Stroh1957 PDF

This article was downloaded by: [University of Pennsylvania]
On: 12 June 2013, At: 07:14

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Advances in Physics
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/tadp20
A theory of the fracture of metals

a
A.N. Stroh
a
Department of Physics, University of Sheffield
Published online: 02 Jun 2006.
To cite this article: A.N. Stroh (1957): A theory of the fracture of metals, Advances in Physics,
6:24, 418-465
To link to this article: http://dx.doi.org/10.1080/00018735700101406
PLEASE SCROLL DOWN FOR ARTICLE
Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-

conditions
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any
representation that the contents will be complete or accurate or up to date. The
accuracy of any instructions, formulae, and drug doses should be independently
verified with primary sources. The publisher shall not be liable for any loss, actions,
claims, proceedings, demand, or costs or damages whatsoever or howsoever caused
arising directly or indirectly in connection with or arising out of the use of this material.
[ ]
A Theory of the Fracture of Metals
By A. N. STROH
Department of Physics, University of Sheffield
CONTENTS
§ l. INTRODUCTION.
§ 2. Pt~ELIMINARY CONSIDERATIOI~S.
2.1. T h e s t r e n g t h o f t h e ideal lattice.
2.2. Griffith's t h e o r y .
Downloaded by [University of Pennsylvania] at 07:14 12 June 2013
§ 3, THE BRITTLE FRACTURE OF METALS.

3.1. Piled u p g r o u p s o f dislocations.
3.2. T h e p r o d u c t i o n o f a crack.
3.3. T h e g r o w t h o f t h e crack.
3.4. C o m p a r i s o n w i t h e x p e r i m e n t a l results.
3.5. H y d r o g e n e m b r i t t I e m c n t .
3.6. Slip n e a r a piled u p group.
§ 4. T ~ E LOWER YIELD POINT.

4.1. T h e s t r e s s n e e d e d to m o v e a dislocation.
§ 5. T ~ E BRITTLE--DuCTILE TRANSITIOn;.
5.1. T h e p r o b a b i l i t y o f brittle f r a c t u r e .
5.2. E f f e c t o f s t r a i n rate.
5.3. E f f e c t o f g r a i n size.
5.4. The activation energy.
5.5. N o t c h brittleness.
§ 6. THE PROPAGATION OF'CRACKS.

6.1. Grifllth's t h e o r y applied to m e t a l s .
6.2. T h e v e l o c i t y of a crack.
6.3. The propagation transition.
6.4. T h e effect o f t r a n s v e r s e stress.
§ 7. DUCTILE FRACTURE.
ACKNOWLEDGMENTS.
APPENDIX A : T h e energy of a crack.

