crack. Ii ..... / LIO . . . .
ume. This is typical and causes disruption in the oil ~.
In a detailed study of butterfly formations in film. The arrow shows the direction of movement
AISI 52100 steel ball bearings by Becker (Ref of the balls over the race. The section through the
11), both through-hardened AIS152100 steel and developing spall shows the subsurface crack V" I I I I I I I
carburized SAE 8620 steel bearing races were propagating down into the race at an angle to the 2 4 6 10 20 40 60 100 200
used. Contact stress was 3280 MPa (480 ksi). surface. Bearing life (in L), inner-race revolutions × 106
Butterflies were found in sectioned posttest races. As the developing spall matures, a surface
layer loosens and eventually breaks out, leaving Fig. 10 Weibull plot of ball-bearing lives distribution.
They were always oriented at about 40 ° to the Source:Ref 12
surface and oriented in the rolling direction. The a pit. While the pit develops, the loose layer
"wings" of the butterflies were found to be com- batters the fracture surface, obliterating the sur-
posed of a mix of heavily strained, ultrafine- face features. Fractographic analysis is not a
grained ferrite and fine carbide particles. Hard- likely option for investigating contact fatigue.
Two life values in the distribution are shown.
ness was measured as close to 1000 HV--harder The L10 life, or the life at which 10% of all the
than the martensitic matrix surrounding the but- bearings have failed is used for bearing selection.
terflies. Fine cracks were also found on the edges Rolling Contact Bearing Life Lundberg and Palmgren (Ref 13) developed a
of the butterfly wings. The same structures were relationship that can be used to predict bearing
found in the carburized case in the 8620 steel. Ball and roller bearings have been subject to life for any load, using the life for standard load
Becker says in Ref 1l: "The breakdown of the the most extensive life testing of all contact fa- in the relation:
matrix microstructure to ferrite and carbide is tigue components. Bearing catalog lives are
caused by very high stress concentration either at based on fatigue failure considerations. It is as- L = (CAP)p
hard inclusions or at pre-existing cracks." sumed that no ball or roller bearing gives unlim-
Contact fatigue is also surface generated. In ited service. Owing to the special stress state where L is the fatigue life in revolutions x 106; C is
fact, surface-originatingspalls are more prevalent experienced by rolling contact bearings, bending the standard load (C is defined as the load that gives
than subsurface-generated cracks. Proving sub- or push-pull tensile fatigue results cannot be ap- an Lt0 life of one million revolutions); P is the
surface fracture origin is difficult because a met- plied to their life calculations. There is significant selected load; and p is 3 for ball bearings and 10/3
allographic section only shows a profile of the scatter in life tests for rolling contact bearings. for roller bearings.
crack which, in three dimensions, may have a The Weibull distribution is used in statistical The predicted life from the above relationship
surface origin. The higher the tangential force or analysis of bearing-life tests. A typical bearing- is, of course, based on bearing tests, analyzed
traction, the more likely will be surface-generated life Weibull plot is shown in Fig. 10 (Ref 12). statistically. It does not take into account other
 Contact Fatigue / 335

Operating region for most

100 portional to maximum Hertz stress to the 9th
I I I
Region of industriala p p l i c a t i o n s ~ P - - - - ' ' " "1"-" I
power. The exponent can be as high as 12 for
roller bearings. Lubricant composition, micro-
possible /
surface distress / structure, and geometry of rolling contacts can
80 for bearings / influence these exponents of life. Zaretsky,
.~th severe /
Region of Region of
Poplawski, and Peters (Ref 15) give a good sum-
lubricated-related increased life mary of life exponents from various tests for a
surface distress number of bearing configurations. This also in-
E 60
cludes case-carburized consumable electrode
c
vacuum arc remelted AISI 9310 spur gears. The
o L10 lives varied inversely with the stress to the 8.4
~ 40
power for the spur gears.
Reducing the scatter in bearing lives so that the
distribution would be compressed toward the
20 longer lives would intrinsically improve bearing
lives. Improving rolling-element precision, sur-
face finish, and homogeneity of microstructure
P should reduce scatter somewhat. Lubrication also
0 I I I I I is effective. Ensuring that a rolling contact com-
0.4 0.6 1.0 2.0 4.0 6.0 10
ponent is operating within satisfactory EHD con-
A = function of film thickness and surface roughness
ditions relative to surface finish is essential.
Fig. 11 Ball-bearing performance map. Source: Ref 2 Cleanliness of the operating environment and
reasonable protection from corrosion are also im-
portant.
