Principles of Tribology

Principles of Tribology
Edited by

J. Halling
Professor of Engineering Tribology
University of Salford

M
© The Contributors 1975, 1978
All rights reserved. No part of this publication
may be reproduced or transmitted, in any form or
by any means, without permission.
First edition /975
Paperback edition /978
Published by
THE MACMILLAN PRESS LTD
London and Basingstoke
Associated companies in Delhi Dublin
Hong Kong Johannesburg Lagos Melhoum~
New York Singapore and Tokyo
ISBN 978-0-333-24686-3 ISBN 978-1-349-04138-1 (eBook)
DOI 10.1007/978-1-349-04138-1

This book is sold subject to the standard conditions of the

Net Book Agreement.
The paperback edition of this book is sold subject to the condition that it
shall not, by way of trade or otherwise, be lent, resold, hired out, or
otherwise circulated without the publisher's prior consent in any form of
binding or cover other than that in which it is published and without a similar
condition including this condition being imposed on the subsequent purchaser.
 This book is dedicated
to the Lima Arms whose
efficacious personal
lubrication service
has often eased the
path to its completion
Contents
Foreword XI
Preface XIII

Introduction 1
J. Halling

1.1 Tribology 1
1.2 Historical 4
1.3 Tribology in industry 7
1.4 Economic considerations 10
1.5 Tribological solutions II

2 Surface Properties and Measurement 16

J. Halling and K. A. Nuri
2.1 The nature of metal surfaces 16
2.2 Surface texture assessment 18
2.3 Surface parameters 22
2.4 The statistical properties of surfaces 25
2.5 Measurement of surface parameters 38

3 Contact of Surfaces 40
J. Halling and K. A. Nuri

3.1 Introduction 40
3.2 Stress distribution due to loading 41
3.3 Displacements due to loading 47

VII
 3.4 Hertzian contacts 48
3.5 The contact of rough surfaces 61
3.6 Criterion of deformation mode 65
3. 7 Thermal effects 67

4 Friction Theories 72
D. G. Teer and R. D. Arnell
4.1 Introduction 72
4.2 Friction measurement 74
4.3 Possible causes of friction 77
4.4 The adhesion theory of friction 79
4.5 Modified adhesion theory 81
4.6 Plastic interaction of surface asperities 87
4. 7 Ploughing effect 89
4.8 Elastic hysteresis losses 91
4.9 Discussion of the various friction theories 91

5 Wear 94
D. G. Teer and R. D. Arnell
5.1 Introduction 94
5.2 Types of wear 95
5.3 Various factors affecting wear 113
5.4 Experimental aspects 120
5.5 Wear prevention 122
5.6 Application of wear relationships to design 122
5. 7 An example of wear in practice- wear of an i.e. engine 125
5.8 Conclusions 126

6 Tribological Properties of Solid Materials 128

R. D. Arnell and D. G. Teer
6.1 Introduction 128
6.2 Tribological properties of metals 129
6.3 Self-lubricating materials 133
6.4 Types of solid lubricant 134
6.5 Tribological properties of plastics 140

7 Friction Instability 147

L. Eaton
7.1 Introduction 147
7.2 Characteristics of friction vibrations 151
7.3 Review of analytical methods 154
7.4 Frictional force models 158

viii
 7.5 Analysis of stick-slip oscillations 161
7.6 Further analysis 170
7.7 Eliminatidn of stick-slip 171

8 Mechanics of Rolling Motion 174

J. Halling

8.1 Introduction 174

8.2 Free rolling 175
8.3 Microslip in rolling 184
8.4 Tyre-road contacts 198

9 Lubricant Properties and Testing 202

R. B. Howarth
9.1 Introduction 202
9.2 Viscosity 202
9.3 Measurement of viscosity 213
9.4 Lubricating oils 224
9.5 Greases 228

10 Hydrodynamic Lubrication 233

T. L. Whomes
10.1 Introduction 233
10.2 Theory 235
10.3 Application of Reynolds equation to sliding bearings 238
10.4 Contacts in the form of non-conforming discs 243
10.5 The journal bearing 247
10.6 Variable viscosity----'the reduced pressure concept 252
10.7 Shear stresses and traction in hydrodynamic films 254
10.8 Finite length bearings 260
10.9 Thermal effects 268
10.10 Gas-lubricated bearings 271
10.11 Hydrodynamic instability 282

11 Elastohydrodynamic Lubrication 288

T. L. Whomes and J. Halling
11.1 Highly loaded contacts 288
11.2 Elastohydrodynamic theory 291
11.3 Comparison of theory and experiment 296
11.4 Traction 300
11.5 Three-dimensional solutions 304
11.6 Fatigue failure 306

IX
12 Hydrostatic Lubrication 308
P. B. Davies and R. B. Howarth
12.1 Externally pressurised bearings 308
12.2 General description of hydrostatic bearings 309
12.3 Viscous flow through rectangular gaps and circular
tubes
310
12.4 Long rectangular thrust bearings in constant flow
system
312
12.5 The need for compensation in multibearing
arrangements 315
12.6 Characteristics of compensated bearings 317
12.7 Comparison of characteristics 325
12.8 Flow, load and power factors for other shapes of
bearings 333
12.9 Sliding effects in thrust bearings 340
12.10 Hydrostatic journal bearings 349
12.11 Other types of hydrostatic bearings 358

13 Selection of tribological solutions 360

J. Halling
13.1 Introduction 360
13.2 Load and speed 361
13.3 The selection of journal bearings 367
13.4 Matching of tribological solutions 367
13.5 Conclusions 368

Appendix A 369

Appendix B 375

Author Index 394

Subject Index 397

X
Foreword
by H. Peter lost
Chairman: Committee on Tribo/ogy, Department of Industry, 1966-74
President: International Tribo/ogy Council

Without tribology, in other words, without 'interacting surfaces in relative

motion', that is, surfaces rolling on each other, surfaces sliding over each
other and surfaces rubbing on each other, life would be impossible.
This truism applies equally to heavy machinery and to a precision mecha-
nism; to a brake and to a rocket; to a mechanical and to a human joint.
Friction and wear, the principal constituents of tribology, have been with
us since time immemorial, so have been our efforts to control the former and
minimise the latter. However, it is only less than nine years ago, that the
modern interdisciplinary concept of tribology was recognised in the United
Kingdom, since when its progress has spread like wildfire throughout the
industrial world.
The main reason behind this development, which is also the principal
reason for the importance of this book, is the recognition of the close inter-
relationship between tribological design principles and practices on the one
hand, and their economic effect on the other. The days of single disciplinary
designs and of design by trial and error are gone forever. Modern products
must, during their design stages, have incorporated all the factors that lead
to a satisfactory control of friction and prevention of wear.
For this to be accomplished, Professor Halling and his co-authors have

xi
 provided, in a single volume, nearly all the basic theory needed for a thorough
grounding in the subject. It will therefore be a book most useful for every
engineering student; in addition, I believe that Principles of Tribology will
be found very useful by practising engineers in industrial design and research
departments. For here, the concentration of valuable information in one
volume, will eliminate the consultation of several sources of books and
scientific papers, a saving of valuable time that will be appreciated by those
employed in industry and others working against the clock.
Similarly, this book should be of considerable value to workers in many
other fields, whose occupation brings them face to face-often unpleasantly
so-with the realities of tribological problems and who may find, in this
volume, fundamental answers to at least some of their problems.
The original Report, that bears my name, estimated that by the better
application of tribological principles and practices, industry in the United
Kingdom could save around £515 million per annum (at 1965 values).
During the years since its publication, it has become apparent that this
estimate of savings has been too conservative. Indeed, in a recent Report,
commissioned by the Congress of the United States of America, it was stated
that there was considerable scope for savings of losses through tribological
causes (friction and wear), which were estimated to cost the U.S. economy
around $100 billion per annum, of which $20 billion were in materials.
The savings through the correct application of tribological principles, as
outlined in this book, can be considerable. A recent report, commissioned
by the S.S.R.C., concluded that 'for individual enterprises-even efficient
ones-the rate of return in improvements in the tribological characteristics
of their capital equipment could be very high indeed-far higher than is
customary on ordinary industrial investments'.
Modern machines, mechanisms and equipment must be reliable. I firmly
believe that the application of the ground rules, contained in Principles of
Tribology, will materially contribute towards greater reliability of industrial
products and therefore to the economy of this country.
I congratulate Professor Halling and his co-authors on their work, which
I can warmly recommend not only to those desiring to become engineers,
but also to those professional engineers and others whose work is connected
with the control of friction and the prevention of wear, in other words, in
work where the minimisation of breakdowns, replacements and outages
through tribological causes are of importance.

London H. PETER JOST

xii
Preface
Tribology is a new word, not yet in common usage, but it deals with problems
which man has encountered throughout the whole of his history. The word
was introduced to focus attention on the problem of carrying load across
solid interfaces in relative motion. The ingredients of the subject are there-
fore well-established batches of knowledge which occur in a wide range of
texts on such topics as lubrication, friction, wear, contact mechanics, surface
physics, and chemistry. The subject is truly interdisciplinary since the basic
knowledge from physics, chemistry, mathematics, materials science and
engineering is used to study problems in all branches of engineering, in
medicine, and in almost all aspects of our daily life from the cleaning of our
teeth to the slicing of our golf drive.
The awareness of the social and economic importance of this subject has
resulted in its introduction into several courses at colleges and universities.
This book is an attempt to bring together in a single volume those topics
which are currently scattered throughout the scientific literature. In particular
this volume concentrates on the basic principles of the subject, using practical
examples only to demonstrate the physical manifestations of these principles.
This book should, therefore, prove a useful supplement to other volumes
dealing with such specific practical applications as the design of bearings.
The book is basically aimed at the final-year level of undergraduate courses
in engineering, but the authors hope that it will prove of interest to a wider
readership. Research workers, new to the subject, and designers and develop-
ment engineers seeking background knowledge to the existing literature
should all find the subject matter relevant.
The authors have deliberately excluded illustrations of standard equipment

XIII
and techniques since these should already be familiar to final-year under-
graduates and practising engineers. The notation may also appear to lack
complete consistency but it has been considered desirable to use that notation
which is already well established in the literature concerned with the various
topics considered. This should facilitate the use of standard references, many
of which are included with each chapter. The symbols used are clearly
identified in the text of each chapter.
It may be that the relatively large number of authors has resulted in some
variation in literary style, but since each chapter is more or less self-contained
this should offer no problem to the reader. Since all the authors belong to the
same department we believe that our collaborative efforts have resulted in a
coherent philosophy for the book. In each topic the authors have concen-
trated on the important physical principles and have not thought it desirable
to include rigorous development of the mathematical treatment, although
the most significant mathematical formulations are included. Since the
authors are mainly concerned with the engineering applications oftribology,
it will be noted that this book does not include much information on the
chemical aspects of tribology. This does not imply any denigration of the
importance of such material but rather that such topics tend to be of a rather
specialist interest.
The final chapter has been included as an indication of the relevance of
the remainder of the book. Indeed this chapter could be taken out of sequence
if the reader seeks to appreciate the value of the topics discussed in the
various chapters.
A number of problems together with outline solutions have been included
in an appendix to this book, since the final proof of assimilation of concepts
must be our ability to handle them in given situations.
Finally the authors would like to acknowledge their deep gratitude for the
enthusiastic and competent way in which Mrs L. M. Chadderton has tran-
scribed their often untidy drafts into a coherent manuscript.

Salford J. HALLING

XIV
1
Introduction
1.1 TRIBOLOGY

What does the word mean? It is derived from the Greek word TRIBOS meaning
rubbing, so that a literal translation would be 'the science of rubbing'. The
word is so new as to appear in only the latest editions of dictionaries where
it is there defined as 'the science and technology of interacting surfaces in
relative motion and of related subjects and practices'. This latter definition,
although embracing the literal translation, is of even wider significance
and was created to bring together the interest in friction and wear of chemists,
engineers, metallurgists, physicists and the like. This wide-ranging concern
with tribology immediately illustrates the interdisciplinary nature of the
subject. In a sense it is the name alone which is new because man's interest
in the constituent parts of tribology is older than recorded history. Clearly,
the invention of the wheel illustrates man's concern with reducing friction in
translationary motion, and this invention certainly predates recorded history.
That man should have been so concerned with the tribological problems of
friction and wear is not surprising because our involvement with such
phenomena affects almost every aspect of our lives.
These problems are not confined to the machines which we use, they also
have profound influences on many other aspects of life. The action of animals'
joints is clearly a tribological situation and cures for such diseases as arthritis
already owe much to the tribologists' expertise. We also rely on the control
of friction in our leisure pursuits, whether they be rock climbing or any form
of ball game-such as the spinning of a cricket or tennis ball, the slicing of
our drive in golf or our proficiency on skates or skis. Holding, cutting and

1
brushing are other manifestations of the impact of tribology on our daily
life, while the cleaning of our teeth is clearly a controlled wear-process where
we wish to avoid wear of the enamel while wearing away unwanted films, etc.
Even our ability to walk is dependent on the existence of appropriate
friction, so that tribological effects have clearly had a major effect on the
whole evolutionary process.
We may examine the effect of friction on the evolutionary process by
considering the way in which developments in translation over the earth's
surface have evolved against the timescale of history 1. In figure 1.1 the resist-
ance to translation is represented by a resistance/weight ratio which might

10- 1

:E
.0' .,
;<
Gs M, Sild1ng
Lubricant
M2
'ec 10- 2 M3 Early wheel
M4 Spoked wheel
~
gj Ms Railway
0:: M6 Modern railways
" G, First reptile
'""' G2 Crawlers
G3 Apes
10- 3 G4 Man
Gs Atheletes
M6
c
-~
.0
u -~ E
0
10- 4 ~ u

~ "' "' Q: "'

1 I·
I
I
10
I
10 2
I
103
Cenozoic era
I
104
I
105
I
10 6
I
107
·I I I, 1..
108 109
Years

Figure 1.1 Improvements in land locomotion throughout

history

be considered as an equivalent coefficient offriction A.. This clearly represents

the ease of translation and its reduction with timescale is seen as evolution
has developed from primeval sliding to the movements of a modern athlete,
that is along line G. In this figure it should be noted that the scales are
logarithmic rather than linear so that the shape of the curve is somewhat
deceptive. Modern man starting some 10 000 years ago has used his inventive-

2
ness to achieve a much better performance than is obtained from the physio-
logical developments in animals. Thus the use of lubricants and improve-
ments in the design of wheels have clearly proved advantageous.
It is also interesting to plot the results in figure 1.1 against the speed
achieved by each method of translation, shown in figure 1.2. Again we note

003 03 3 30
Speed m/s

Figure 1.2 The results from figure 1.1 indicating speeds

achieved

the two curves, one representing the evolutionary process and the other due
to man's inventiveness. What is most interesting in this logarithmic plot
is that the evolutionary line is approximately straight and has a negative
slope of 45°. This means that
log A = - 1 x log S + log C (1.1)

where S is the speed of translation and C is some constant. This equation

may thus be expressed
AS= C ( 1.2)
But AS is the resisting force multiplied by speed and divided by the weight,
that is, power divided by the weight carried, thus this straight line suggests
that the evolutionary line is governed by a substantially constant power/
weight ratio. Clearly this is reasonable, since the power/weight value is

3
defined by the same physiological processes in all animals, including man.
The curve due to man's inventiveness does not suffer from such a restriction,
since our inventions have benefited from a more successful application of
scientific principles and the use of other than purely physiological materials.