B : T h e initial g r o w t h of a crack.
REFERENCES
§ 1. ~[I~TRODUCTION
THE topics which will be considered here can b r o a d l y b e described as the
transcrystalline fracture of polycrystalline materials; an attempt will
be made to show how tile principal features observed follow from a
simple model. Single crystalls will not be considered in detail ; though
they have been the subject of some recent experimental studies, e.g.
Deruytt6re and Greenough (1955), Allen et al. (1955), it has not yet
proved possible to give a satisfactory account of their properties. Other
forms of fracture, for example, fatigue, fracture under creep conditions,
will also be excluded, even though the), may be of great technical
Theory of the Fracture of Metals 419
importance and the subject of much experimental work, for it seems likely
t h a t quite different mechanisms operate in such circumstances. A recent
paper by Mott (1956) includes a discussion of fatigue in addition to
those topics now considered while a review b y Perch (1954) deals with
the whole problem of fracture and includes a more complete account of
the experimental results.
I t should be mentioned t h a t o~her theories of fracture, such a s t h o s e
of Fisher (1955) and KochendSrfcr (1954) which are quite different from
t h a t presented now, .have been proposed; such treatments will not, be
discussed here as t h e y have yet to be developed in sufficient detail to
permit of a quantitative comparison with experimental results.
Two main types of fracture fall within out' consideration. First
the metal m a y be brittle and break with only little plastic deformation ;
after fracture, the surfaces show bright crystalline facets, on which
cleavage has occurred. Here definite crystalline planes are involved,
for example, body-centred cubic iron cleaves on [001} planes, and zinc
(hexagonal) on the basal plane. No metal with a face-centred cubic
structure is known to break in this way, and other metals do so only s.t
low temperatures. Such fractures spread with great rapidity, and in
practice m a y produce spectacular failures. They attracted particular
notice during the last war in connection with welded ships; some 200
Liberty ships developed serious cracks, while about a dozen broke com-
pletely in two.
The other possibility is t h a t the metal is ductile and plastic deformation
precedes fracture. In a tensile test the specimen first necks and later
a crack develops in the centre of the neck ; after the specimen has grown
to sufficient size, the mechanism changes and the final separation is
by shear at. an angle of 45 ° to the stress axis ; thus the familiar ' cup and
c o n e ' is produced. In contrast to the brittle case, the fracture surface
now has a m a t t or fibrous appearance. Ductile cracks spread only
slowly and can be stopped by a decrease in the load.
For the very reason t h a t complications due to plastic flow are absent,
brittle fracture is the simpler to interpret, and we shall find it possible
to give a much more complete account of it than we can of ductile fracture.
The temperature at which the transition from ductile to brittle behaviour
occurs is a m a t t e r of some importance for which our model will be able
to account. Finally it is found that, with a material such as iron having
a sharp yield point, the lower yield point, t h a t is the stress at which a
band of deformation (a Liiders band) spreads through the specimen,
has a number of features in common with the brittle s t r e n g t h ; it is
therefore convenient to include a discussion of this topic.
§ 2. PRELIMINARY CONSIDERATIONS
2.1. The Strength of the Ideal Lattice
I t has been known for a long time t h a t a substance having a perfect
crystalline structure should exhibit a strength very much greater than is
420 A . N . Stroh on a
observed in practice. The ideal strength is easily estimated by a simple

argument due to Orowan (1949, 1955). The force between two adjacent
atomic planes must vary with distance in the manner shown in fig. 1;
Fig. 1
Distance
The interatomic force as a function of distance.
the force is zero when the planes have their natural spacing b, rises to a
maximum as the separation increases and then falls again at greater
distances. The area below the curve represents the work done per unit
cross sectional area in fracture and so must be equal to 2r where y is
the energy per unit area of the new surfaces formed. To estimate the
order of magnitude of the magnitude of the maximum stress a,n let
us assume t h a t ttooke's law, which must be trhe for small displacements,
holds until the stress reaches the value a m ; then the elastic energy Stored
between the two planes is bam~/2E per unit area where E is Young's'
modulus. Further, assume t h a t the area to the left of the dotted line in
fig. 1 is about half the total area, and so equal to y. Then a m is given by
bam~/2E:y,
or
~,~=v~(2~r/b) . . . . . . . . . (1)
Typical values of the constants for a metal are E : 1 0 1 2 dynes/cms,

y : 1 0 3 e r g s / c m 2 and b~2)<10-Scm, giving a m ~ 3 X 1 0 n dynes/cm2
2000 tons/in, s, a value from 20 to 1000 times the observed strength.
Though our calculation gives only a rough estimate it is difficult to
believe t h a t it can be wrong by so large a factor. Indeed the large
value of the theoretical strength of the ideal lattice has been confirmed
by Zwicky (1923) and by Born and F u r t h (1940), for ionic crystals, by
detailed calculations of the interatomic forces.
2.2. Gri~th' s Theory