Because of the sensitivity of contact fatigue life
to contact stress, reduction of contact stress can
5 process, steelmaking methods have been modi- significantly improve bearing life. Of course, ac-
4.5:~~ Catalogl ife[3
fied to eliminate or substantially reduce the pro- curate estimation of the actual operating contact
n 4' duction of hard inclusions. Consumable electrode stress is important. Contact stress can be reduced
O ~ . B/Experimental vacuum melting has produced bearing-grade by spreading out the area contact with a soft solid

i
steels that have dramatically improved bearing thin film applied to the surfaces (bearing races,
reliability. In many cases, bearing failure is now for instance). Conversely, hard coatings have
o. 2.5 [3 related to wear rather than to contact fatigue. been used to improve fatigue life of bearing steels
Good surface finish is necessary for long bearing (Ref 16).
"1- 2 life. As was noted, contact fatigue is initiated by High-speed ball bearings have an increased
1.5 surface defects like dents and deep scratches. ball contact stress owing to centrifugal forces.
1 I I I I Surface defects not only cause asperity contact in Such increased stress levels are sufficient to c a u s e
106 107 108 109 1010 1011 thin-film lubrication, but dents have been shown significant reduction in fatigue life even in very
Life, stress cycles to disturb the EHD film and cause local film clean precision ball bearings. Reduction in the
breakdown. ball mass can reduce this effect and increase life
Fig. 12 Contact stress-life plots for lives based on the in-
The possibility of a fatigue limit for rolling
verse power load-life law and bearing tests with to reasonable levels. Significant advances have
ideal operating conditions. Source: Ref 14 contact (deviation from the inverse load power
been made by the use of silicon-nitride balls for
law) has been investigated. Tallian (Ref 14) has
high-speed bearings. Because of the lower den-
analyzed test data from bearing tests run at high
sity of silicon nitride, centrifugal forces in the
A values under conditions free of contaminants
factors that impact on bearing life. Lubrication is bearing are reduced. Hybrid ball bearings with
and debris and found deviations from the theo-
a powerful factor in beating life. Since the discov- silicon-nitride balls have surpassed bearing grade
retical life suggesting a fatigue limit. This is
ery of thin-film lubrication, elastohydrodynamic steel in rolling contact performance (Ref 17-19).
shown in the plot in Fig. 12. Further information
(EHD) lubrication of rolling contact bearings, the on bearing life is described in the article "Fatigue These bearings are finding use in gas turbines and
effect of film thickness on bearing life, has re- and Life Prediction of Beatings" in this Volume. high-speed machine tools.
ceived considerable attention. Tests have shown Race fracture in high-speed ball bearings can
that the lubricant film thickness is influenced by be avoided by using a carburizing grade steel
bearing speed and lubricant viscosity and less by Minimizing Contact Fatigue with increased fracture toughness (M50 NiL)
load. A beating performance map was developed (Ref 20) instead of through-hardening steels like
by Harris (Ref 2). The performance map is shown The study of rolling contact behavior has indi- AISI 52100. Carburizing to a depth below the
in Fig. 11. It has been in general use for a number cated new approaches that might further improve estimated maximum shear depth will provide the
of years. The lubricant film coefficient, A, deter- the contact fatigue resistance of these systems. required resistance to contact fatigue. Cleanliness
mined by dividing EHD film thickness by a sur- One important problem in the application of roll- of the steel will still be an important factor in
face roughness factor, relates to the present film ing contact systems is the wide scatter in failure bearing life.
or percentage time the surfaces are totally sepa- lives. Zaretsky (Ref 15) indicates that in a group The residual stress state in the near surface of
rated by a lubricant film. If A is less than 1, the of 30 ball bearings the ratio of the longest to the rolling contact elements resulting from heat treat-
beating is likely to not attain the Lip life predicted shortest life may be as much as 20 times. Bear- ment and machining have an influence on contact
by the Weibull distribution. If A exceeds 4, then ing-catalog ratings are based on L10 lives or the fatigue life. By imposing compressive residual
one might expect longer life than predicted. time in which 10% of the bearings have failed. stresses, gear life can be improved. This can be
Steel microstructure also has a significant ef- The same scatter can be expected in other rolling accomplished by shot peening and burnishing.
fect on beating life. Of greatest importance is the contact components such as gears and cams. Nitriding gear steel will produce the desired com-
cleanliness of the steel. Because hard inclusions Bearing-fatigue life is sensitive to bearing load. pressive residual stresses to inhibit crack propa-
have been found to enhance the fatigue crack Generally, it is assumed that life is inversely pro- gation.