1.2 HISTORICAL

Man's invention of the wheel as one of the earliest tribological devices has
already been mentioned, but friction affected the development of civilised
man in many other ways 2 • It is known that drills made during the palaeo-
lithic period for drilling holes or producing fire were fitted with bearings
made from antlers or bones, and potters' wheels or stones for grinding ·
cereals, etc., clearly had a requirement for some form of bearing 3 . Records
show the use of wheels from 3500 B.C., and yet it is interesting to note that the
very advanced Inca civilisation of more recent date never did discover the
principle of the wheel. Lubricants were also used from about this period,
and a tomb in Egypt provided evidence of this fact. A chariot in this tomb
still contained some of the original animal-fat lubricant in its wheel bearings
and it is also interesting that this lubricant was contaminated with road
dirt in the form of quartz sand and compounds of aluminium, iron and lime.
In their monumental tasks of building the Egyptians also showed a clear
appreciation of tribological principles 4 . Surviving illustrations in the form of
bas-reliefs show the use. of rollers and sledges to transport their heavy
weights. Figure 1.3 illustrates one example of such transportation; here 172
slaves are being used to drag a large statue weighing about 6 x 10 5 newtons
along a wooden track. Closer examination shows one man (standing on the
sledge supporting the statue) pouring a liquid into the path of motion;
perhaps one of the earliest lubrication engineers. If we assume that each
slave pulls with about 800 newtons we can see that this picture would suggest
a coefficient of friction of
172 X 800
J1. = 6 X 105 ~ 0.23

This is just about the value we would expect for a lubricated wooden slide,
thus we can infer that this picture is a true record of what actually occurred.
The somewhat artificial arrangement of the slaves probably arises from the
artist's inability to draw perspective.
In 1928 fragments of what must have been a ball thrust-bearing were
found in Lake Nimi near Rome, and probably date from about A.D. 40.
The bearing is shown in figure 1.4. It was probably used to support a statue
in a sculptor's workshop thus facilitating rotation during the sculpting
process. There is little further evidence of tribological development until
the time of Leonardo da Vinci (1452-1519), who first postulated a scientific
approach to friction. He recognised that the friction force is proportional to

4
5
 I

I
---0---- ~ 1
- fi1
I

Figure 1.4 Detail of fragment and theoretical

reconstruction of Lake Nemi hearing, estimated
diameter nearly 1 metre (reproduced from Ucelli,
Le Navi di Nemi, Libreria Delio Stato, Roma)

load and independent of the nominal area of contact. It was almost two
hundred years however before these two laws were enunciated by Amon ton
in 1699, who, independently of Leonardo da Vinci, postulated these laws and
is usually credited with their discovery. The eighteenth century saw consider-
able tribological development because of the increasing involvement of man
with new machines, and about 1780 the work of Coulomb led to the third
law of friction which suggested that friction was independent of velocity.

6
 These three laws are still used today as reasonably true and are to be found
in the teaching of friction in elementary texts in physics and engineering.
Many other developments occurred during this century, particularly in the
use of improved bearing-materials. As early as 1684 Robert Hooke suggested
the combination of steel shafts and bell-metal bushes as preferable to wood
shod with iron for wheel bearings. Further developments were undoubtedly
associated with the growth of industrialisation in the latter part of the
eighteenth century.
With lubricated bearings, although the essential laws of viscous flow
had earlier been postulated by Newton, any scientific understanding of
their operation did not occur until the end of the nineteenth century. Indeed
our understanding of the principles of hydrodynamic lubrication can be
seen to date from the experimental studies of Beauchamp Tower (1883) 6
and the perspicacious theoretical interpretation by Osborne Reynolds
(1886) 7 • Other work by Stokes and Petrof'~ about this time was also very
pertinent to these developments. Subsequent development in hydrodynamic-
bearing theory and practice was extremely rapid in an attempt to meet the
increasing demand for reliable bearings in the new machinery then being
developed, some of which is still operational to this day.
Since the beginning of this century, and spurred on by the industrial
demand for better tribology, our knowledge in all areas of tribology has
expanded enormously. In this context ball-bearings, which were first intro-
duced for industrial applications about 1700, have now reached an unusual
peak of efficiency. They are available in a wide range of sizes and today offer
a very cheap and flexible solution to many tribological design problems.
Other developments in the quality and service characteristics of lubricants
have also added to our toolkit of tribological solutions, but the pace of
modern industrial society is such that increasing demands for higher speeds
and loads, often in hostile environments such as nuclear reactors and space
vehicles, still necessitate further development of the subject. Today we take
as the norm that a car engine should last about 100 000 miles, whereas less
than twenty-five years ago the life expectancy was only a third of this value.
It is also interesting to note that a modern car contains upward of 2000
tribological contacts, so that it is not surprising that this subject is of increas-
ing importance to engineers engaged in a wide variety of engineering
disciplines.

1.3 TRIBOLOGY IN INDUSTRY

Before discussing the prevalence and types of tribological problems in
industry we must first take a somewhat global view of the phenomena. In
essence we are dealing with the interaction of two solid surfaces within a
given environment which results in two outward manifestations.
(a) There is an energy dissipation which is the resistance to motion and is
indicated by the coefficient of friction. This energy dissipation results

7
 in a heat release at the contact and a small, but sometimes significant,
amount of noise. It should be emphasised that since two solid surfaces
are always involved, parameters such as the coefficient of friction must
relate to the pair of interacting materials. To talk about the coefficient
of friction of steel without reference to the mating solid is scientifically
incorrect and misleading. It might also be remarked that the idea of
frictionless surfaces is scientifically impossible, and the often-stated
implication that low friction is associated with the smoothness of the
surface is also basically incorrect.
(b) During the sliding process all surfaces are to a greater or lesser extent
changed in their basic characteristics. They may become smoother or
rougher, have physical properties such as their hardness altered,
and some material may be lost in the so-called wear processes. Such
surface changes may be either beneficial, as happens when surfaces
'run in' to produce near ideal operation, or disastrous, as happens when
surface failure occurs, necessitating the replacement of components.
From the foregoing it may be thought that both friction and wear are
always disadvantageous, but this is not the case. In many engineering appli-
cations we are able to employ friction to fulfil required functions. Thus brakes,
clutches, driving wheels on trains, cars, etc. all operate because ofthe existence
of friction, while the commonplace nut and bolt only work because of the
friction between the two. In the same way the wear of machinery is some-
times advantageous. The initial wear resulting in better mating of com-
ponents (running-in), is evidently desirable, while the fact that components
wear-out provides a strong motivation to replace obsolescent machinery.
In its most extreme form this leads to the concept of 'planned obsolescence'
where designers endeavour to use the wear phenomena to provide machinery
with a specified life-span.
A fairly general misconception exists that friction and wear, which must
be related in some way ~ince they both arise from the interaction of surfaces,
are simply related such that high friction means high wear. That this is not
the case is clearly shown in table 1.1 where it is seen that the lowest friction
is not associated with the lowest wear. These results are also interesting in
that they show the very large differences in wear rates for materials whose
friction coefficients vary in a fairly modest manner. The complexity of the
relationship between friction and wear is also demonstrated by the peculiar
results which sometimes occur when, with the same materials, friction may
decrease after a'given running time and this reduction in friction is associated
with an increase in the wear rate.
In almost all industrial situations the wear effects are more important than
the frictional losses because they tend to have the greater economic conse-
quences. Higher friction can often be tolerated, with its slightly higher running
costs, provided there are consequent savings by increased wear life of the
machine. Some of the industrial situations where friction and wear are
important are indicated by the following categories.

8
 TABLE 1.1 FRICTION AND WEAR FROM PIN ON RING
TESTS. RINGS ARE HARDENED TOOL STEEL EXCEPT IN
TESTS 1 AND 7

Coefficient of Wear rate

Materials friction cm 3 jcm x 10- 12

I Mild steel on mild steel 0.62 157,000

2 60/40 leaded brass 0.24 24,000
3 PTFE 0.18 2,000
4 Stellite 0.60 320
5 Ferritic stainless steel 0.53 270
6 Polyethylene 0.65 30
7 Tungsten carbide on itself 0.35 2

Load 400 g Speed 180 cmjs

1.3.1 Energy Losses

While friction may not always be a prime consideration, the tribologist has
continuously to be aware of the absolute magnitude of the energy losses
involved and any likely consequences of the ensuing heat release. Thus in a
500 MW generating-set a frictional dissipation of only one tenth of one per
cent represents 0·5 MW or the heat equivalent of 500 single-bar electric
fires. This heat is confined to a fairly localised region of the system and
therefore requires special consideration from the designer. Indeed in many
machines the tribological heat release can result in undesirable thermal
distortions.

1.3.2 Wear
In industrialised societies about one-half of the gross national product is
used to replace the results of wear and similar effects, so that these phenomena
are central to the development of such societies. Savings by improvements
in the life of machinery also have the economic advantage that they are
deflationary, a particularly attractive feature in these days of inflation. It
is of course true that wear may also be desirable as a spur to the replacement
of obsolescent machinery, that is, the concept of 'planned obsolescence'.
Such concepts do not however exclude the need to understand wear processes,
since planned wear or wear control of any kind implies an appreciation of the
basic rules of the wear processes.

1.3.3 Control
There is a requirement in sophisticated machjnery for the precise control of
parts in relative motion. This clearly involves one in proper tribological
design and is a factor which is increasingly important for systems using

9
automatic control methods for the positioning of machine elements. Where
parts are moving at very slow speeds such problems are accentuated due to
instabilities in the frictional behaviour.

1.3.4 Environment
It is not possible to solve all engineering tribological problems by a liberal
supply of oil. In many applications the use of such a lubricant is either
excluded or technically undesirable. Typical examples of such situations arise
either in whole or in part with space vehicles, nuclear reactors, chemical
plants, textile plants, food processing machines and many other forms of
machinery. Indeed the tribologists' problems arise increasingly from the
designers' requirement for higher load-capacity, higher speeds and the
operation in difficult and sometimes 'hostile' environments.

1.3.5 Friction Devices

As already mentioned, in many situations friction is a prior requirement for

the functioning of the device. In such cases the heat release and the rate of
wear become the primary design considerations, as in the design of auto-
mobile brakes and clutches.

1.4 ECONOMIC CONSIDERATIONS

The economic significance of tribology is so obvious as to require little

detailed discussion, but it is still true that the savings in individual cases are
generally small, and we all find difficulty in applying the maxim 'save the
pennies and the pounds will look after themselves'. It is because of the
enormous prevalence of tribological contacts in machinery that the small
savings on each contact can add up to significant sums for the nation as a
whole. Thus it is not surprising to find that about one-half of the world's
energy production is used to overcome friction in some form. In a very real
sense we therefore see that savings by better tribological design could have
considerable significance in the conservation debate which concerns the
whole future of mankind. For these reasons governments of industrial socie-
ties have placed increasing emphasis on the economic aspects oftribology.
In Britain the recent acceleration in tribological activities owes much to
the 1966 D.E.S. report, Lubrication (Tribology), Education and Research. 9
This report (the Jost Report) suggested that this country could save no less
than £515 million per annum by better tribological practices. This enormous
sum arises from the constituent savings shown in figure 1.5, and it must
always be remembered that such costs include the loss of production etc
which is consequent on tribological failures in modern industry. It is also

10
 Manpower sav•ngs
Lubricant - " -
Investment - " -

Less frrct1ona1
diSSIPOI10n

Longer life
of machines

Fewer breakdowns

Less mamtenonce
and replacement

Figure 1.5 Sarings indicated hy

the Jost Report 9

apposite to note that these savings do not require new research but only the
application of current knowledge. As research in this area expands, even
greater savings could be anticipated. Experience since 1966 suggests that
these figures were not unduly optimistic and the introduction of the word
tribology and the attendant publicity have already had considerable effect
on entrenched attitudes and in the best interest of the nation's economy.

1.5 TRIBOLOGICAL SOLUTIONS

Perhaps the most significant effect of the introduction of the word tribology
has been to introduce a problem-orientated view of any system. We thus
tend to think: 'What is the best solution to the problem of carrying load
across the interface with acceptable friction and wear?' rather than heretofore

11
where the lubrication engineer naturally tended to have a predisposition to
adopt a lubrication solution. We now therefore categorise the available
solutions to tribological problems in the manner shown in figure 1.6 and
discussed in the following text.

(a) Dry contoCf ( bl Chem1col f1lms

/////

~
\\\\\

(c) Lamellar sol1ds (d) Pressurised lubncont

f1lms

(el Elostometers ( fl Flexible stnps

(g l Roll1ng elements (h) Mognet1c f1elds

Figure 1.6 Methods of solution of tribological problems

Figure 1.6a We may choose the contacting materials because they have
intrinsically low friction and/or wear characteristics, although this may
mean accepting lower load-carrying capacities as for instance when plastic
materials are employed. In many cases it is possible to use materials as
surface layers supported on substrates which fulfil the basic structural
requirements of the particular component. This method is employed in the
well-known bearing s~el used in automotive engines.
Figure 1.6b We may apply chemical films which protect the surfaces

12
and in part reduce the intimate contact of the base materials. In such systems
the thermal stability of these films is important due to the high local tempera-
tures which are created at the points of intimate contact during sliding..
Figure 1.6c Solid surface coatings may be used since they have low
resistance to transverse shear, for example soft metal layers or lamellar
solids such as graphite and molybdenum disulphide. These latter materials
have a layer structure rather like a pack of playing cards with strength to
carry normal load and weakness along planes at right-angles to facilitate
sliding.
Figure 1.6d The surfaces may be separated with a continuous film of
fluid, which may be either a liquid, a vapour or a gas. In such systems the
fluid film must have a built-in pressure to withstand the effects of the applied
normal load. Such pressures may be provided by two distinct mechanisms.
The most obvious is to supply the fluid at a pressure generated by an external
pumping system as used in the externally pressurised bearing, often referred
to as the hydrostatic or aerostatic bearing. Alternatively the pressure may be
generated by the motion of the surfaces themselves as they tend to drag the
fluid into a converging gap. This action of the hydrodynamic bearing is more
dependent on the viscosity properties of the fluid than in the hydrostatic case.
In both these fluid-film applications a wide range of fluids such as water,
oil, air and even liquid metals in nuclear reactors, have been successfully
employed.
Figure 1.6e Where the degree of transverse displacement is of fairly
small amplitude the surfaces may be separated by elastometers bonded to the
two surfaces. This clearly offers an excellent tribological solution and alter-
native designs might incorporate flexible elastic strips as shown in figure
1.6f.
Figure 1.6g A widely employed tribological solution is to interpose
rolling elements such as balls, cylinders and the like between the two surfaces.
The wide range of rolling contact bearings available is evidence of the value
of this particular solution.
Figure 1.6h The carrying of load without mechanical contact is clearly
possible by using magnetic and similar force fields. Such a bearing is to be
found in the domestic electricity-supply meter.
Having this range of solutions in mind the designer may now consider
such factors as the load to be carried, the speed, the nature of the environment
and any limitations on friction and wear, in arriving at the most appropriate
answer to his design problem. It sometimes happens, however, that by a
complete change of design philosophy particular tribology problems may
be either eliminated or at least modified while still achieving the desired
effect from the resulting machine. This concept is illustrated in figure 1. 7
where a power unit is required to drive an aeroplane. Here the tribological
implications of three possible engines are considered, namely, the internal
combustion reciprocating engine, the jet engine and the ram jet. The shaded
areas illustrate where tribological contacts have to be considered between the

13
 '-'
~
(f)
<l
u
'-'
z w
z
Vi
<l
u
~
w
w
z
i3
z
w

Propel lor

I C ENGINE

1'7"'7"'7"71 Energy
t:::..L...L.L.J diSSipOfiOnOf Propuls1ve
beonngs and Jel
sl1d1ng surfaces
RAM JET

Figure 1. 7 Energy dissipation in devices for propulsion

various engine components and it is seen how these are reduced in number,
although sometimes becoming more complex, as we move from one design
to the next.
In the foregoing it has been assumed that surfaces are simply smooth
planes defining the boundaries between solids and their environment.
Unfortunately all engineering surfaces are rather more complex than this,
having geometries with hills and valleys and equally complex physical and
chemical properties which are seldom uniform throughout the depth of the
material. It is these features of the surfaces which contribute to the complexity
of the subject of tribology. The remainder of this book is therefore devoted to
considering the nature of such surfaces and to the underlying scientific
principles involved in the various topics discussed above.

REFERENCES

I. J. Halling, The roll (role) of the wheel. Wear, 24, (1973), 53.
2. D. Dowson, Tribology-Inaugura1 Lecture. University of Leeds Press, (1969).
3. C. St. C. Davison, Bearings since the Stone Age. Engineering, (Jan. 1957), 2.