To account for the discrepancy between the observed and calculated
values of the strength, Griffith (1920) assumed t h a t a real material
contains a number of small cracks which can act as stress concentrators
raising the stress at their tips to the ideal strength a m ; in this way the
crack can extend and produce fracture of the material.
Consider the two dimensional (plane strain) problem of a crack of
length c extending through a thick plate, and subjected to a tensile
stress a normal to the crack. From Inglis's calculation (1913) of the
stress distribution, the elastic energy is decreased due to the presence
of the crack by an amount 7r(1--v)a2c2/8G, where G is the rigidity modulus
and u Poisson's ratio. I n addition the crack has surface energy 2c~,
so t h a t the total energy of the crack is
W h --~r(1 --v)a~c~/8G-~-2c~,. . . . . . (2)
I f the crack is to extend, W must decrease as c increases ; the equilibrium
length of the crack is given by dW/dc=O, and from eqm (2) equilibrium
will occur when
The crack can extend if this stress is exceeded, and so eqn. (3) gives the
fracture stress of the material. There is clearly always a crack length c
which will lead to the observed strength.
Sack (1946) has extended the calculation to a circular (penny-shaped)
crack ; the only change is that the critical stress is greater by a factor
½7r(= 1-57) t h a n is given by eqn. (3). This result is interesting because
most calculations on the behaviour of cracks have, for simplicity, dealt
with a two-dimensional system only; Sack's work suggests t h a t no very
great error will be made thereby.
An alternative approach is to use the condition that, for the crack to
extend, the stress at its tip must have the value %~ (given by eqn. (1)) ; over
an area of atomic dimensions. By considering the stress magnification
produced by the crack, Orowan (1949, 1955) has shown that this
leads, within the accuracy of the calculation, to the same critical stress
as before (eqn. (3)). There is thus general agreement as to the stress
needed to propagate a crack through a brittle body.
Though the Griffith theory may possibly give a satisfactory account
of the fracture of a completely brittle solid, such as. glass on which
Griffith's own experiments were done, it seems unlikely t h a t pre-existing
cracks are important in metals. To account for the observed strength in
this way the cracks would often have to be unreasonably large--several
miUimetres in length, for example, for a weak single crystal of zinc.
Such cracks could hardly always be present and escape detection. Never-
theless Griffith's ideas occupy a fundamental position in the theory of
fracture and have greatly influenced much of the subsequent work.
In particular, an application of Griffith's theory to the propagation
of brittle cracks in metals will be discussed in § 6.1.
§ 3. THE BRITTLE FRACTURE OF ~ETALS
3.1. Piled Up Group~ of Dislocations

The presence of dislocations in metal crystals is now well established,
and considerable progress has been made in elucidating the part they
play in plastic deformation and work hardening ; full accounts of their
properties are given in the books by Cottrell (1953), Read (1953) and
Friedel (1956). I t m a y be expected that dislocations play an equal!y
important role in fracture. Here the edge dislocation, at which a h a l f
plane of atoms terminates, is the more important type of dislocation,
since it is surrounded by regions of tensile (and compressive) stresses.
I f a large number of such dislocations arrange themselves with their
stress fields reinforcing one another, we may expect that they will jointly
be able to produce a stress great enough to cause fracture.
Suppose t h a t a number of dislocations are generated in the same slip
plane, and t h a t these spread out until they meet a grain boundary. On
reaching the boundary they will be unable to advance further, as in
general the slip planes of the next grain will not be suitably orientated.
The dislocations will, therefore, be piled up on the grain boundary by
the applied stress, and will take up the arrangement shown in fig. 2;
Fig. 2
J
<rs
_l_ _l_ _i_ _a_ I
Piled up group of disloca.tions.
those at the head of the array will be crow<led close together by those
behind them, the spacing gradually increasing towards the rear. Such
piled u 1) groups, which were first considered by Cottrell (1949) and
Eshelby et aI. (1951), have been observed by lacquer (1954), and fig. 3,
PI. 14, which is reproduced from one of his photographs, shows a row
of etch ]>its each formed on one dislocation of a piled up group ; the
whole arrangement agrees so closely with that predicted (a detailed
comparison has been made by Bilby and Entwistle 1956) t h a t there
can be little doubt that this interl>retatiou is correct.
Theory o]' th~ Fracture of Metals 423
N e a r a piled up g r o u p t h e stresses will be large : their n a t u r e is m o s t

easily seen f r o m Zener's r e m a r k (1948) t h a t such a b l o e k e d s l i p line will
resemble a freely slipping crack u n d e r a af~ear stress1; for the fact t h a t
the slip line contains a large n u m b e r of dislocations implies t h a t the
materiM on one side of it can readily slip relative to t h a t on the other.
W e have, therefore, f o u n d an a r r a n g e m e n t of dislocations which is
v e r y similar to one o f Griffith's era~cks, and which m a y be e x p e c t e d to
serve equally well in p r o d u c i n g fracture.
3.2. The Prod~ctio.n of n Crac]c