336 / Fatigue Strength Prediction and Analysis

As power systems become lighter and more 6. K.H. Kloos and F. Schmidt, Surface Fatigue Model with Fatigue Limit, Wear, Vol 107,
compact, bearings, gears, and other rolling ele- and Wear, MetallurgicalAspects of Wear, K.H. 1986, p 13-36
ments will have to operate at higher speeds. Al- Zum Gahr, Ed., Deutsche Gesellschaft fur Met- 15. E:V. Zaretsky, J.V. Popiawski, and S.M. Peters,
though even at this time not all is understood allkunde, 1981, p 163-182 "Comparison of Life Theories for Rolling-Ele-
about the mechanisms of contact fatigue, ad- 7. K.H. Kloos and F. Schmidt, Surface Fatigue ment Bearings," Preprint 95-AM-3F-3, Society
vances in improved reliability and component life and Wear, MetallurgicalAspects of Wear, K.H. of Tribologists and Lubrication Engineers, May
are being made. Research and testing continue to Zum Gahr, Ed., Deutsche Gesellschaft fur Met- 1995
try to narrow the life scatter and increase the allkunde, 1981, p 163-182 16. A. Erdemir, Rolling Contact Fatigue and Wear
predicted life of rolling contact parts. 8. J.A. Martin, S.F. Borgese, and A.D. Eberhardt, Resistance of Hard Coatings on Bearing Steel
Trans. ASME, Vo159, 1966, p 555 Substrates, Su~ Coat. Technol., Vo154-55 (No.
9. H. Swahn, P.C. Becker, and O. Vingsbo, Met. 1-3), 1992, p 482-489
Sci., Jan 1976, p 35 17. R.J. Parker and E.V. Zaretsky, Fatigue Life of
REFERENCES 10. W.D. Syniuta and C.J. Corrow, Wear, Vol 15, High-Speed Ball Bearings with Silicon Ni~de
1. V. Bhargava, G.T. Hahn, and C.A. Rubin, Roll- 1970, p 187
Balls, J. Lubr. Technol. (Trans.ASME), 1975, p
11. P.C. Becker, Microstructural Changes Around
ing Contact Deformation and Microstructural 350-357
Non-metallic Inclusions Caused by Rolling-
Changes in High Strength Bearing Steel, Wear, 18. F.J. Ebert, Performance of SiliconNitride Com-
Contact Fatigue of Ball-Bearing Steels, Met.
Vol 133, 1989, p 69 Technol., June 1981, p 234-243 ponents in Aerospace Bearing Applications,
2. T. Harris, "The Endurance of Modem Rolling 12. R.J. Boness, W.R. Crecelius, W.A. Ironside, Proc. Gas Turbine and Aeroengine Congress,
Bearings," AGMA paper 269.01, Oct 1964, C.A. Moyer, E.E. Pfaffenberger, and J.V. 11-14 June 1990, American Society of Me-
Rolling Bearing Analysis, John Wiley, 1966 Poplawski, Current Practice, Life Factors for chanical Engineers
3. S.H. Roby, Investigation of Sequence IIIE Rolling Bearings, E.V. Zaretsky, Ed., Society of 19. M. Hadfield, S. Tobe, and T.A. Stolarski, Sub-
Valve Train Wear Mechanisms, Lubr. Eng., Vol Tribologists and Lubrication Engineers, 1992, p surface Crack Investigation of Delaminated Ce-
47 (No. 5), p 413-430 5-7 ramic Elements, Tribol. Int., Vol 27 (No. 4),
4. T.E. Tallian, Failure Atlas for Hertz Contacts, 13. G. Lundberg and A. Palmgren, Dynamic Ca- 1994, p 359-367
ASME, 1992 pacity of Rolling Bearings, Acta Polytechnia, 20. C.A. Moyer and E.V. Zaretsky, Failure Modes
5. K.R. Kilburn, An Introduction to Rail Wear and Mechanical Engineering Series 1, R.S.A.E.E., Related to Bearing Life, Ltfe Factorsfor Rolling
Rail Lubrication Problems, Wear, Vol 7, 1964, No. 3, 7, 1947 Bearings, E.V. Zaretsky, Ed., Society of Tri-
p 255-269 14. T.E. Tallian, Unified Rolling Contact Life bologists and Lubrication Engineers, 1992, p 67