14
4. C. St. C. Davison, Transporting sixty-ton statues in early Assyria and Egypt. Tech-
nology and Culture, 2, (1961), No. I.
5. J. H. Harris, The Lubrication of Rolling Bearings. Sheli-Mex and B.P., London,
(1967).
6. B. Tower, Report on Friction Experiments. Proc. lnstn mech. Engrs, (1884), 632.
7. 0. 0. Reynolds, On the theory of lubrication and its application to Mr. Beauchamp
Tower's experiments. Phil. Trans. R. Soc., London, (1886), p. 117.
8. N. P. Petroff, Friction in machines and the effects of the lubricant. Engng J., (1883),
St. Petersburg.
9. Lubrication (Tribology) Education and Research, D.E.S. Report, (1966), H.M.S.O.

15
2
Surface Properties
and Measurement

2.1 THE NATURE OF METAL SURFACES

In the field oftribology it is usually necessary to widen the simple interpreta-

tion of a surface as being a geometric plane separating two media. A surface
must be recognised as a layer that grows organically out of the solid and has
physical properties of considerable functional significance. The surface
layer of metals is known to consist of several zones having physico-chemical
characteristics peculiar to the bulk material itself.
At the base of the surface layer, (see figure 2.1) there is a zone of work-
hardened material on top of which is a region of amorphous or micro-
crystalline structure. This so-called Bielby layer is produced by the melting
and surface flow during machining of molecular layers which are subsequently
hardened by quenching as they are deposited on the cool underlying material.
This basic structure is usually contaminated by the products of chemical
reaction with the atmosphere and is covered by dust particles and molecular
films deposited from the environment. Finally, the surface contains atoms of
gas, which live with the surface and have properties somewhat dissimilar to
those of the gaseous environment. In addition, the whole texture of the surface
layer has a geometric property characterised by a series of irregularities
having different amplitudes and frequency of occurrence. This particular
property, the surface texture, is of fundamental importance in the study of

16
 Bielby
layer

Deformed
layer

Base
molena!

Figure 2.1 Typical surface layers

friction, wear and lubrication and thus a knowledge of, and a definition of,
the topographic features of surfaces is vital.
Some feel for the scale of the various surface properties is given in figure
2.2, and it should be noted that the vertical scale is a log scale in ascending
orders of magnitude. Furthermore such terms as oxide films must be care-
fully defined since they are complex layers which, on steel for example, tend
to have an increasing oxygen/metal ratio as the atmosphere is approached.

u
.
.g
~
;;:::
u
~
;;:::
5
:I
E
0 0
c:,..
"' E
ilQ "0
e 5
loG Q z
10 5
Iii
,.. E
!!
Q
.2 ",..
s::
.2 0 g
E
104 "'0
"' g
0
~ "' w
"'0
10 3
. E
...: Air

·'~
c:

102
8" eu Fe 3 0 4
~
10 FeO
Surface
Metal Fe
10

102
Iii
103 ~
,..
D
104 9i
ID c:
Q
105 g
10 6
i,..
>
r
2

Figure 2.2 Order of magnitude of surface features

17
 The geometric texture of ordinary surfaces is controlled by the character-
istics of the finishing process by which they are produced. Close examination
of these surfaces, even after the most careful finishing, shows that they are
still rough on a microscopic scale. The roughness is formed by fluctuations
in the surface, of short wavelengths (microroughness), characterised by hills
(asperities) and valleys of varying amplitudes and spacings and these are
large compared to molecular dimensions. On many surfaces a longer wave-
length roughness called waviness is also observed and is often referred to as
macroroughness (figure 2.3). In addition the surface also contains undulations

Macroroughness
(waviness)

M icroroughness

Ill

Resultant
surface

Figure 2.3 Components of surface geometry

of very long wavelengths caused by the vibrations of the workpiece or tool

during the preparation of the surface. The distribution of the asperities over
the surface can be either directional or homogeneous in all directions,
depending upon the nature of the processing method. Surfaces which have
been submitted to directional methods of processing, such as turning,
milling and planing, exhibit a definite orientation of asperity distribution.
Those which have been submitted to non-directional methods, such as
electropolishing and lapping, show an isotropic or equiprobable distribution
in all directions along the surface.
From the foregoing description it can be appreciated that the surfaces of
metals can be as complicated as the surface of the earth and, obviously, in
order to make a precise assessment of their topographic features, three-
dimensional maps are needed. Unfortunately, simple methods for producing
such maps have not yet been developed, and for the time being we must rely
on a two-dimensional profile of the surface together with observations using
the various microscopic techniques which are discussed later.

2.2 SURF ACE TEXTURE ASSESSMENT

The quantitative assessment of the topographic features of surfaces is of
vital importance for solving a wide variety of problems in tribology. Trib-
ological phenomena such as friction and wear depend primarily on the nature
of the real area of contact between the surfaces which is, in turn, dependent

18
upon the distributions, sizes and shapes of the asperities. Measurements of
these features will, therefore, prove indispensable in the study of all kinds of
surface contact phenomena. Indeed, they offer a valuable analysis of the
conditions for elastic or plastic contact of metal surfaces and information on
the size of the interstices between them. Furthermore, the application of
surface measurement to such problems has produced correlation, both
qualitative and quantitative, between theoretical arguments and experi-
mental evidence.
Many methods are available for the measurement of the micro- or macro-
geometrical features of surfaces. They include optical methods using electron,
interference or reflection microscopy and mechanical methods such as
oblique sectioning and profilometry. The optical methods show certain
advantages in that they can provide a three-dimensional appreciation of the
surface. However, the obtaining of quantitative assessment of surfaces by
these methods is still relatively tedious.
Because of the difficulties in representing every irregularity within the
whole plane of the surface, it is necessary to accept measurements which are
based on a small sample of the surface, always provided that its size is
large enough to be representative, that is, there must be a high degree of
probability that the surface lying beyond the sample is similar to that which
lies within it. Most of the existing methods of surface measurement give
either a high-resolution picture of a small, and often unrepresentative,
sample of the surface or a representative sample with correspondingly lower
resolution, (table 2.1). This difficulty is overcome with profilometers which
provide a representative length of the surface and a high resolution in a plane
normal to it 1 . The higher vertical magnification results in a distortion of the
recorded shape of the profile and thereby sometimes leads to a physical
misconception as to the general character of the actual surface profile. Most
surfaces have asperities with gentle slopes rather than the jagged character-
istics observed on profilometer traces, (figure 2.4). The major disadvantage of
profilometers however, is that they are restricted to a single line sample which
may not be representative of the whole surface if the texture has character-
istics dependent upon the orientation of the record. Nevertheless, for iso-
tropic surfaces they provide useful measurements of the various parameters
which are necessary for their characterisation.
It is also appropriate to record a difference between a peak on a profilo-
meter trace and an actual summit on the real surface. Since the stylus will
register a peak even when it only traverses a shoulder of an actual summit,
it will be clearly recognised that the number of true summits is considerably
less than the number of peaks recorded. This distinction may easily be shown
by taking a series of line profiles in the same direction but with each traverse
slightly displaced from its neighbour. In this way, the information in each of
the line tracings may be consolidated to produce a true three-dimensional
contour map of the surface and thereby clearly identifies the actual summits,
(figure 2.W.

19
 TABLE 2.1 SURFACE MEASUREMENT METHODS

Resolution (J.!m)

Method Lateral Vertical Comments

Results depend on the quality

Optical of the optical system and
0.25-0.35 0.18-0.35
microscope hence depth of field of
photomicrographs

Light Using optimum microscopic

0.25 0.25
profile conditions

Sectioning angle ~ tan- 1 0.1

Oblique
0.25 0.025 and optimum microscopic
section
conditions

Interference Requires high specular

0.25 0.025
microscope reflectivity of surface

Multiple Requires high specular

beam 5 0.005 reflectivity of surface and no
interference angular deviations greater than 5

Reflection O.o3 Vertical resolution from

electron 0.03-0.04 profiles.
microscope 0.02-0.008 Vertical resolution from shadows

0.005 Vertical resolution using

Electron stereo device.
0.005 Vertical resolution using shadowed
microscope 0.0025
replica angle tan- 1 0.3

Finite size of the stylus

Profilometers 1.3-2.5 0.005-0.25 is ultimate limit on resolution

The best-known profilometry method for obtaining surface profiles is the

stylus method, in which a fine diamond stylus traverses the surface and its
vertical movements are recorded, usually by electrical systems 3 . The
Talysurf (Rank, Taylor, Hobson) is one of the most popular instruments of
this type. In such systems the vertical measurements must be recorded with
respect to some appropriate datum. The two most common methods for the
establishment of such datums are
(a) The use of datum-generating attachments which ensure accurate
horizontal motion of the stylus-support system.
(b) The use of large-radius skids or flat shoes which rest on and traverse the
surface being measured, thereby generating the general level of the
surface texture.

20
 SlyiUSIIP
I
rodn;s
I

~ I I
I
I
,,,.,....·vocr•rnv;
I
Ac1uo1 prof1le I
wuhout dJSTOr"fron
D•slor eo I
~ed by I
prof1lometer 1

Figure 2.4

Adjocenr prof•les

F(qure 2.5 Typical surface map (lighter areas corre-

spond to higher surface regions)

21
 Clearly the use of such devices creates some experimental errors, although
(a) is rather more precise than (b) for accurate work. Another major error
in these instruments arises from the size limitation of the measuring stylus.
Its finite size clearly precludes it from complete penetration of all the valleys,
thus their recorded shape will often appear narrower than their actual shape.
In a similar way the peaks are distorted in shape such that they appear wider
than they actually are. This effect, however, is less harmful with smoother
surfaces where the slopes of the asperities are known to be fairly gentle.
In spite of its defects, the stylus instrument, as far as engineering purposes
are concerned, remains the outstanding instrument for studying the nature
of surface geometry and for evaluating its parameters. The main difficulty
does not, in fact, lie in reproducing surface profiles accurately, it lies in using
these profiles for the evaluation of the important features which will determine
the functional behaviour of real surfaces.

2.3 SURFACE PARAMETERS

The choice of surface parameters is necessarily influenced to a certain extent

by the method chosen to reveal the surface features. The stylus method reveals
only a single plane property of the surface topography and, consequently,
its characterisation has to be based on the nature of the ensuing single line
profile. In production engineering one of two parameters is usually used to
define the texture of surfaces. These parameters are the C.L.A. (centre-line
average) roughness value, and the R.M.S. (root mean square) value. The
C. L.A. value is defined as the arithmetic average value of the vertical devia-
tion of the profile from the centre line, and the R.M.S. value as the square
root of the arithmetic mean of the square of this deviation. In mathematical
form they can be written as
• l n
C.L.A. = - L lzil
ni=l

(2.1)

L (zY ]1/2
l n
R.M.S. = [ -
n i=l

where n is the number of points on the centre-line at which the profile devia-
tion zi is measured, (figure 2.6). The centre-line is taken as a line which
divides the profile in such a way that the sums of the enclosed areas above and
below it are equal.
These parameters are seen to be primarily concerned with the relative
departure of the profile in the vertical direction only; they do not provide
any information about the slopes, shapes and sizes of the asperities or about

22
 X

Figure 2.6

the frequency and regularity of their occurrence. It is possible, therefore, for

surfaces of widely differing profiles to give the same C. L.A. or R.M.S. values,
(figure 2.7). These single numerical parameters are mainly useful for classi-
fying surfaces of the same type which are produced by the same method.

(a)

(b)

(c)

(d)

(e)

(f)

Figure 2. 7 Various surfaces having the

same C.L.A. value

Lapped surfaces, for example, differing only in the grades of the lapping
compounds to which they have been submitted, will have the same pattern of
roughness and a single parameter will be adequate to characterise them.
For a more rigorous definition of surface profiles, however, the C.L.A. and
R.M.S. parameters are inadequate and more information will be required.
It is important to realise that the choice of parameters for specifying the

23
surface texture will be controlled by the functional requirements of the
surface, that is, by measurements of those features of the surface which are
known to be significant in any given practical situation.
Abbott and Firestone 1 , the founders of profilometry, were concerned with
the wear behaviour of surfaces, and they chose their parameters accordingly.
Using the profile of a surface they constructed its bearing-area curve by
measuring the fraction of the sample length which lies inside the profile
at various positions above the lowest point on the profile, (figure 2.8). By

~- - - - - - ~me
L
~nh- - - - - - i -roo per cen~

Figure 2.8 Method of deriving the bearing area

CUrl'e

dividing the surface into three height-zones, containing the highest 25 per
cent, middle 50 per cent and lowest 25 per cent of the bearing area, they identi-
fied three parameters-the peak, medial and valley occurrences respectively.
It is interesting to find that they were well aware, some thirty-nine years ago,
of the inadequacy of single numerical parameters to define surface textures.
Meyers 4 was concerned with friction between sliding surfaces and suggested
that the slope and curvature of the profile should affect the frictional be-
haviour of surfaces and found strong correlation between slopes and friction.
Recent work on the dry contact of surfaces has shown that this pheno-
menon may be largely explained by the shapes of the peaks and their distri-
bution through the upper decile of the texture. Indeed, in such problems the
properties of the valleys are almost insignificant. For such a situation,
therefore, the required surface parameters must incorporate knowledge of
the number of peaks of any given height level in the upper part of the texture,
a shape factor for the peaks and a statement of the frequency of occurrence of
such peaks in the plane of the surface. This information would, be particularly
important in considering a large range of engineering problems such as
tribological situations, electrical contacts, thermal contacts and the stiffness
of joints created by mating surfaces. Conversely, in the context of stress
concentrations in surfaces under load, it would be anticipated that the shape
and occurrence of the valleys would be the most significant parameters. For
surfaces separated by lubricated films, or those which are to be covered by
layers such as paint, it is apparent that we shall be concerned with the whole
of the texture characteristics.

24
 It should be pointed out that the profiles of surfaces can be considered,
in statistical terms, as stationary processes of the random type and therefore
in order to formulate the various parameters discussed earlier, it is necessary
to apply statistical methods in describing the properties of surfaces.

2.4 THE STATISTICAL PROPERTIES OF SURFACES

It was mentioned earlier that surface profiles often reveal both a periodic and
a random component in their geometric variation. Thus the periodicity is
most marked in fine shaped surfaces and in diamond turning, whereas a
considerable degree of randomness is apparent on abraded surfaces such as
those produced by grinding. It has become common practice to break up
the periodic variations into frequency bands such as waviness and roughness.
In a sense it would be desirable not to make such distinctions, but this would
necessitate the evaluation of a sample of the surface which includes all the
frequency variations, and this inevitably leads to a need for the measurement
of the whole of the surface. This latter requirement is unacceptable and the
division into frequency bands is therefore a practical necessity. In what
follows, we shall restrict ourselves to consideration of the higher frequency
spectra, that is, the roughness of the surface.
Before any quantitative statements can be made about surface profiles
a datum line must first be established. For this purpose the centre-line of the
texture offers the most usual choice, although any line parallel to this line
would be satisfactory. Let us first of all consider the properties of surface
profiles in the vertical plane.