Let us now consider in w h a t circumstances t h e stresses r o u n d a piled-up
g r o u p of dislocations will be sufficient to generate a crack. T h e v a l u e

o f the stresses m a y be f o u n d either b y using t h e a n a l o g y m e n t i o n e d
a b o v e b e t w e e n a slip line a n d a crack, or b y directly s u m m i n g t h e stresses
due to the individuM dislocations (Stroh 1954). T h e result is t h a t a t a
distance r a n d f r o m the h e a d of the piled u p g r o u p t h e n o r m a l stress
across the p l a n e OP (fig. 2) is
cs=as(L/r)l/2f(O), ( r < L) . . . . . . . (4)
where the dislocations o c c u p y a length L in the slip plane> a s is the resolved
shear stress on the slip plane, and f(0) is a factor d e p e n d i n g on t h e orienta-
tion of OP. I n a p o l y c r y s t a l l i n e m a t e r i a l there will be grains in all
orientations a n d those m o s t f a v o u r ~ b l y s i t u a t e d will b r e a k first ; hence
we m a y e x p e c t f r a c t u r e to s t a r t on a cleavage p l a n e for which f(O) is
n e a r its m a x i m u m value ; in the case o f an elastically isotropie materiM
this will be w h e n 0 ~ 7 0 . 5 °. T h e n if a crack s t a r t s at O a n d e x t e n d s a
distance c along OP the a v e r a g e value of the stress (4) on it m a y be
written
cr=y.(g/e)l/e%, . . . . . . . . (5)
where ~. is a numerieM c o n s t a n t . I n s e r t i n g this stress in eqn. (2) we
13nd t h a t the p r o d u c t i o n of a crack will result in a decrease of the e n e r g y if
%2--16y(@r(1--,)~2L . . . . . . . . (6)
a n d this is t h e n the condition for initiating a crack. I n A p p e n d i x A
the c o n s t a n t ~ is e v a l u a t e d a n d f r o m t h e results o b t a i n e d t h e r e eqn. (6)
m a y be w r i t t e n
%~=37rya/8(1--~,)L . . . . . . . (7)
E q u a t i o n s (6) a n d (7) do n o t c o n t a i n t h e length c o f the crack, a n d so if
t h e y are satisfied for one length of the crack, t h e y will be satisfied for all
lengths ; thus the crack can grow to a finite size w i t h a decrease of e n e r g y
a t e v e r y stage. This g r o w t h can continue until t h e length of the c r a c k
is c o m p a r a b l e to the length of t h e slip line ; t h e n since the stress is no
$ A crack and a slip lille will, of course, behave quite differently under a
normal stress, since such a stress can act across the slip plane just as if the
dislocations were absent: in c~)mparing a slip line with a crack we must therefore
disregard any normal component ~f the stress.
longer given by eqn. (5) so far from the piled up group, the ideas we have
been using cease to apply ; the further growth of the crack needs separate
consideration, and will be discussed in § 3.3.
I t is of interest to consider how m a n y dislocations are needed to initiate
a c r a c k . Eshelby et al. (1951) have shown t h a t a piled up group of n
dislocations will occupy a length of slip-line given by
L==Gbn/Tr(1--v)% . . . . . . . . (8)
and on substituting this value into eqn. (7), the condition for initiating
a crack becomes simply
n%b-~Tr2~, . . . . . . . . . (9)
As an indication of the magnitude of n suppose fracture occurs under a

tensile stress of say 20 tons/in. 2, t h a t is a shear stress of 10 tons/in. ~
(1.6x 109dynes/cm s) ; then with b ~ 2 × l0 -s cm and ?----103ergs/cm z,
eqn. (9) indicates t h a t about 100 dislocations will be required. This
number of dislocations escaping from a free surface would produce a
step of height 200 4.
)~[ost of the observations which have been made on slip line heights
relate to face-centred cubic metals ; here there is a good deal of evidence
showing t h a t steps corresponding to 30 dislocations are the rule (cf. for
example, Seeger et al. 1957). However these results are not very relevant
to the present discussion, as brittle fracture does not occur in face-
centred cubic metals. Much less work has been done on other metals.
Figure 4, P1. 14, due to ~'utting and Brandon (private communication),
shows t h a t in iron coarse slip lines m a y occur ; a rough estimate of the
height of the slip line shown is about 800 ~, corresponding to about
300 dislocations, but this is subjected to considerable uncertainties as
the direction of shadowing is unknown. However, it would appear
that in iron, at least, piled up groups of the size needed to initiate a crack
will occur.
Since dislocations are used to produce the crack, it is necessary t h a t
some plastic flow should occur before fracture, but the amount needed
is quite small. I f a grain of diameter d contains a single slip line, it
will undergo a plastic shear nb/d. Since the slip distance L will be of
order d, eqn. (8) gives t h a t in order of magnitude this plastic strain is
nb/L ~ % / G ,
that is, of the same order as the elastic strain. But as it is by no means
necessary t h a t every grain should slip, the average plastic strain in a
polycrystalline material m a y be very much smaller. Thus it is possible,
on the model, for a crack to form after only a very small plastic strain.
In fact, metal fractures which are classed as brittle are frequently
preceded by plastic strains much larger than the minimum estimated
here to be necessary. Thus Crussard et al. (1956) find t h a t when notched
specimens of steel break in a brittle manner, a plastic strain near the
notch equivalent to an elongation of at least 3 % occurs. Even when the
ductility is very slight, there is some evidence to show t h a t fracture is

always preceded by a small amount of plastic flow, as the model requires.
Low (1954) has found t h a t fracture will not occur until the yield stress
has been exceeded. Earlier work has been reviewed b y Zener (1948)
who emphasizes that a large variety of experiments show t h a t a n y process
which increases the yield stress will also increase the fracture stress,
so t h a t initial yielding always occurs before fracture.
3.3. The Growth of the Crack

There is an alternative and possibly more vivid way of picturing the
origin of a crack, due to Zener (1948) and Stroh (1955a). An edge
dislocation represents a half plane of atoms wedged between two complete

planes. Now the leading dislocations of a piled-up group are forced
very close together; two dislocations will be as close as possible when
the extra half plane of atoms associated with each arc adjacent, and in
this case there is a wedge two atomic spacings thick between the nearest
complete planes of atoms. By forcing more dislocations together a
wider wedge can be produced until ultimately the neighbouring planes
are split apart, so t h a t a crack is formed (fig. 5). Once the crack has
started the stresses round it will be relieved and so make it easier to
bring in more dislocations, which in turn will widen the wedge and cause
the crack to spread further.
Fig. 5
• • • • • • $ • •
• • • • • • • • •
• • • • • •
A dislocation with Burgers vector 3b, showing an incipient crack.

On this picture the stress needed to start a crack is the stress which
will bring the two leading dislocations of the piled up group within one
atomic spacing of each other. Now according to Eshelby et al. (1951),
the distance between the two leading dislocations is 3.67 Gb/4Tr(1--v)%n,
and equating this to b, the condition becomes
nb%:3.67 Gb/4rr(1--v).
This agrees within the accuracy of the calculation with our previous
condition (eqn. (8)) if we make use of the empirical result t h a t Gb/7 "~ 8 for
most metals.
I f this wedging action of the dislocations were the only factor causing
it to spread, the crack would come to rest after attaining some definite
length depending on the thickness of the wedge, t h a t is on the number
of dislocations which have entered the crack. However, the applied
stress will in general have a normal component acting across the crack
and this will also help open up the crack. The normal stress will be
more important the greater the length of the crack, and once the crack
has attained the Griffith length (i.e. the length given by equ. (3)) the
normal stress alone will be sufficient to make it spread through the
material. (It is assumed here t h a t the metal is ' brittle ', i.e. that the
growth of the crack is not damped out by plastic flow.) Thus when the
crack is small its growth is controlled by the slip line, when large, by the
tensile stress. Then' depending on the nature of the stress system we
may have either of the alternatives : (i) the crack is able to grow ibr all
values of the length and so produces fracture ; or (ii) there is some inter-
mediate length at which neither of the mechanisms considered is sufficient
to e x t e n d the crack, so t h a t it comes to rest in the material.
The problem of a specimen tested under simple tension is considered in
Appendix B where it is found that case (i) should apply; hence eqns.
(7) and (8) will in this situation give not only the condition for initiating
a crack but also that for producing fracture.
3.4. C o m p a r i s o n with E x p e r i m e n t a l Results