2.4.1 The Height Distribution of Surface Textures

Surface textures can be adequately described in terms of the distribution
function of the heights of their profiles. Indeed, Abbott and Firestone's
bearing-area curve is, in statistical terms, the cumulative distribution of the
all-ordinate distribution curve. This can be written as
X

F(z) = J1/J(z) dz
where z refers to the heights of the profile measured from the centre-line and
1/J(z) is the probability density function of the distribution of these heights.
The practical derivation of such distribution curves involves taking
measurements of z 1 , z 2 , etc. at some discrete interval I and summing the
number of ordinates at any given height level, (figure 2.9). In effect this
may be interpreted as converting the continuous analogue signal of the
profile into a discrete digital record taken at intervals I. Self-evidently the

25
 1nterval d1s rn but1on
histogram

Figure 2.9 Method of deriving the all-ordinate distribution

distribution curve is then seen to be the best smooth curve drawn through the
histogram produced by such a sampling procedure.
Many surfaces tend to exhibit a normal, gaussian, distribution of texture
heights. Figure 2.10 shows how a gaussian distribution is a reasonable fit

Figure 2.10 Typical distribution curve for a ground

surface

for the histogram obtained for the all-ordinate distribution of a ground

surface. The curve of the gaussian distribution or its density function is given
by
1/J(z) = I/Jo(z)e-z2/2uz

where rr is the standard deviation of the distribution, which is defined in the

same way as the R.M.S. value given by equation 2.1, and rr 2 is the variance.
I/J0(z) can be calculated from the fact that the area of the curve must be equal

26
to the total number of data summed over the chosen scale. The area of the
gaussian curve is

f
ex,

l/lo(z)e-xz/2az = l/lo(z)u(27t)t/2
-oo
Therefore the curve

(2.2)

has unit area and the gaussian distribution curve is usually written in this
standard form. Since this curve encloses unit area, the foregoing procedure
has simply adjusted the vertical scale of the distribution curve to produce a
probability density function, that is the size of any ordinate on this curve
indicates the probability of occurrence of that event in the total population.
It is worth recording that the origin ofthe above curve is located at the centre-
line or the mean ofthe distribution. If we wish to write the curve with reference
to some other point as origin, then
1
1/J(z) = u(27t)t/2

where m is the distance of the mean from the value chosen as origin.
The values of the ordinates 1/J(z) of the gaussian distribution curve and
those of the corresponding areas are found in most books on statistics. The
form of the gaussian distribution necessitates a spread for the events from
- oo to + oo which cannot happen with practical surfaces; in practice the
distribution curve is truncated to finite limits of about ± 3u. Fortunately
about 99.9 per cent of all events lie within this region and consequently the
truncation would lead to negligible error while providing useful simpli-
fication.
It must be emphasised at this stage that although several common surface
preparations produce near-gaussian distributions, many do not. There are
surfaces whose textures show certain departures from the gaussian distri-
bution and it is necessary to define some statistical parameters which are
used for measuring such departures, as well as to consider some other
simple distributions 5 .

Moments, Skewness and Kurtosis

(a) Moments
The nth moment of the distribution curve 1/J(z) about the mean is defined as

f
CJc

Mn = z"l/l(z) dz (2.3)
-oo

27
so that

R.M.S. =a =[_£ '~()

= [2nd moment of t/t(z)]l' 2
d'r
and

C. L.A. =2 Jzt/t(z) dz
0
= twice the I st moment of half t/t(z)
Obviously, the lst moment of the whole of t/t(z) about the mean is zero, and
is in fact the method by which the centre-line of the distribution can be
located.

(i) Gaussian Distribution For a gaussian distribution curve the nth moment
is given by
X

M =
n
1
0"(2rr)l/2
f z"e-z2/2a dz
-oc

If n is odd this vanishes, as it must for any symmetrical curve. If n is even then

M = n'. (Jn (2.4)

n 2"12 (n/2)!
Thus the 2nd moment becomes simply 0" 2 , the variance.

(ii) Rectangular (Uniform) Distribution The rectangular or uniform distribu-

tion implies that at any level there is an equal number of events, (figure 2.11 ).
We can write
t/t(z) = c = constant
so that
'L

Jt/t(z) dz = 1

For practical purposes let us suppose that this distribution is cut off
at the finite limits ±h, and then work out these limits in terms of the standard
deviation O". From the definition of the second moment we have
h

0" 2 = Jz t/t(z) dz
2

-h
h

= J zzc dz = 2c3h3
-h

28
that is

Now, making the area under the curve equal to unity gives
30"2
2h 2h3 = I

or
h = (3)1/20"

and the distribution curve can be written in the standard form

(iii) The Triangular (Linear) Distribution If the number of events varies

linearly with height then the distribution of the events is said to be triangular
or linear, (figure 2.11).

!
I
-t>
l
0 +t>

Gouss,on Uniform Lmear

Figure 2.11 Forms of distribution

Again let us consider the finite limits ±h of such a distribution. For this
we can write
c
t/J(z) = c - - z O<z<h
h
c
= c + -z -h < z < 0
h
such that
h

Jt/l(z) dz = I
-h

29
Now
h

a 2 = J z 1/J(z) dz
2

-h

that is
6a 2
c = /;3

Making the area under the distribution curve equal to unity gives
h x c =1 or h = (6) 1 ' 2 a
The triangular distribution can therefore be written as
1 z
1/J(z) = (6)112a - 6a2 0 < z < (6) 112 a

1 z
= (6)112a + 6a2

(b) Skewness
The skewness is a measure of the departure of a distribution curve from
symmetry. This is defined as
en
J z 1/J(z) dz
3
S = _-_oo.:__---c;-----
(13

3rd moment of 1/J(z)

(J3
(2.5)

For a large class of moderately skewed distribution curves the skewness

can be calculated from an empirical relationship given by

3 (mean - median)
S=-------
(J
(2.6)

where the median represents that value of the variable whose ordinate divides
the area under the distribution curve into two equal parts.
Clearly, symmetrical including gaussian distribution curves have zero
skewness; unsymmetrical curves can have either negative or positive skew-
ness, as shown in figure 2.12.

30
 -ve Skew

Figure 2.12
(c) Kurtosis
The kurtosis is a measure of the hump on a distribution curve. This is defined
as
ex,

J z 1/J(z) dz
4

k = ----"(L'------o---
a4
(2.7)
4th moment of 1/J(z)
a4
For a gaussian distribution, using equation 2.4, we find that
I 4x3x2 4
k = a4 X 2 X 2 X 2a = 3

that is a gaussian distribution curve has a kurtosis of 3 which is usually

taken as a standard value for the kurtosis. Curves with values of k less than 3
are called platykurtic and those with k greater than 3 are called lepto-
kurtic, (figure 2.13).

Figure 2.13

31
 In plotting distribution curves of events on actual surfaces it has been
shown that this involves drawing smooth curves through the histograms
obtained by summing the frequency of such events at discrete intervals. This
is obviously a fairly tedious procedure and a more useful presentation is to
plot the distributions on probability paper. This has the advantage that the
scales are adjusted such that a truly gaussian distribution degenerates into a
straight line. Figure 2.14 shows the distribution of events on probability
paper for gaussian, rectangular and triangular distribution of events.

Gauss1an d•str•but•on
Tnangular dlstnbut1on
Un1form dtstnbutton

80

01

--o -+
Figure 2.14 Distributions plotted on probability
paper

2.4.2 The Autocorrelation Function of Surface Profiles

The effect of increasing the sampling interval /, although to some extent

affecting the ensuing frequency histogram and distribution curves, can be
most clearly seen by considering the correlation between adjacent ordinates 7 .
This may best be achieved by plotting the variation of the autocorrelation
function, R(/) against the sampling interval /. The autocorrelation function
for a single profile is obtained by delaying the profile relative to itself by some

32
fixed interval, then multiplying the original profile by the delayed one and
averaging the product values over a representative length of the profile.
Thus
R(l) = E[ z(x)z(x + I)] (2.8a)
where E indicates the expected (average) value and z(x) is the height of the
profile at a given coordinate x along the mean line z(x) = 0 and z(x + I) is
the height at an adjacent coordinate (x + I) taken at an interval I from the
previous one. If the value of the ordinates at discrete intervals I is known
this may be interpreted as
J N-1
R(l) = N _ I x~! z(x) x z(x + I) (2.8b)

where N is the total number of ordinates in a sample length L. For a contin-

uous function, equation 2.8b may be more usefully written
L/2

R(l) = lim
L-oc
_!_
L
J z(x) x z(x + I) dx (2.8c)
-L/2

It can easily be seen that when I = 0, R(l) reduces to the variance u 2 or mean
square value of the profile. The autocorrelation function is therefore usually
plotted in its standardised form, r(l), where
r(l) = R(l) = R(l)
R(lo) u2

A typical plot of the autocorrelation function for two different profiles is

shown in figure 2.15. The shape of this function is most useful in revealing

k(~'-­
~;tv¥MA Surface A Surface B

R(()
R(l)

Autocorretotion AutocorrelatiOn
funct1on funct10n

Figure 2.15 Typical surfaces and the resulting autocorrela-

tion functions

33
some of the characteristics of the profile. The general decay of the function
indicates a decrease of correlation as I increases and is an indication of the
random component of the surface profile, while the oscillatory component
of the function indicates any inherent periodicity of the profile. Figure 2.16
shows some typical results for actual machined surfaces where the auto-
correlation functions can be seen to be demonstrating the general features of
the profiles 8 . For this reason it is possible to describe the features of any
surface profile by two characteristics: the height distribution function t/J(z)
and the autocorrelation function r(l).
The main difficulty which may prevent the use of the autocorrelation
function in practical applications lies in the large volume of digital informa-

zoL ~An
so·~UV w~ V~ VV VlJ~
h Shaped surface
C.L A lf;pm

,~
(a)

DI\Af\1\f\ot
V'VV V'VVlJ v

(b)

sL ~ Ground surface

(c l
~o/\ ""LC.LA IOpm

CLA
Superfin1shed
018pm

(d)

Figure 2.16 Examples of engineering surfaces, their

distributions and autocorrelation functions

34
tion which must be obtained from the surface profile for its evaluation.
However, the autocorrelation function may be broken up into two terms
such that the random component of the surface profile may be expressed by an
exponential decay term and the periodic component by a trigonometric
term. For many practical applications, for example the dry hertzian contact
of surfaces, the random component of the surface may be of primary im-
portance, and consequently the problem can be simplified by neglecting the
periodic component of the profile. The surface will then be statistically defined
by an exponential autocorrelation function.
In certain situations the profile may be required to be presented in terms of
its frequency domains. To this end a plot of the power spectra C is useful.
The power spectrum is obtainable from the autocorrelation function using
the expression
c('

C(w) = ~ JR(l) cos wl dl

0

where w = 2rrf, f being the frequency of occurrence. The power spectrum

is in fact the Laplace transform of the autocorrelation function. For the
purposes of this book the use of the power spectra will not be pursued further,
although it clearly has considerable possibilities for the defining of surface
contours.

2.4.3 The Distribution of the Peaks, Valleys, Curvatures and Slopes

Surface profiles might be considered as comprising a certain number of

peaks of varying heights and an equal number of valleys of varying depths.
These features may therefore be assessed and represented by their appro-
priate distribution curves which can be described by the same sort of char-
acteristics as were used previously for the all-ordinate distributions, for
example, aP and av are the standard deviations of the peak and valley
distributions. These distribution curves will obviously involve summing
all the peaks or valleys at any given height level, (figure 2.17).

Figure 2.17 Peak and valley distributions of a surface

35
 As with the all-ordinate distributions of surface textures, the peak and
valley distributions often follow the gaussian curve. Figure 2.18 shows how
close to gaussian distributions are the distributions for the peaks and all-
ordinates for a ground surface.

95 ~L
2501-'m

80

50

••

••
• •- ••

0
01 0
0

0 2 4 6 8
He1gh1 above datum (pm)

Figure 2.18 Distributions of an engineering surface

Although the distribution functions of the peaks and valleys provide the
basic information relating to such features of the surface profile, they do not
include any specific statement on the order in which the particular peaks and
valleys occur along the surface; although it will be recognised that any peak
must necessarily have valleys on either side at a lower level than the peak.
Thus the juxtapcsition of the peak and valley distribution implies some limit
on the choice of order for the peaks and valleys. This is readily seen by the
simple example shown in figure 2.19, where the same distribution of peaks
and valleys always occurs but their separation K varies. Where there is no
overlap of these distributions, figure 2.19a, any peak may be associated with

36
 . . ---1-
.~
....
~ ~-=
.... __ __
......
-~,
1/1 (v) (a)

Figure 2.19 The effect of the interaction of the peak and

the valley distributions

any two valleys so that the order of occurrence of the peaks along the surface
is in no way defined. Where the distributions overlap, figure 2.19b, it is seen
that in regions A and C the foregoing still applies. However in region B
one cannot associate a valley in the upper part of this region with a peak in
the lower part of the same region, since peaks cannot be lower than their
adjacent valleys. Thus it is seen that the greater the degree of overlap of the
peak and valley distributions the more the order in which the peak and
valley occur along the surface is implied.
Similar distribution curves can be obtained for the curvatures of the peaks
and valleys. This may be done by simple curve-fitting-such as parabaloids
or circular arcs-to the peaks or valleys and assessing the way in which the
curvature varies along the profile. A typical distribution curve for the curva-
ture of the peaks of real surfaces is shown in figure 2.20, together with the
gaussian distribution of the same mean and standard deviation.
The way in which the slope varies along the profile can also be described
in terms of its distribution curve. The derivation of this curve will involve
measuring the slope at each point along the surface profile. The distribution
curve may then be characterised by its standard deviation d- (figure 2.21 ).

37
 Radius of asperity peok (mm)

Figure 2.20 Distribution of the asperity peak radii

Meon Mean
ord1note slope

-z -i
All-ordinate distnbut1on All-slope dlslnbut,on

Figure 2.21 Derivation of the slope distribution of

surfaces

2.5 MEASUREMENT OF SURFACE PARAMETERS

Abbott and Firestone measured their bearing-area curve manually using an

optical comparator. For each height considered a line was drawn along the
profile and the fraction of the sample length embraced by the profile at that
height was measured. However, the practical derivation of such curves, in

38
the form of all-ordinate distributions of surface profiles, and indeed of the
distribution curves of all surface parameters, peaks, valleys, curvatures, etc.,
involves the measurement of a sequence of ordinates along the profile.
Such a process can be very tedious since a considerable amount of digital
information is required, but with the aid of modern electronics, the stylus
instrument can be modified to make the operation fairly easy. The usual
practice is to feed the analogue electrical signal from the Talysurf instrument
through a digital voltmeter to a high-speed paper-tape punch to get a per-
manent numerical record of a sequence of profile ordinates at some discrete
sampling interval. The data on the paper tape can then be analysed by a
digital computer for the required information. The height distributions for
the peaks and valleys and their curvatures can be derived by using the
technique of three-point analysis. A peak may be defined when the middle of
three successive ordinates is higher than those on either side. Similarly, the
peak curvatures can be obtained by fitting a curve to the profile using the
three-point analysis.

I. E. J. Abbot and F. A. Firestone. Specifying surface quality. Mech. Engng, 55, (1933),
569.
2. J. B. P. Williamson. The microtopography of surfaces. Proc. lnstn mech. Engrs,
182 (3K), (1967-8), 21.
3. R. H. Reason. The trend of surface measurement. J. lnstn. Prod. Engrs, (May, 1954).
4. N. 0. Meyers. Characterization of surface roughness. Wear, 5, (1962), 182.
5. J. Halling. The specification of surface quality-quo vadis?. Prod. Engr, 51, No. 5,
(1972), p. 171. •
6. J. Halling and M. El-Refaie. A statistical model for engineering surfaces. Tribo/ogy
Convention, Paper No. C60/71, Instn. mech. Engrs., (May, 1971).
7. D. J. Whitehouse and J. F. Archard. The properties of random surfaces of signifi-
cance in their contact. Proc. R. Soc., 316A, (1970), 97.
8. J. Peklenik. New developments in surface characterization and measurements by
means of random process analysis. Proc. lnstn mech. Engrs, 182 (3K), (1967-8), 108.