A slight modification of eqn. (7) permits of its being compared directly
with the experimental values of the fracture strength. Tile largest
piled up groups can be obtained when the slip line extends right across
a grain; in, this case the calculations of Eshelby et al. show that we
must take L--~~d. Also the maximum sheer stress % is half the tensile
stress ~. Inserting these values in eqn. (7) we obtain
( ~ j : K d -1/~ . . . . . . . . . (10)
where
K={67rTG/(1--v)}l/e . . . . . . . (11)
Equation (10) giving the relation between fracture stress and grain
size was first deduced by Orowan (1933) from (~riffith's theory, by
assuming that the initial crack length was equal to the grain diameter.
The relation found experimentally is slightly more complicated than that
given by eqn. (10), and is
af=Kd-1/2+% . . . . . . . . (12)
where a0 is characteristic of the material and temperature but independent
of the applied stress. This modified equation gives satisfactorily the
strength of a number of metals (figs. 6-9). A relation of the form (12)
Theory of the Fracture of Metalz 427
would have been obtained if in eqn. (fi) we had replaced a, b y %--%0, where
a ~ is a constant shear stress ; t h a t is if we h a d assumed t h a t only p a r t of
the resolved shear stress was effective in piling up the dislocations. The
stress a~0, which will be discussed f u r t h e r in § 4.1, m a y be regarded a~ a
frictional force opposing the m o v e m e n t of t h e dislocations.
Fig. 6
IE
tOO
..% ~ 0 ~, ~ ~0
T
/ IO
]
i
• f
c / ,oS
~0
o. gO / / .
~0
2 4 6 il I0
t "~ M~,?"
The relationship between fracture strength and grain size for a low carbon
steel at (a) --196°c (brittle), (b) 18% (ductile), (0) --115°c (ductile).
Perch (1956 a).
Fig. 7
i2 i i ! !
~8
z_
%~ -196 ° C .
o J I I I I
$ IO 15 2o 2s
II,/'d, iN. "~
The relationship between fracture strength and grain size for zinc.
and Quarrel (1954).
Greenwood
P.M.S. 20
428 A. No S t r o h on a
The experimentally determined values o f the slope K o f the line

m a y be c o m p a r e d ~4th those calculated:~ from eqn. ( I 1 ) ;
aI v s d - * / ~
as it is seen from table 1, the a g r e e m e n t is v e r y satisfactory.
Fig. 8
~2
z8
J
ff
/ J
W)"
20
E
"' $ Q 78°K ~ IO@%/mie.
. . . . O,OI% Imin.
<> tSS°K ~ : I%/min:
~ 3 4 5 6 7
The relationship between fracture strength and grain size for magnesium.
ttauser et al. (1956).
Table 1. Observed and Calculated Values of c o n s t a n t K
Calculated Observed
Metal Surface energy 7 according to (11 ) from gl"aphs
erg/cm 2 c,g.s, milts e.g.s, units
Fe 1600 1.9 × l0 s 2-1 × iO s

Zn 800 0.79 1.05
Mg 570 0.52 0.85
Mo 2240 ! 2.5 2-6
~ N o dircct determination of 7 for molybdmmm is available, and the

vahm used is t,hat estimated by Taylor (195i}.
:~ 111 calculating the values shown, the elastic constants used ~re those fi)r
the polycrys!'dline aggregate rather than for a single crystal for ~e are cow
eerned with a stress field extending across two grains of different orientation.-.
and st) want some sort of mean between the elastic constants of the individuai
grains.
Fig. 9
i |- i
!
-
f
[ 2 -~/2 3 -tla 4 5 6
d mm
The relationship between fracture strength and grain size for molybdenum.
ge,chtold (1953, 1954).
3.5. Hydrogen Embrittlement

The importance of the surface energy in determining tile fracture
strength is seen very clearly in the phenomenon of hydrogen embrittle-
ment. The presence of hydrogen in steel causes cleavage and inter-
crystalline fracture even at room temperatures, and Petch and Stables
(1952) have attributed this to adsorption on the newly formed fracture
surfaces with consequent lowering of the effective surface energy. Since
the hydrogen will take time to diffuse to the crack this will grow only
slowly at first, though once it has grown sufflcient]y (viz. to the Griffith
length corresponding to a clean surface) sudden fracture can occur.
Thus there will be a delay time in developing the embrittlement and
so the effect will be apparent only for sufficiently long durations of loading ;
this prediction is confirmed by the results of Perch and Stables shown
in table 2. Moreover: diffusion can take place more easily along the
grain boundaries than through the interior of the grains so t h a t the
2G2
concentration of hydrogen will be greatest near the grain boundaries;

"~his will favour intercrystalline fracture as is in fact observed.
Table 2. The Effect of Duration of Loading on the Hydrogen Embrittle-

ment of Steel as Measured by the Reduction of Area per cent.
Perch and Stables (1952)
Hydrogen free
Hydroge~l charged
0, 60.5
60-5
mt
61.0
27.8
I f the whole of the effect of the hydrogen can be attributed to a change

in the effective surface energy, we should obtain the same dependence
of fracture strength on grain size as before (eqn. (12)) but with a reduced
value of K ; ~hat this is so, is seen from the results of Petch (1956 b)
shown in fig. 10.
Fig. 10
1 . 0 ~" ~Z)O
50
80
0 4C .0
,0 ~ ~o ~, ~,0 -
oj IE
0
i J 40 ~
0 ~'~ .0 ~0
b" ~ C
20
2 4
1
6 8 IO
"]?he relationship between fracture stress and grain size of a mild steel at
18°c (a) with 10 cms hydrogen/100g, (b) without hydrogen. Petch
(195'6 b).
I f the adsorption takes place at temperature T from a diatomic gas

at. a pressure p, the gas dissociating on adsorption, Langmuir's isotherm
leads to the effective smfface energy
ra=~,-- 2 / ' ~ T log {1.Jr (p/px/~)l/2},
where/'s is the number of molecules adsorbed per unit area at saturation,
and Pl/2 is the pressure at which the number of molecules adsorbed is
½/'s. Values of the surface energy calculated in this way for iron with
adsorbed hydrogen are shown in fig. 11 ; the effect is large. By sub-
stituting ~'1 for ~ in eqn. (11) the change of strength produced by the
Theory. of the Fracture-of Metals 431
h y d r o g e n m a y be c a l c u l a t e d ; for a grain size corresponding to

d-1/~--=-5 m m -1/2, the calculated value o f al--a o is 9.5 tons/in. 2, agreeing
well with the observed v a l u e o f 11 tons/in. 2.
:Fig. 11
- 200
i
U
-400
w
Z
I - 600
~J
-800 ~
n~
4 6 8
c.c. H 2 / ~ 0 0 cM.
Lowering of surface energy of iron by hydrogen adsorption. Petck (1956 b).
3.6. Slip near a Piled Up Group

I f a crack is to form near a piled up group o f dislocations in the m a n n e r
we are considering, then it is necessary t h a t the material round the
group should r e m a i n u n y i e l d i n g ; for plastic flow will t e n d to relieve
the stresses here so t h a t t h e y will be insufficient to initiate a crack.
Now plastic flow could conceivably s t a r t in either of two distinct ways :
existing Frank-Ree~d sources n e a r b y might generate new dish)cations~
or a l t e r n a t i v e l y the stresses might be sufficient to initiate slip in the
perfect lattice. The latter possibility will be discussed first
F r a n k (1950) has shown t h a t t h e initiation of slip in the perfect lattice
should occur when a shear stress of order G/20 acts over a region whose
dimensions are at least 5 to 10 atomic spacings. I t is difficult to say
from p u r e l y theoretical considerations whether such slip wiU occur
more easily t h a n fracture. The problem has been considered by Stroh
(1955 a), who concludes ~hat it is not unlikely t h a t fracture will occur
first, b u t no completely satistactory t r e a t m e n t exists. In bott~ slip
and fracture, the same forces, namely the forces binding the atoms
together in the lattice, must be overcome, and so the two processes m a y
be expected to be of comparable difficulty; but present calculations
are not sufficiently precise to decide unambiguously which should occur
first. We may, however, ask whether the production of new dislocation
loops in the perfect lattice helps to explain the observed behaviolrr
of metals ; the view taken here is that it does not. For, if slip could
be initiated in this way, this would cause the deformation to spread
through the material from one grain to the n e x t ; it would be natural
to identify this process with the spread of a Lfiders band. But the
s~ress at which this occurs (the lower yield point) is strongly temperature-
dependent, while the energy of a dislocation loop is so large that temper-
ature is unlikely to play any appreciable part in its production; it is