39
3
Contact of Surfaces
3.1 INTRODUCTION

It is clear that any study of tribology must incorporate a detailed under-

standing of the mechanics of contact of solid bodies. This involves an under-
standing of the nature of the associated deformations and the stresses induced
by any applied loading to bodies of a wide variety of geometric shapes. In
particular we are concerned not only with the deformation and stresses at the
surfaces of solids but also throughout the depth of the surface layers. Any load
inducing a deformation of solids may readily be resolved into a normal and a
tangential component, and it is generally convenient to consider these two
influences separately with respect to the stresses and deformation which
they induce, and then by superimposing the two obtain the total effect. In
such cases the principle of superposition is acceptable since the systems are
essentially statically determinate.
Solid materials subjected to loads deform in either an elastic or plastic
manner; the former deformation being characterised by simple linear rela-
tions between stress and strain and being basically reversible. With plastic
deformation the stress-strain relations are more complex and some deforma-
tion persists even after removal of the load. In most contact situations we
find a mixture of both elastic and plastic deformations. Thus the loads applied
to solids in contact may induce a general elastic behaviour in the bulk of the
solid bodies, but since the actual contact must occur at the tips of the surface
asperities these may be subjected to localised plastic deformation at their tips.
The amount of plastic/elastic deformation must obviously depend on the
value of the applied load and the degree of plastic deformation increases as

40
the load is increased. Thus, in metal-working processes when the nominal
contact pressures are exceedingly large, the amount of plastic deformation of
the surface is increased.
In much of the following we shall be considering the deformation patterns
induced by loads applied to cylinders and spheres. This study is valuable for
two reasons
(a) Many engineering contacts are concerned with the contact of bodies
defined by circular arcs such as wheels on tracks, rolling element type
bearings, gear-teeth contacts, many variable-speed drives and belt
and rope drives.
(b) All solid bodies have surface asperities which may be considered as
very small spherically shaped protuberances. Thus, the contact of
essentially flat bodies reduces to the study of an array of roughly spheri-
cal contacts where we shall concentrate on the deformation of the tips
of such spherical asperities.
Finally, in sliding contacts we are all aware that the work done in over-
coming any friction is finally manifested as a heat release; try sliding down a
rope and this will become painfully obvious. We shall therefore be very
interested in the nature of this heat release, the temperatures which are
induced and the distribution of these temperatures throughout the bulk of
the contacting solids. In sliding down a rope the surface temperature of our
skin is clearly much in excess of the t~mperau of the bulk of our hands.
From the foregoing it is clear that a detailed study of surface contacts will
necessitate a relatively detailed understanding of such topics as elastic and
plastic deformations and the nature of heat conduction due to moving
heat sources. Clearly it is not possible to develop all these arguments in depth
in a book of this type, so that recourse will often be made to physical argu-
ments where the essential conclusions will be stated. In each case the reader
will be able to find extensive verification of such results in standard works on
elasticity, plasticity and heat conduction. It is not the purpose of this book to
develop such arguments, but only to demonstrate the value of such know-
ledge in the particular problems with which we are concerned.

3.2 STRESS DISTRIBUTION DUE TO LOADING

In almost all our studies we shall be concerned with effects within the outer-
most layers of the surface (typically within the outermost millimetre or so of
the surface). The effects at several centimetres below the surface are of only
secondary importance so that we may often treat the surfaces, from a physical
standpoint, as though they represent the surfaces of bodies of essentially
infinite depths, that is, they may be considered as semi-infinite bodies. This
device enables us to concentrate on the details of the surface contact of solids
rather than considering their overall geometric shape and thereby leads to
considerable mathematical simplification.

41
 Consider a single normal line load P per unit length in the plane xz and
applied at a point 0' defined by the coordinates (t:, 0) on the surface (z = 0)
of a semi-infinite solid and having the same value for all values of y, see
figure 3.1 a. The elastic stress field in the plane xz is readily obtained. Consider-
ing a unit length in they direction, the radial stress (J, will be given by 1

2P
(J, = - -cos 0, (3.1)
nr
the tangential stress (J 9 and the shearing stress r, 0 being equal to zero.
This represents a state of simple radial compressive stress, the stress
increasing with decreasing radius r and decreasing angle 0. The use of the
two-dimensional Mohr's circle of stress, figure 3.1 b, for these stresses gives

z
(a)

(b)

(b)

Figure 3.1 Stress distribution due to a line load acting on a

semi-infinite body

42
us the resulting cartesian stresses with respect to 0'
ux = i (1 -cos 20) = u, sin 2 0

2
= _ 2P (sin 0 cos 0) = 2P [ zx 2 J
1t r 1t (xz + z2f
u. = i (1 + cos 20) = u, cos 0 2

=- :
2
t<x2 z;xz2)2] (3.2)

or with respect to the original origin 0 as

(3.3)

A similar approach may be applied to obtain the stresses due to a single

tangential line load T acting at 0', (figure 3.2) where
2T
u =--cos O'
r nr
(3.4)

and

(3.5)

If we make T = p.P, where J.l is the appropriate coefficient of friction and

add the stress components due to P and T at any point (x, y) we will clearly
43
 T
o·
8' X

Figure 3.2 Stress due

to a tangential line load
acting on a semi-infinite
body
have the stress distribution arising in a simple frictional contact. The solu-
tion, however, suffers from one serious drawback. If we examine equations 3.1
and 3.4 we find that at 0' (r = 0) the stresses are infinite and such a situation
is obviously inadmissible. This arises from our basic assumption that the
load acts at a single point, that is, over zero contact area. In real cases we
must therefore always have some finite area of contact and this clearly
changes our initial problem. Fortunately, we are still able to use our original
solution for the new situation.
Consider a uniformly distributed load giving rise to a contact pressure p
over a region 0 to a on the surface (z = 0) of a semi-infinite solid, figure 3.3.
Taking a length along the y direction equal to unity, it will be recognised
that the total load P is given by
a

P = J pdx = pa
0

If we consider a vanishingly small load p dt: at some point defined by the

coordinates (t:, 0) we may obtain the stress at any point (X, Z) due to this
Pa

0
X

.- --j 1-
d£_3 z

x-.-
X

Figure 3.3

44
load using equations 3.3. In this case P will be replaced by p de. The total stress
at a point X, Z due to the distributed load P is then clearly obtained by the
summation of the effects of all the p de loads acting at different values of e
from 0 to a, or mathematically speaking

(3.6)

rxz = -1[
2p fa [ Z 2 (X - e)
[(X - e)2 + z2]2
J de
0
In a similar way if we consider a tangential load T = JJ.P distributed over
the region 0 to a, then at every point it follows that t dx = Jl.P dx (figure 3.4)
and
a a

T = f
0
t dx = f
0
Jl.P dx = JJ.P

1----- X ----4
z

Figure 3.4

By using equations 3.5 for each elemental tangential load t de acting on

element de (0, e) we can obtain the stresses at any point (X, Z) due to the total
distributed load T, thus

(3.7)

•xz = - n2t fa [ [(X_Z(Xe)2- +zip

e) Jde 2

0
45
 For a sliding contact subjected to a normal load P uniformly distributed
over the contact 0 to a, the total stresses are the sum of the stresses given by
equations 3.6 and 3.7 above.
It is clear that the use of the basic normal and tangential point load solu-
tions may be used to obtain the resultant stress distribution for any type of
load distribution over the contact region. All of the preceding solutions have
assumed elastic behaviour of the bodies, but we have already anticipated that
we are also interested in the possible plastic effects in such situations. The
simplest criterion for the onset of plastic deformation assumes that this occurs
when the maximum shear stress reaches the critical shear stress for the
material k, where k = Y /2, Y being the tensile yield stress. For the cases
considered above, where plane strain conditions apply, the maximum shear
stresses always occur in the xz plane. The maximum shear stress in this plane
is simply the radius of the Mohr's circle of stress, (figure 3.1 b), that is

a, p (}
r max =2=- 1tr cos

If we consider a circle of diameter b drawn in the manner shown in figure 3.5a,

we find that r = b cos (} and

that is, the stress remains constant at all points on the circle. It is therefore
useful to plot the stress distribution as isochromatics or lines of constant
rmax• and it is then possible to determine the location at which rmax will reach
its limiting value of k, that is, the location of the onset of plastic deformation.
These plots are useful since they also indicate the pattern of isochromatics
obtained in photoelastic stress analysis. For a point normal load, not strictly
achieved in practice, and a uniformly distributed normal load, calculations of
rmax yield the pattern of isochromatics shown in figures 3.5b and c. It is seen
that in both cases the material will first reach a yield condition at the surface
where increasing load gives rmax = k, the yield shear stress of the material.
p p p

(o) (b) (c)

F~qure 3.5 Patterns of lines of constant maximum shear

stress ( isochromatics)

46
3.3 DISPLACEMENTS DUE TO LOADING
Having obtained the stress distribution we can now obtain the displacements
in a solid using the usual equations which relate the strain e and the cor-
responding displacements. Thus for a single normal load P acting at 0',
figure 3.la, the horizontal and vertical displacements u and w are given by

ou l
- = e = - (u - vu8 ) = - -cos 0
2P
or ' E ' rrr E

u ow l
- + -- = e = - (u
2P
vu) = v - cos(}
0 8 -
r rt!O E ' rrrE

au c'w w l
r o~
c
+ -;;-
ur
- -r = 'Yr8 = -G r,o = 0

To solve these equations we require a knowledge of the boundary conditions.

For this we assume that points on the z-axis, that is, at 0 = 0, have no lateral
displacements and that at a point on the z-axis at a distance b from the origin
there is no vertical displacement. Clearly we are interested in the displace-
ments occurring at the boundary, z = 0, of the solid, thus by putting
0 = ± rr/2 in the solution of the above equations it can be shown that the
horizontal displacement is given by
(l - v)P
(u)z=O = - 2£ (3.8)

This indicates that at all points on the boundary of the solid there is a constant
displacement directed toward the origin. We can also find that the vertical
displacement of a point on the boundary z = 0 at a distance x from the
origin is given by
2P b (! + v)P
(w)z=o = ~E log~- rr£ (3.9)

Here we also find that at the point of load application (x = 0) the vertical
displacement becomes infinitely large. As mentioned earlier, this is due to our
assumption of a point load, but in practice the load is usually distributed over
a finite area. In this case if the load is distributed over the region 0 to a,
figure 3.3, giving rise to a contact pressure p, the vertical displacement at any
point (X, 0) produced by an element of load p de at a distance c from point
0 is known from equation 3.9 by substituting p d<: for P and (X - c) for x
so that the total displacement at point (X, 0) is given by
a a

(w)-= 0 = 2
•
-
rrE
Jp l oX gb- -c - d c(I-rrE+- -v) Jpdc (3.10)
0 0

In the foregoing we have concerned ourselves with problems of a two-

dimensional nature where the deforming solid is considered to be subjected

47
to plane strain conditions. It is clear that the three-dimensional analogues
of these problems necessitate a more complex treatment, however, the results
are of great importance and are often met in tribological situations. For the
purposes of this book it will suffice to state these results, detailed treatments
can be found in standard books on elasticity. Thus, for a point normal load
P acting on a semi-infinite solid, the horizontal and vertical displacements
along the boundary z = 0 at a distance x from the point of load application
are found to be given by
(I - 2v)(l + v)P
(u)z = o = - 2rr£x (3.11)

(I - v2)P
(w)_=o = EX (3.12)
- 1r

For a load distributed over a region of the boundary, giving rise to a pressure
p acting on an element of area dA taken at some distance x from a point,
the vertical displacement at that point is given by
_ (I - v2 )
(w)z=O- - - -
f p dA (3.13)
rr£ x
The integral in the above equation is, in fact, an elliptic integral.

3.4 HERTZ IAN CONTACTS

As mentioned earlier we are particularly interested in the problems of contact

between bodies whose geometry is defined by circular arcs. This problem was
first solved by Hertz for elastic contact and is generally referred to as hertzian
contact.
Consider the contact of two identical cylinders under conditions of plane
strain. From arguments of symmetry we can see that the zone of contact will
be created by compression of the cylinders to generate a straight line, that is,
produce a plane contact zone, figure 3.6a. Although this is no longer strictly
true for a cylinder in contact with a plane, the error is such as to be negligible
and for this situation a plane contact zone may be assumed. One feature of
such contacts becomes immediately apparent, in that as we increase the load
the width of the contact zone will increase and, by noting that the deforma-
tion at the centre of the zone is larger than that at the extremities, we would
expect a contact pressure which is no longer constant. This problem is there-
fore more complicated than the preceding cases and, before considering the
stress distribution, we need to define both the contact pressure distribution
and the actual size of the contact zone for any given applied load. Detailed
treatment of this problem is found in books on elasticity, but for our pur-
poses we shall use a simpler physical argument.
For two identical elastic cylinders in contact under a normal load P

48
 p

1--- 2o ------.j (b)

(a)

Figure 3.6 Pressure distribution for the contact of two

cylinders

per unit axial length, let the resulting plane contact zone have a width of 2a,
figure 3.6b. Since the normal deformation at the centre of the contact zone is
greater than at the extremities, it is not surprising that the actual contact
pressure distribution p has the form 1
p = 2P
n:a
(1 - a~)12 2
(3.14)

Without this information it is physically apparent that the stresses in such

a system would be such that

stress oc ( ~)
Considering the deformation, we note that increasing the load would increase
a and thereby increase the strain so that we would expect the non-dimensional
strain to be related by

strain oc ( i)
where R is the radius of the cylinder. From these relations for stress and
strain we have
2 PR
or a oc-
E
The actual solution for this case is in fact 1

2 4PR(l - v2 )
a=----- (3.15)
nE
49
which clearly substantiates the above simplified argument. The solution
defined by equations 3.14 and 3.15 is almost true for other than identical
cylinders, that is, plane contact geometries. Provided the angle subtended
by the contact width at the centre of the cylinder is less than 30", the results
may be reasonably used for other contact geometries, figure 3.7, by using E'
and R' where
1 I - vi 1 - v~
E'=~+
I 2

and

whence
4PR'
az =~-
rtE'

For the case of a cylinder on a plane, the radius of the plane is taken as
infinity, thereby R' becomes the radius of the cylinder only and for concave
curvatures the radius is taken as negative. It is also worthwhile noting that
when E --+ ·X, the solids become rigid leading to a single point contact where
a--+ 0.

G
(a) (b) (c)

Figure 3.7

3.4.1 Stress Distribution in Hertzian Contacts

Earlier in this chapter we have seen that the onset of plastic deformation may
be associated with the maximum shear stress reaching a critical value k. We
therefore proceed to study the distribution of the maximum shear stress for a
body loaded by a pressure distribution given by equation 3.14, acting over the
contact zone -a to +a. Using equations 3.6 for elemental loads p dt: and
integrating for the actual distribution of P will give the cartesian stress distri-

50
bution within the body. Thus

-a

(3.16)

Again, the maximum shear stress for plane strain conditions is given by the
radius of the Mohr stress circle, see figure 3.1 b, that is

rmax = [( -"'~
(J - (J )2 + Jli2
r;z (3.17)

where (JX' (Jz and rxz are defined by equations 3.16. Equation 3.17 therefore
defines the values of rmax at all points. The evaluation of this equation will
enable us to draw the isochromatics and obtain the somewhat surprising
result shown in figure 3.8. We note immediately that the greatest value of
rmax occurs below the surface at a distance of 0.67a. Furthermore, we find
that as the load is increased rmax at this point will reach the value k when the
maximum pressure at the centre of the contact zone Po is 3.1k. This is again a
surprising result since if we had simple compression conditions we would
expect the surface material to yield when Po reaches a value of 2k. This does
not oceur because the surface elements are subjected to compressive stresses
in all three orthogonal directions, allowing Po to be greater than 2k without
producing yield. This is equivalent to saying that the hydrostatic component
of stress at any point cannot contribute to the plastic deformation of the
material at that point. This is. a very important result since it means that
contact pressures in excess of the yield value for the material do not result in
plastic deformation, so that higher loads than might have been expected can
be carried elastically with hertzian type contacts. Moreover, it will be recog-
nised that even when yielding has taken place below the surface, very little
plastic deformation can occur because the plastic zone is constrained by
elastic material on all sides.

51
 Lines of
&.-o~ consranl
rmax

Figure 3.8 Actual isochromatics obtained for the

contact of a cylinder and a plane due to norma/load
alone

As the load is increased further, the plastic zone will increase in size and
ultimately spread to the surface of the body. Plastic flow may then occur
fairly readily and the cylinder will indent the surface of the body. This will
happen when the mean contact pressure Pm is about 6k, that is, more than
twice the contact pressure at which initial yield occurred 2 . The mean pressure
under these conditions is essentially the indentation hardness value of the
material, H, and this is why for metals we find that
H::::: 6k::::: 3Y
where Y is the uniaxial tensile yield strength of the material; see figure 3.9.

Fully plaSIIC
behaviour

PlasiiC behav1our
cons1ra1ned by
surroundmg elasiiC malenal

Purely elasiiC
behaviour

Figure 3.9

52
 We have as yet only considered normal loads applied to hertzian contacts
and it is apposite to ask what happens in the presence of both normal and
tangential loads. The stress field due to a tangential load !J.P may readily be
obtained by the method discussed earlier, since at all points it is clear that the
tangential traction t = !J.P· Combining the stress distribution due to the
normal and tangential loads and calculating the values of rmax leads to the
isochromatic pattern shown in figure 3.10. It is seen that the location of the

Lines of
constant
Tmox

Figure 3.10 Actual isochromatics obtained for the

contact of a cylinder and a plane due to combined
normal and tangential loads where T = 0·5!J.P

greatest value of the maximum shear stress is now much nearer to the surface,
thus the plastic deformation can take place more readily than in the previous
case. In other words, macroscopic plastic deformation is facilitated by the
presence of such friction tractions.

3.4.2 Conditions When T ~ !J.P

In many practical situations contacting bodies are subjected to tangential
loads less than !J.P so that macroscopic sliding does not occur. This happens in
situations where friction is used as the mechanism for preventing sl;p
between mating components, for example, nuts and bolts, interference
fits and friction drives such as clutches. Most textbooks explain this situation
by allowing the coefficient of friction to increase from zero to a limiting value
at which slip occurs. This is clearly inadmissible since it makes what should
be a physical constant into a variable. In what follows we show that this
assumption is not necessary if we recognise that the contacting bodies are
deformable rather than rigid. The mechanism will be explained by consider-
ing a cylinder pressed against a plane and subjected to a tangential load less
than !J.P.

53
 We shall start by assuming the answer to our problem and subsequently
justify its validity. What happens is that within the contact zone there exists
a central area in which no slip occurs, while at the two extremities a small
degree of slip takes place, see figure 3.11. The coexistence of a zone of sticking
and zones of microslip is possible because of the deformable nature of the
materials in contact and the deformation pattern being such as to allow slip
at the extremities of the contact zone. As the value ofT increases, the areas of
slip increase until, when T = pP, they meet at the centre of the contact

(a)

(b) (c)

~)
[ 0:
\ I
I I
®
Contocf
zone
1
1
fP
Ip
Contact
zone
[CD ,
1
~
® 1-1 --1-L-- ®I
1
Comp ~
1e
~
1
Comp Comp
le
~
'
~Tensil
I I I I I I I I

~
1eA' 1
Comp ~Camp Tensil~ ,....__,
:eA'zl Comp

; BOdy I :

~l•p
~ Body 2

Figure 3.11

and macroslip takes place throughout the contact zone. With this model it is
possible for J1 to have a constant value wherever slip occurs, that is, within
the slip regions t = JlP while within the stick region t < pp. Since T is the
integral oft over the contact zone this can be seen from figure 3.11 to satisfy
the requirements of the problem with J1 always having a constant value.
For the case when T = pP, the distribution of normal pressure p and
tangential traction t = JlP are as shown in figure 3.11 b and 3.11 c. Increasing
the normal load induces equal compression strains ex in both bodies so that

54
no slip occurs due to this effect. With the tangential load, however, we see that
since these must be opposite in direction on the two bodies they will cause
the patterns of strain shown in figure 3.11 c, so that slip must occur every-
where throughout the contact zone. For T < JLP we therefore see that the
central stick areas must have zero strain in both bodies and this consequently
defines the type of distribution for the tangential tractions.
The distribution of tangential traction which results in zero strain over the
central stick area is illustrated by the argument shown in figure 3.12. Con-
sider a tangential load T = JLP which results in the traction distribution

Col

Slope
(b)

(c l
I
I
K:±=:k
I I I
I
I
r"

(d)

~
I I I I I
Slope

Cel I I~ I I I I
T=Ti-T"

I I I I I
I I I I

(f)

~ I
Slip
I SliCk I
Slip
I
ex= e; +e;

-a -a a a

Figure 3.12

shown in figure 3.12a. The strain e~ resulting from such a distribution is

shown in Figure 3.l2b. It will be noted that this strain follows a linear law
within the width of the contact zone and that the slope is, from -a to +a,
e~ = J1X/2R. Now consider a similar shape of tangential traction distribution,
T", in the opposite sense applied over the region -a. to +a.. The strains, e~,
resulting from such a distribution are as shown. and it will be seen that the
slope of the strain distribution is exactly opposite in sign, but of the same
magnitude, as that due to T'. Thus by adding the strains e~ and e~ one

55
obtains a zero strain over the region - rx to + rx and a tangential traction in
this region where t < llP· In the remainder of the contact zone one obtains
t = llP and strains which are non-zero and consequently area of slip, figures
3.12e and 3.12f.
The above arguments show that even when no macroscopic motion occurs
some degree of microslip exists when T < 11P and this gives rise to a mech-
anism known as fretting. For more complicated contact geometries these
arguments are still qualitatively correct and microslip will occur at the
extremities of the contact zone.

3.4.3 General Three -Dimensional Contact

For simplicity we have so far concerned ourselves with two-dimensional

hertzian contacts. In many practical situations we must be able to deal with
the more complicated three-dimensional problems and the following out-
lines the main results of such analysis. In general, the patterns of behaviour
are analogous, but some of our preceding formulae must be modified.
If we subject two identical spheres to a normal load N, the area of contact
will clearly be a plane circle of radius a and the pressure distribution is
hemispherical in form, figure 3.13, and is given by 1

p= 3N ( 1 (3.18)
2rra 2
The value of a is given by

a= (3NR)
8£'
n 1
(3.19)

Contact
zone

Figure 3.13 The pressure distribution due to

the contact of spheres

56
We can recognise that this is correct by applying the simple analysis used
previously.
Although the contact of two dissimilar spheres does not result in a plane
contact area, the results of equations 3.18 and 3.19 still hold with substantial
accuracy, but in this case we have

a= (3NR')1/3 (3.20)
4E'
where R' is related to the radii of the spheres R 1 and R 2 by
1 1 1
R'=p:+-R
1 2

For the contact of a sphere and a plane, R' is merely the radius of the sphere.
A more general approach to the problem is to consider the contact of
two bodies 1 and 2 whose contact geometry is defined by the principal radii
of curvature of each body in two orthogonal planes, figure 3.14. The contact

I
I I
I ..L....l.
rz4.2b
-I 2a 1-r
Plane 1 Plane 2

(R12 radous of body 1 in plene 2 ercl

Figure 3.14

geometry is now clearly elliptical in shape and the contact pressure distri-
bution is given by
p= 3N (1 - x2 - z2)1/2 (3.21)
21tab a2 b2
The size of the contact ellipse is defined by the semi-major and semi-minor
axes a and b which are given by
3N ] 1' 3
a = ka[ 4E'(A + B)

h=
r 3N ]1'3
kl4E'(A + B)
(3.22)

57
where ka and kb are constants which depend on the values of the principal
curvatures of the contacting bodies and on the angle </> between the normal
planes which contain these curvatures. If we denote the principal radii of
curvature of body 1 by R 11 and R 12 and those of body 2 by R 21 and R 22 ,
then the constants A and B may be defined by

1 1 )2 I 1 )2
B- A = 21 [( R,, - R,2 + ( R2, - R22

+ 2( - 1 - - 1 )(- 1 - - 1 ) cos 2</> ]1/2 (3.23)

R,, R,z Rz, R22

A+ B = !(_1_ + _1_ + _1_ + _1_)

2 R 11 R 12 R 21 R 22
In these equations any concave curvature is taken as negative. The coeffi-
cients ka and kb in equations 3.22 are numbers which depend on the ratio
(B - A)/(A + B) and these coefficients can be obtained by introducing an
auxiliary angle y defined by
B-A
cosy= A+ B

Thus, using equations 3.23, the value of y can be easily determined. To

evaluate the values of ka and kb corresponding to a certain value of y we
require complicated numerical calculations involving elliptic integrals. The
results of such calculations are given in figure 3.15.

20
\ ~
~

10
"' ~
!'-- ...............
..--- __..
-- _:::- ~

40 50 60 70 80
r•
Figure 3.15

58
 When dealing with such complicated geometries it is clear that our assump-
tion of a plane area of contact will no longer be true. As mentioned earlier,
while the pressure distribution and the size of the contact as determined from
hertzian theory are still substantially correct, we sometimes need to know the
actual shapes of such contacts. For materials having the same elastic proper-
ties it is sufficient to assume that the deformed surface, which has some com-
mon radius Rc, is about midway between the two original surfaces, figure
3.16. Thus the value of the common radius of curvature is given by 1

R =. 2RIR2 (3.24)
c Rt - R2

It is clear that for two identical spheres the above equation gives the expected
result of a plane contact area. Where concave curvatures occur, the radius is
taken as negative.
I
I
I
/
., /

Figure 3.16

Subsequently, we shall find that it is necessary to define the normal approach

of a sphere due to the application of normal load and the consequent deforma-
tion. Consider the contact of a sphere and a plane, figure 3.17. It is easily
seen that the separation u of the surfaces at a distance r from the centre of the
contact zone is given by
u = R - (R2 - r2)tt2

If r is small compared to R then

r2
U=- (3.25)
2R

59
 ----r---···._··...... ~· ....... •

(a)
~a
(b)

Figure 3.17 The elastic contact between a sphere and a plane

The normal approach is defined as the distance which points on the two
bodies remote from the deformation zone move together on application of a
normal load. This arises from the flattening and general displacement of the
surface within the deformation region. If a is the radius of the contact zone
and w is the displacement of the sphere at the boundary of this zone then the
normal approach c5 will be given by

(3.26)

Obviously, at the centre of the contact zone, c5 is given by the degree of deform-
ation and it is therefore reasonable to assume that the normal approach will
be proportional to the flattening of the sphere. In other words
a2
c) oc-
R
From the results given by equation 3.20 we have

aoc - (NR)t/3
E'
so that
c)
N2
oc ( - -
)t/3
E'2R
The exact results show that

or
(3.27)

60
Combining equations 3.20 and 3.27, the area of contact, A, will be given by
A = rra 2 = rrRJ (3.28)
It will be noted that the relation given by equation 3.28 indicates that the
surface outside the contact region is displaced in such a way that the actual
area of contact is only one half of the geometrical area, which is clearly
equal to 7tiX 2 = 2rrRJ.

3.5 THE CONTACT OF ROUGH SURFACES

Although in general all surfaces have roughness, we shall find some simpli-
fication if we consider the contact of a single rough surface with a perfectly
smooth surface. The results from such an argument are then reasonably
indicative of the effects to be expected from real surfaces. Moreover, the
problem will be simplified further by introducing a theoretical model for the
rough surface in which the asperities are considered as spherical segments
so that their elastic deformation characteristics may be defined by the
hertzian theory. We shall also assume that there is no interaction between
separate asperities, that is, the displacement due to a load on one asperity
does not affect the heights of the neighbouring asperities.
Consider a surface of unit nominal area to consist of an array of identical
spherical asperities all of the same height z with respect to some reference
plane X X', figure 3.18. As the smooth surface approaches due to the applica-
tion of load we see that the normal approach will be given by (z - d), where

Figure 3.18 Contact between a smooth plane and a idealised

rough surface

d is the current separation between the smooth surface and the reference
plane. Clearly, each asperity is deformed equally and carries the same load
N i so that for '1 asperities per unit area the total load N will be ·equal to
IJNi. For each asperity, the load Ni and the area of contact Ai are known
from the hertzian theory, see equations 3.27 and 3.28. Thus if f3 is the asperity
radius we have

and
Ai = rrfJ(z - d)

61
and the total load will be given by

N = 1t~E'f32(/JY 12

that is, the load is related to the total real area of contact A = t~A; by

(3.29)

This result indicates that the real area of contact is related to the two-thirds
power of the load, when the deformation is elastic.
If the loads are such that the asperities are deformed plastically under a
constant flow pressure H, which is closely related to the hardness, we assume
that the displaced material moves vertically down and does not spread hori-
zontally so that the area of contact A' will be equal to the geometrical area
2nf3t5. The individual load N; will then be given by

N; = H A; = 2Hnf3(z - d)
Thus
N' = t~N; = t~HA; = HA' = 2HA (3.30)

that is, the real area of contact is linearly related to the load.
It must be pointed out at this stage that the contact of rough surfaces
should be expected to give a linear relationship between real area of contact
and load, a result which is basic to the laws of friction (chapter 4). From our
simple model of rough surface contact we see that while a plastic mode of
asperity deformation gives this linear relationship, the elastic mode does not.
This is due to our simple and hence unrealistic model of the rough surface.
Later we shall find that the proportionality between load and real contact
area can in fact be obtained with an elastic mode of deformation when we
consider a more realistic surface model.
It is well known that on real surfaces the asperities have different heights
indicated by a probability distribution of their peak heights. We must there-
fore modify our previous surface model accordingly and the analysis of its
contact must now include a probability statement as to the number of
asperities in contact 3. If the separation between the smooth surface and
reference plane is d, then there will be contact at any asperity whose height
was originally greater than d, figure 3.19. If cf>(z) is the probability density
of the asperily peak height distribution then the probability that a particular
asperity has a height between z and z + dz above the reference plane will be
c/>(z) dz. Thus, the probability of contact for any asperity of height z is

prob(z > d) = f"'

d
cf>(z) dz

62
 z
Smooth surface

x'

Otstnbutton
of peak hetghts ¢ (zl

F(qure 3.19 Contact between a smooth plane and a rough

surface

If we consider a unit nominal area of the surfaces containing 17 asperities,

the number of contacts n will be given by
oc

n = 11 Jcj>(z) dz (3.31)
d

Since the normal approach is (z - d) for any asperity and Ni and Ai are known
from equations 3.27 and 3.28, the total area of contact and the expected load
will be given by
x;

A = 1t17[J J(z - d)cj>(z) dz (3.32)

d

and

f
oc

N = 111Pl!2E' (z- d)3/2cj>(z) dz (3.33)

d

It is convenient and usual to express these equations in terms of standardised

variables by putting h = dja and s = zja, a being the standard deviation
of the peak height distribution of the surface. Thus
n = 17F 0 (h)
A= 1t17{JaF 1(h)
N = 111P 1' 2a 312 E'F 312 (h)
where
"
Fm(h) = f (s - hrcf>*(s) ds
h

cf>*(s) being the probability density standardised by scaling it to give a unit

standard deviation.

63
 Using these equations one may evaluate the total real area, load and
number of contact spots for any given height distribution. Experimental
confirmation of the validity of this method has recently been demonstrated 4 .
An interesting case arises where such a distribution is exponential, that is

In this case we have

so that
n = '1e-h
A = 1t1Jf3ue-h
N = n•'z'1f3t12u312E'e-h
These equations give
N = C 1A and N = C2 n
where C 1 and C 2 are constants of the system. We find, therefore, that even
though the asperities are deforming elastically, there is exact linearity
between the load and the real area of contact.
For other distributions of asperity heights such a simple relationship will
not apply, but for distributions approaching an exponential shape it will be
substantially true. For many practical surfaces the distribution of asperity
peak heights is near to a gaussian shape (see chapter 2) and this itself is
near enough to exponential at the outermost decile of the distribution for the
above result to be valid.
Where the asperities obey a plastic deformation law, equations 3.32 and
3.33 are modified to become

f
00

A' = 2rt'1fJ (z - d)¢(z) dz (3.34)

d

N' = 2rt1Jf3H J(z - d)¢(z) dz (3.35)

d

We see immediately that the load is linearly related to the real area of contact
by N' = H A' and this result is totally independent of the height distribution
¢(z), see equation 3.30.
We have so far based our analysis of surface contact on a theoretical model
of the rough surface. An alternative approach to the problem is to apply the
concept of profilometry using the surface bearing-area curve discussed in
chapter 2. In the absence of asperity interaction, the bearing-area curve
provides a direct method for determining the area of contact at any given

64
normal approach. Thus, if the bearing-area curve or the all-ordinate distri-
bution curve is denoted by t/l(z) and the current separation between the
smooth surface and the reference plane is d, then for a unit nominal surface
area the real area of contact will be given by

f
00

A = t/f(z) dz (3.36)
d

so that for an ideal plastic deformation of the surface, the total load will be
given by
00

N = H J t/f(z) dz (3.37)
d

We may summarise the foregoing by saying that the relationship between

the real area of contact and the load will be dependent on both the mode of
deformation and the distribution of the surface profile. When the asperities
deform plastically, the load is linearly related to the real area of contact for
any distribution of asperity heights. When the asperities deform elastically,
the linearity between load and real area of contact occurs only where the
distribution approaches an exponential form and this is very often true for
many practical surfaces. These results will be found to be of considerable
significance when considering such effects as friction and wear.

3.6 CRITERION OF DEFORMATION MODE

It will of course be recognised that in most practical situations the higher
asperities could be plastically deformed while the lower contacting asperities
are still elastic. Thus we obtain a mixed plastic-elastic system in most real
contacts, where the greater the load and hence the normal approach, the
greater the number of plastic contacts. We therefore see that the normal
approach will in a sense be an indicator of the degree of plasticity which
exists. Referring to equations 3.27 and 3.28 we find that for an elastic asperity
contact the mean pressure Prn is given by
4E'b'i2
Prn = 3rt{Jl/2
or
bl/2 = 31t /3 Prn
1/2
(3.38)
4£'

For a spherical contact we know that the transmission from purely elastic to
completely plastic behaviour occurs over a range of loading. Plasticity is
initially sub-surface when the maximum contact pressure is 3.lk or a mean
pressure of about Y, and becomes macroscopic when the mean pressure is

65
about 3 Y, that is, the hardness value of the material (see section 3.6). Thus in
equation 3.38 we see that the transition from elastic to fully plastic behaviour
occurs in a range of values of <5 112 , and the initial deviation from elastic
behaviour occurs when Pm = H/3 where

<)1/2 = 0.78c1;H)

Recognising that the elastic-fully plastic transition IS not instantaneous

we shall assume a transition point where
<)1/2 ~ ~ ({3)1/2
It is convenient to standardise this by dividing both sides by <J 112 so that

<5* 112 = (~) 1/2 = ; (~Y/2

This parameter decreases as the surface roughness as specified by <J increases,
and it is usual to define a function Q, the plasticity index, as the inverse of
<5* 112 so that
Q-_E' - ((J)1/2
- (3.39)
H {3
This index is clearly indicative of the onset of significant plastic deformation,
being large when the contact is basically plastic and small, say less than unity,
when the contact is essentially elastic. The plasticity index is clearly a useful
parameter since it is a non-dimensional grouping relating both the current
physical and geometrical properties of the surface. In a wear-process study
there are obviously advantages in making measurements of Q at discrete
time intervals, since the trend of such a parameter will then indicate whether
the surfaces are approaching an elastic state, that is, a run-in condition, see
figure 3.20.

~ "' 6
u

~
"'

-
4

2
\ r- ~
.............

0
2 4 6 8
T1me (h)

Figure 3.20 The variation of the plasticity index during

a wear process

66
3.7 THERMAL EFFECTS

In a sliding situation most of the work done against friction will be mani-
fested as a heat release with a consequent rise in temperature. Clearly, since
the heat release is basically a continuous process, temperature gradients will
develop in the contacting bodies with the highest temperature occurring at
the point of heat release (heat source), that is, the contact surface. The points
at which heat is released will obviously depend on the overall geometry of the
contacting bodies. If we consider, for instance, our previous surface contact
model of a smooth plane and a rough flat surface then heat will be released
at each contact spot whose size is determined either elastically or plastically.
Thus each contact spot may be treated as an independent heat source and the
individual temperature analysis may then be applied to obtain the tempera-
ture distribution for the general contact region. The same argument is
obviously true for the contact of curved surfaces, however, in this case the
contact spots will be so close together that at high normal loads the whole
contact zone may be regarded as a single heat source. This argument is
justified by the fact that in an analogous electrical example the results show
that the collective resistance of a large number of closely grouped constrictions
is nearly the same as the resistance of one equivalent large constriction. It
will also be recognised that the temperature effects will depend on whether
a body is stationary or moving with respect to the heat source. Thus if we
consider the simple example of two discs in contact we shall obtain a physical
feel for the problem. In such a consideration we must realise that the thermal
effects happen to discrete particles of material and therefore by contemplating
their contact history we can anticipate their overall behaviour which is
clearly related, at any given time, to the integrated effect of their history.
Consider the situation shown in figure 3.2la where two contacting discs
are assumed to be rolling with very small relative slip. It is clear that all

(al (b)

Figure 3.21

67
particles on the surfaces of both discs pass through the contact zone where
heat is generated and afterwards undergo considerable periods of rest. Their
temperature rise will therefore be relatively modest, partly because the
frictional work is small, owing to the small relative sliding, and partly because
the generated heat will be readily dissipated by heat losses during the rest
periods. If one disc is fixed as in figure 3.21b, we will clearly have a state of
pure sliding. In this case, surface particles on disc 2 will be subjected to rela-
tively high temperatures, sometimes called flash temperatures, when passing
through the contact zone and thereafter have a considerable rest period where
cooling takes place. Surface particles on disc l, on the other hand, never
escape from the contact zone and are therefore subjected to a continuous
build-up of temperature towards a steady state determined by the thermal
properties of the whole system. In this example the contact zone, which is
clearly fixed in space, can be regarded as a stationary heat source with respect
to disc I, and a moving heat source with respect to disc 2, and we shall treat
the problem simply by considering these two distinct effects on the two bodies.

3.7.I Stationary Heat Source

Consider a semi-infinite body subjected to a stationary heat source acting

over a small circular surface area of radius a. Clearly, in this case heat will be
supplied to the body through a fixed area and steady state conditions will
exist. This problem is analogous to its electrical counterpart and the heat
flow is considered as a flow of thermal current through a thermal resistance.
If Q is the rate of heat supply then it can be easily shown that the mean
temperature rise, MJ, at the surface is given by

fl.()
Q
= --
4aoc
where oc is the thermal conductivity of the body. If the temperature of other
points on the body at a far distance from the source is taken as zero, then
the above equation will give the mean surface temperature of the body, that is
Q
() = - (3.40)
• 4aoc
The same argument will reasonably apply to a slow-moving heat source
provided that its speed V is so small that at each position of contact there is
sufficient time for the temperature to acquire the same distribution as that
induced by a statK->nary heat source. For a fast-moving source equation 3.40
no longer applies and this occurs at some speed which is related to a certain
parmet~ by 5
(3.41)

where p is the density and c is the spe:::ific heat of the body. For ~ > 5 the
speed is considered high and we must apply moving heat source analysis.

68
3.7.2 Moving Heat Source

For a moving heat source traversing the surface of a semi-infinite body at a

relatively high speed V, we can neglect the effects of the transverse flow of
heat and the problem can be regarded as one of a linear heat flow. In this case,
if the heat is supplied at a constant rate of q per unit area, then the mean
temperature rise of a point on the surface of the body is given by 5
2qt112
!:!f) = ----='------:--;-,--
(mxpc)112

where t is the time during which heat has been supplied. If heat is supplied
through a circular area of radius a, then by putting q = Qjrr.a 2 and by
considering the effective value oft for all points within this area we can obtain
the mean surface temperature. The traverse time for any point (x, y) is
given by

and therefore the mean effective time is

a

1
t-- -
~
J-a
2(
V
2 -y 2)1/2 d y- -an
-
4V
0
Therefore
2Qa112rr.112 0.318Q
f) = ----,;--:--,-,;----;-;-;;- (3.42)
m 2rr.a 2rr. 112 (apcV) 112 a(aapcV) 112

We may summarise the foregoing results by introducing a standardised

surface temperature, 8*, defined by
f)*= pcV 0
rr.q
so that for a stationary heat source, using equation 3.40, we have

o: = a~:v = 0.5~ (3.43)

and for a moving heat source, using. equation 3.42, we have

f)* = 0.318 X (2apcV) 112 = 0.4 38 ~ 12 (3.44)

m (2a)1/2

where ~is as defined by equation 3.41. Examining equations 3.43 and 3.44
we find that for small values of~ or low speeds both stationary and moving
heat sources produce similar temperatures on the surface of a semi-infinite
body, while for larger values of~ a stationary heat source produces consider-
ably higher temperatures.

69
3.7.3 Application to Sliding Bodies

Let us consider the contact model shown in figure 3.22 where a single
spherical asperity of body 1 makes contact, under a normal load P, with the
surface of body 2 which is assumed to be sliding with a constant speed V.
The amount of heat generated over the contact area A will be given by

Q = J1PV (3.45)
J

where J1 is the coefficient of friction and J is the mechanical equivalent of

heat. During sliding, the contact surface of body 1 will be continuously
subjected to part of Q, say A.Q, while the remainder, (1 - A.)Q, will be flowing

Figure3.22

into body 2. For such a contact both bodies can be reasonably considered as
semi-infinite so that their surface temperatures may be obtained by applying
the results derived for stationary and moving heat sources. Thus, using
equations 3.40 and 3.42 the surface temperature of bodies 1 and 2 will be
given by

(3.46)

and
- 0.318(1 - A.) ( v ) 112 (3.47)
02- --
aJ aa.pc
Now, it remains to define the value of A.. Clearly, this will depend on the heat
transfer characteristics of the contacting bodies, but for simplicity we can

70
assume that A. is defined by the ratio of the thermal diffusivities of the two
bodies, that is
A. rx.tfPtCl
1 -A= rx2/P2C2
Therefore, we may easily use equations 3.46 and 3.47 to determine the surface
temperatures for the two bodies in contact.

REFERENCES

1. S. Timoshenko and J. N. Goodier. Theory of Elasticity. McGraw-Hill, New York,

(1951).
2. D. Tabor. The Hardness of Metals. Oxford University Press, (1951).
3. J. A. Greenwood and J. B. P. Williamson. Contact of nominally flat surfaces.
Proc. R. Soc., 195, A, (1966), 300.
4. K. A. Nuri and J. Halling. The normal approach between rough flat surfaces in
contact. Conference on Mechanics of Contact Effects, Kiev, (June, 1973).
5. J. F. Archard. The temperature of rubbing surfaces. Wear, 2, (1958-9), 438.

71
4
Friction Theories
4.1 INTRODUCTION

Friction is the resistance to motion which is experienced whenever one solid

body slides over another. The resistive force, which is parallel to the direction
of motion is called the 'friction force'. If the solid bodies are loaded together
and a tangential force is applied, then the value of the tangential force which
is required to initiate sliding is the 'static friction force'. The tangential force
required to maintain sliding is the 'kinetic (or dynamic) friction force'.
Kinetic friction is generally lower than static friction.

4.1.1 The Laws of Friction

It has been found experimentally that there are two basic 'laws' of friction,
which are obeyed over a wide range of conditions. There are, however,
a number of notable exceptions, examples of which will be included later. It
should be stressed at this point that the two laws of friction are empirical in
nature and, of course, no basic physical principles are violated in those cases
where the laws of friction are not obeyed.
The first law states that the friction is independent of the apparent area of
contact between the contacting bodies, and the second, that the friction force
is proportional to the normal load between the bodies. Thus a brick can be
slid as easily on its side as on its end and if the load between two sliding
bodies is doubled then the friction force is doubled. These laws are often
referred to as 'Amontons laws' after the French engineer Amontons who
presented them in 1699 1 • Coulomb (1785) introduced a third law, that the

72
kinetic friction is nearly independent of the speed of sliding, but this law has
a smaller range of applicability than the first two.

4.1.2 Coefficient of Friction

The second law of friction enables us to define a coefficient of friction. The

law states that the friction force F is proportional to the normal load W.
That is
F ex W
therefore
F = f.LW (4.1)

where ll is a constant known as the 'coefficient of friction'. It must be stressed

that f.L is a constant only for a given pair of sliding materials under a given set
of ambient conditions and varies for different materials and conditions. For
example, a hard steel surface rubbing against a similar surface under normal
atmospheric conditions would typically have a value of ll equal to about 0.6.
The same combination rubbing under very high vacuum conditions would
have a much higher value of fl. A graphite-on-graphite combination in
normal atmosphere has a value of ll equal to about 0.1 but this rises to over
0.5 if the atmosphere is very dry.
It is the object of this chapter to develop theories of friction which can
explain the variations in friction coefficient between different materials and
under different conditions. Now let us examine the first law a little more
closely and attempt to explain why the friction is independent of the apparent
area of contact.
In chapter 2 it has been shown that nearly all surfaces are rough on a micro-
scopic scale and true contact is obtained over a small fraction of the apparent
contact area. Furthermore the real area of contact is independent of the
apparent area of contact. Thus the first law of friction is explained since
friction is related to the real area of contact.

4.1.3 Surface Roughness and Real Area of Contact

When two bodies are rubbed together, some form of interaction takes place
at the contacting surfaces resulting in a resistance to relative motion. Most
friction theories assume that the resistive force per unit area of contact is a
constant. Thus
F = As
where F is the friction force
A is the real area of contact
sis the constant force per unit area, resisting relative motion, that is,
the specific friction force
73
If the assumption that s is a constant can be justified, then we can see the
importance of A. Let us now summarise some of the more important findings
of chapter 3 with respect to the real area of contact.
(a) For a single spherical contact or an array of similar spheres all at the
same height under loading conditions which produce elastic deformation
only

(b) When the contact is wholly plastic

A cc W
(c) For a surface whose asperity height distribution can be represented by
an exponential function
A cc W
whatever the mode of deformation, that is, either elastic or plastic.
(d) For a surface whose asperity height distribution is gaussian, we can
again write with sufficient accuracy
A cc W for all modes of deformation
(e) Many practical surfaces have an asperity height distribution that is
close to gaussian.
Results (a) and (b) above, combined with experimental findings confirming
the second law of friction, led in the past to the conclusion that asperity
contacts must be plastic. However, various physical arguments and experi-
mental results indicated that under certain conditions, for example, well
run-in surfaces, where it was found F cc W, elastic conditions must exist.
This apparent anomaly is now explained by (d) and (e) above. The analysis of
the deformation of model and real surfaces has now made possible the formu-
lation of reasonably realistic friction theories, and it is certain that any future
improvements on current theories must be based on such analysis.

4.2 FRICTION MEASUREMENT

It is not the object of this chapter to present a comprehensive survey of

friction measurement but a brief account of some of the available methods
will now be given.
Any apparatus for measuring friction must be capable of supplying relative
motion between two specimens, of applying a measurable normal load and of
measuring the tangential resistance to motion. There are a large number of
methods available and the final choice will depend largely on the exact
conditions of rubbing contact under investigation. For example, probably the
simplest arrangement is the tilting plane where a specimen is placed on a

74
flat surface which is gradually tilted until sliding starts, figure 4.1. The
coefficient of friction is then tan e. This method is obviously unsuitable in
those cases where a study of the variation of friction with continued rubbing
is required but its simplicity makes it attractive in many cases.

Figure 4.1 Measurement

of friction using tilting
plane, Jl. = tan (}

Where continuous friction measurement is required over a period of time

then an alternative approach must be used. Here one specimen, usually a
disc or a cylinder, is driven continuously, while a second specimen, nominally
stationary, is loaded against it. Commonly used combinations are crossed
cylinders, pin-on-cylinder or -disc, and disc-on-disc. The loading of the
stationary specimen can be by simple deadweight or, if the experimental
conditions demand it, by some more complicated method such as hydro-
static or magnetic loading. The measurement of the friction force is usually
accomplished by mounting the nominally stationary specimen so that a
very small tangential movement, proportional to the frictional force, occurs.
This small movement is measured and recorded. Two typical arrangements, a
crossed cylinder and a pin-on-disc are illustrated in figures 4.2 and 4.3.
In each case the specimen is mounted on leaf springs which allow a small
movement in the direction of the friction force. The movement can be

Load

Rotalong
cylinder

Figure 4.2 Simple crossed cylinder arrangement for the

measurement offriction and wear

75
 tload

Rotating
diSC

Figure 4.3a Arrangement

of pin-and-disc machine
Loading moror

Disc To leaf
spring for
friction
force
measurement

Triode 10n pump

Figure 4.3b Schematic diagram of ultra-high vacuum pin-and-

disc machine

calibrated to give the friction force and measured by a capacitance or in-

ductance method and continuously recorded.
The apparatus shown in figure 4.2 is a very simple and convenient method
of measuring friction. That shown in figure 4.3b has been used for ultra high
vacuum and controlled atmosphere friction tests. The motion is provided

76
by a magnetic drive through the walls of the chamber and the load applied
and friction are measured outside the chamber wall, each force being trans-
mitted through a bellows. ·
Just as is the case in wear testing, great care must be taken in ensuring
cleanliness during these tests, since small amounts of contamination can
significantly affect the measured friction, and this is why so much emphasis
is placed on controlled atmosphere tests.

4.3 POSSIBLE CAUSES OF FRICTION

In section 4.1.3, we stated that friction must be due to some interaction

between the opposing surfaces and that this results in resistance to relative
motion. As the surfaces move relative to one another, work is done by
the forces causing the motion, that is, there is an energy loss at ihe contacting
surfaces. In considering the possible causes of friction it is convenient to
consider separately the surface interaction and the mechanism of the energy
loss.

4.3.1 Surface Interactions

When two surfaces are loaded together they can adhere over some part of
the contact and this adhesion is therefore one form of surface interaction
causing friction.
If no adhesion takes place then the only alternative interaction which
results in a resistance to motion is one in which material must be deformed
and displaced to accommodate the relative motion. We need consider only
two interactions of this type. The first is asperity interlocking. Considering the
situation illustrated in figure 4.4, it is obvious that relative motion cannot
Otrect1on of mot1on
A

Figure 4.4 Asperity interlocking-motion cannot

take place without deformation of the asperities

take place between surfaces A and B without displacement of the material of

the asperities.
A second example of the displacement type of interaction is illustrated in
figure 4.5. Here a hard sphere A is loaded against a relatively soft flat surface,
B. In order for relative motion to take place some of the material B must be
displaced. Although the surfaces of both sphere A and flat B will be rough on

77
 Figure 4.5 Macro-displacement-a
hard sphere A, loaded against a softer
surface B, causes displacement of
material B during motion

a microscopic scale the material displacement at the individual asperities

will, in this case, be small compared to the 'macro-displacement'. Thus we
have only two types of interaction, Adhesion and Material Displacement,
although we will find it convenient to think of the material displacement as
either, Asperity Interlocking or Macro-Displacement.

4.3.2 Types of Energy Loss

There are only three mechanisms which can cause appreciable loss of energy
at the interacting surfaces. As relative motion takes place, material must be
deformed. The deformation can be either elastic or plastic; additionally the
material may be fractured. Plastic deformation will always be accompanied
by a loss in energy and it is this energy loss which accounts for the major
part of the friction of metals under most practical circumstances. Fracture
must occur when the surface interactions are adhesive and can also take place
due to relative motion of interlocking asperities. The formation of wear
debris is of course evidence that fracture has taken place. However, the energy
losses associated with fracture will, in most cases of sliding metals, be small
in comparison with those due to plastic deformation. One reason for this is
given in chapter 6 where it is shown that a wear particle is not formed at
each asperity contact but that for most metals in normal atmospheric con-
ditions, an asperity makes more than 1000 contacts before the formation of a
wear particle.
Although energy is required to deform a metal elastically, most of this
energy is recoverable and elastic energy losses are negligible compared with
the energy losses associated with plastic deformation. However, some rubbers
exhibit large irreversible energy losses due to their elastic deformation
(elastic hysteresis) and in certain cases this is the major source of friction,
(see section 4.8).
Summarising the above considerations shows that there are two sources
of surface interaction, that is, adhesion and material displacement, and these
can cause energy losses due to both elastic and plastic deformation and to
fracture.
We will now consider various proposed mechanisms of friction in the light
of the above considerations.

78
4.4 THE ADHESION THEORY OF FRICTION

4.4.1 Simple Theory

This theory due to Bowden and Tabor 2 , uses as a starting point the fact that
when metal surfaces are loaded against each other they make contact only
at the tips of the asperities. Because the real contact area is small the pressure
over the contacting asperities is assumed high enough to cause them to
deform plastically. This plastic flow of the contacts causes an increase in the
area of contact until the real area of contact is just sufficient to support
the load, figure 4.6. Under these conditions, for an ideal elastic-plastic
material
Ap0 = W

where A is the real area of contact, p o is the yield pressure of the metal W is
the normal load.

Pressure Po

lllll!!ll
Area A

Figure 4.6 A single asperity

contact-the asperities yield
plastically until the area of
contact has grown sufficiently
to support the load

Over the regions of intimate metal-to-metal contact, Bowden and Tabor

state that strong adhesion takes place, and that the junctions 'cold weld'.
If s is the force per unit area of contact necessary to shear the junctions,
that is, s is the shear stress necessary to cause plastic flow and final fracture
and F is the friction force, then
F =As+ Pe
where Pe is a term introduced by Bowden and Tabor to take account of the
force required to 'plough' hard asperities through a softer surface, that is, a
material displacement interaction causing plastic deformation. Bowden and
Tabor state that for most situations involving unlubricated metals Pe is small
compared with As and may be neglected. (The ploughing term is discussed
in section 4.7.) Ignoring the ploughing term we can write
Ws
F =As=-
Po
79
and
F s
f.1.=-=-
w Po
Thus this simple theory provides an explanation of the two laws of friction,
that is, that the friction is independent of the apparent area of contact and
the friction force is proportional to the load.
In the above analysis we have considered an ideal elastic-plastic material,
and have ignored the effects of work-hardening. Therefore it is reasonable to
takes equal to S0 the critical shear stress, and both Po and S0 must refer to the
softer of the two metals.
Now

(4.2)

This ratio sofPo is fairly constant for most metals and the above analysis
gives an indication of why the friction coefficient of a large range of metals
varies little, while their mechanical properties, for example, hardness, vary
by orders of magnitude. In the case of two hard metals in rubbing contact
Po is high, A is low and S 0 is high. For soft metals, Po is low, as is S 0 , but A is
large.
One way of obtaining lower coefficients of friction is to deposit a thin layer
of soft metal onto a hard metal substrate. Now, the load carrying capacity is
really due to the substrate and Po is the yield pressure for the substrate.
However, shearing takes place in the soft metal layer and the critical shear
stress of the soft metal is the value we must use, and therefore

f.1. ~ s0 (soft) = low

Po (hard)

4.4.2 Discussion of Simple Theory

There is no doubt that junction welding can take place during the rubbing of
metals. For cleaned metal surfaces in high vacuum very high adhesion and
friction coefficients have been recorded. For metals rubbed in normal atmos-
pheric conditions, adhesion and transfer of metal fragments have been
demonstrated using radioactive tracer techniques. However, the simple
adhesion theory has been criticised for a number of reasons and it can be
shown to be inadequate by a comparison of the absolute values of the friction
coefficient predicted by the simple theory and those found exprim~ntaly.
For most metals, S0 is about one-fifth of Po and therefore the simple adhesion
theory predicts that f.1. ~ 0.2.
Many metal combinations in air give a friction coefficient higher than 0.5,
and metals in high vacuum give much higher values of f.l.· This led Bowden

80
and Tabor to re-examine some of the assumptions in the simple theory, and to
present a modified and more realistic description of friction in terms of
adhesion 2 .

4.5 MODIFIED ADHESION THEORY

The fact that very high values of friction are obtained for metals under high
vacuum conditions, where adhesion is unimpeded by oxide films or other
contaminants, indicates that the real contact area must be considerably
larger than is indicated by the simple theory. In the simple theory it was
assumed that A was defined by the yield pressure of the softer metal Po
and the normal load W. This is approximately true for static contact but in the
case of friction, where a tangential force is also applied, yielding must take
place as a result of the combined normal and shear stresses. To illustrate this
let us look at the simplified two-dimensional stress system illustrated in
figure 4.7a and assume that yielding occurs when the maximum shear stress
attains a critical value 3 .
Shear
stress

llllllll
Normal
s stress

(al

(b)

Figure 4.7 The Mohr's circle construction for finding the

maximum shear stress for an idealised two-dimensional
junction under normal and tangential stress

We can find the maximum shear stress in the system by using the Mohr's
circle 4 construction, figure 4. 7b. The maximum shear stress is the radius R
of the circle therefore

When R reaches the critical resolved shear stress, yielding takes place. From
this it is easy to see that yielding is dependent on the action of the combined
stresses and not on p alone.

81
 We will now examine how consideration of the combined stresses affects
the value of the real area of contact in an asperity junction.
Consider first a single asperity contact under a normal load W. The area
of contact will be A where W/A =Po· If a tangential force is now gradually
applied up to a value F, further plastic flow will take place. This flow causes
an increase in the contact area, that is, junction growth is brought about by
the superposition of the shear stress on the normal stress. The normal and
shear stresses caused by the normal and shear forces must decrease as the
area over which the forces act increases, and junction growth continues until
the combined stresses obey a relationship of the form previously given for the
two-dimensional system. The exact solution for the three-dimensional case
is not known but we will assume that it is of the form
p2 + rxs2 = k2
where rx and k are constants yet to be determined, that is

(4.3)

where A is the area of contact of the junction. Now, if s is zero, the pressure
over the junction must be Po and therefore
k2 = p;
that is
(4.4)
If F increases to very large values then junction growth continues until
W/A is small in comparison with F/A in which case we can write

In this case s must be approximately equal to S0 , the critical shear stress.

Therefore

or
p;
(1,~2
so
Now Po ~ 5s0 therefore rx ~ 25. However, experiment indicates that rx should
have a value somewhat lower than 25 and Bowden and Tabor assume that
rx = 9. (This also implies that Po= 3s0 .) We will see later that the exact value
of rx does not greatly affect the amount of junction growth taking place in
many practically important cases.
From equations 4.3 and 4.4 we have

82
W !Po is the area of contact derived from the simple theory in which only the
effect of normal load is considered, and the additional term a(F/p 0 ) 2 repre-
sents the increase caused by the shear or friction force.
It can be seen from the above discussion that for clean metal surfaces
(that is, metals in high vacuum), large-scale junction growth is possible,
resulting in very high friction coefficients. This has been confirmed experi-
mentally5. In normal atmospheres metals are covered by an oxide or other
contaminant film. We will now see how the above theory can be applied to
such cases.

4.5.1 Adhesion Theory of Metals with Contaminant Films 2

Consider an asperity junction under normal load W and a gradually increas-

ing shear force F. We assume that at the junction there is a thin contaminating
film with critical shear stress sf. We also assumes f = cs 0 where S 0 is the critical
shear stress for the metal and cis less than unity. While F and A, have values
such that F I A < sf, then junction growth will proceed as described pre-
viously for uncontaminated metals. However, when F I A = sf then the con-
taminating film will shear, junction growth will end, and gross sliding will
occur.
Thus the condition for the start of gross sliding is
p2 + IY.SJ = p;
But it has already been shown that
p; =as;
Therefore

or

Therefore
sf= c
p [a(l _ c2)Jli2
The coefficient of friction 11 = (F /W) = (s fA/pA). Therefore
c
(4.5)
11 = [a(l _ c2)Jli2

As c tends to 1 then 11 tends to infinity in agreement with results obtained for

uncontaminated metals. We can plot 11 against c for various values of 1:1.,
figure 4.8, and we see that the value of 11 drops rapidly as c reduces from unity.

83
 0

Figure 4.8 The variation of Jl. with c for different values

of IX. It can be seen that except at large values of c, the
exact value of IX is not of major importance

Thus a small amount of weakening at the interface produces a drastic reduc-

tion in Jl..
When c is small then equation 4.5 can be written
c
Jl. = (1X)l/2

but

Therefore
cs0 s1
)1.=-=-
Po Po
84
that is
critical shear stress of the interface
~ =-~
yield pressure of the bulk metal
which is essentially the same result as obtained by the simple theory (com-
pare discussion of soft metal films, or hard substrates). This is understandable
since when the interface is weak in shear, appreciable junction growth does
not occur and the real contact area depends only on the normal load and the
yield pressure po.
This then is a more realistic theory. Although it is still based on a simple
model and contains a number of assumptions, it is able to explain a remark-
able range of friction phenomena.
However, it is no longer exclusively an adhesion theory. Let us examine
what are the main points of the theory
(a) The real contact area is defined by plastic deformation.
(b) The two rubbing surfaces are separated by a film of shear strength
which can vary from low values up to the bulk shear stress of the sub-
strate material.
(c) The friction force is the force required to shear the separating film.
At one extreme, where metal-to-metal contact and junction welding take
place, we can consider that either the separating film is of zero thickness,
or that it is composed of the softer of the two metals. Under atmospheric
conditions of dry sliding where the metals are covered by oxide films it is
only at those points where the oxide film is broken that metal-to-metal
contact and welding take place. In this case the effective shear strength of the
interface will lie between the shear strength of the softer metal and that of
the oxide film; its exact value depending on the relative amount of metal to
metal and metal to oxide contact.
We will now discuss a number of criticisms of the theory that have been
advanced.
(i) If two hard metals are rubbed together under atmospheric conditions,
in general no adhesive component normal to the surfaces can be
detected when the normal force has been released. Bowden and Tabor
answer this by pointing out that the adhesion under rubbing conditions
occurs while the normal force is exerted. To measure adhesion in the
normal direction the normal load must first be released and elastic
recovery will break many of the bonds during this process.
(ii) The adhesion theory of friction can explain the transfer of material
from one rubbing surface to the other but offers no explanation of the
formation of loose wear debris. Experimental work has shown that
transfer occurs during single traversals but subsequent traversals on
the same track produce loose wear particles. This suggests a possible
change of mechanism during the rubbing process. It may be significant

85
 that much of the experimental work carried out by Bowden and Tabor,
giving results consistent with the adhesion theory, consisted of single
rather than multiple pass experiments.
(iii) The adhesion theory is based on plastic deformation of asperities.
On continued rubbing under conditions of low wear, work hardening
will take place and a proportion of asperity contacts will now deform
elasticially. However, although the deformation of the asperities is now
largely elastic, the area of contact is still related to the previous plastic
deformation. This can be understood by considering an individual as-
perity contact under an initial normal load W. The deformation curve
is as shown in figure 4.9. If the load is now released and reapplied up to
a maximum value of W, the deformation is elastic but the total deforma-
tion or strain is exactly the same as in the plastic deformation case.

We will return to these criticisms in section 4.9 which contains a general

discussion of friction theories.

I
I

I
I
I
I
I
I
I
I
I
I
I
I
I 8

Figure 4.9 Typical stress-

strain curve of metals. If the
stress is relaxed at A and then
reapplied the deformation is
elastic along BA, but the resul-
tant strain is equal to that
caused by the initial plastic
deformation

86
4.6 PLASTIC INTERACTION OF SURFACE ASPERITIES

In the Bowden and Tabor theory, the normal and yield stress on a single
asperity were assumed to be representative of the stresses of all asperities.
No consideration was given to the possibility that at a single asperity,
s and p could vary with time over large ranges so that s/p could also vary.
This evolves as a feature of a theory based on the plastic interaction of
asperities. This type of theory was first introduced by Green 6 and has been
extended by Edwards and Halling 7 whose treatment is followed here. The
basis of this analysis is that in the sliding of macroscopically flat surfaces the
motion is parallel to the surface.
Consider two wedge-shaped asperities of equal angle interacting as shown
in figure 4.10. Assuming that motion is in the direction shown and that

Figure 4. 10 The idealised wedge-shaped as-

perities studied in the 'plastic interaction theory'

deformation is plastic it is possible to calculate the instantaneous shearing

force F 1 and normal force W1 over the complete horizontal displacement l
from the first contact until the asperities separate, both for the case where the
asperities are in intimate contact and adhesion is possible and for the case
where the asperities are separated by a film of shear strength cs 0 where c and
s0 have the same meanings as in the Bowden and Tabor theory. The develop-
ment of the expressions for F 1 and W 1 is beyond the scope of this book
but the form of the plot of normal and shear forces against displacement is
shown in figure 4.11 for junction angles of 10° and for various values of c.
The coefficient of friction is the sum of instantaneous shearing forces F 1
for all contacting asperities divided by the sum of instantaneous normal
forces W 1 for all contacting asperities. Alternatively J1 could be regarded as
the mean FjW value for a single pair of interacting asperities, which are
assumed to be typical of all the interacting asperities. Thus the coefficient of
friction can be calculated from figure 4.11 and plots of J1 against c, so obtained
are shown in figure 4.12.
It is interesting to note that if the results are extrapolated the curve cor-
responding to junction angle (}equal to zero, corresponds identically with the

87