therefore difficult to account for t h e spread of a Liiders band in this
way. On the other hand, as will be seen in § 4, the correct temperature-
dependence is obtained if slip is always supposed to start from F r a n k -
Read sources already present in the material. We shall therefore reject
the possibility that the stresses of the piled up group initiate slip in the
perfect lattice.
It follows then that we need consider only plastic flow due to disloca-
tions from F r a n k - R e a d sources, If the material is brittle and plastic
flow does not occur round the piled up group of dislocations, then the
sources must be locked so as to be unable to operate. This locking
m a y arise either b y the segregation of impurities to the dislocations
(Cottrell locking), or because some slip systems are inherently difficult
to operate (e.g. non-basal slip in the hexagonal close packed lattice).
We consider these in turn.
(i) Cottrell locking of dislocations

Cottrell (1948) and Cottrell and Bilby (1949) have shown how im-
purities (and especially interstitial impurities such as carbon or nitrogen
in iron), can segregate to the dislocations and anchor them. The disloca-
tions can only move by |caving their atmospheres of impurities behind,
and so a stress :is required to pull the dislocations free ; this stress is the
upper yield point. Once the dislocations are free they can move more
easily ; the stress then falls after the yield point has been reached and the
deformation continues at a lower stress (the lower yield point).
Yield points have been observed in a number of metals (zinc, Wain
and Cottrell 1959, cadmium, Gibbons and Cottrel! 1947, titanium,
Churchman 1955, ~-brass, Ardt.ey and (~ottrell 1953), but they are
especially prominent in those having a body-centred cubic structure;
in addition to iron, molybdenum (Trury and Kraus 1936, 1937), tungsten
(Bechtold and Shcwmon 1954) and ~antalum (Bechtold 1955 ) all show
pronounced upper and lower yield points. The occurrence of such
yield points is an indication of the ~trength of the Cottrell locking in these
metals.
There are two reasons why interstitial atoms should interact parti-
cularly strongly with dislocations in the body-centred cubic structure.
First, this lattice has smaller interstices t h a n do other lattices so ~hat an
interstitial atom has to expand the lattice to a greater extent giving a
corresponding larger energy of interaction. Secondly an interstitial
atom in this structure not only produces a pure volume change, but also
distorts the lattice, changing it locally from cubic to tetragonal ; thus
the interstitials will interact with the dislocations through the shear
stresses as well as through the hydrostatic component of the stress.
If, then, Cottrell locking occurs, it will anchor the Frank-i~ead sources
against any stress less than the upper yield point, and so can provide
the rigid matrix round the piled up group which is needed if fracture
is to occur. The Cottrell locking will also help produce fracture in

another w a y : the most difficult step in operating a source will be to
free it from its locking impurities ; this however needs to be clone only
for the first dislocation loop produced by the source. Hence once a
source has been freed, it will be easier to continue operating this source
than to free another ; and so a small number of sources will operate,
each producing m a n y dislocations. Cottrell locking will ~hus lead to
the formation of the large piled up groups needed for fracture. This
conclusion is confirmed by the observation of course shp (fig. 4, P1. 14).
The view that the brittle behaviour of these metals depends on the
dislocations being locked by impurities is supported by the observation
of P r a t t (private communication) that if a single crystal of iron is deformed
shghtly at room temperature (so as ~o displace ~he dislocations and
free them from their impurity atoms) ~nd then cooled to liquid air
temperature, the metal is still ductile though tested in an orientation in
which brittle fracture would normally occur. Smith and I~utherford
(1957) have found that brittleness is eliminated in removing impurities,
and zone refined iron could be deformed at 4"2°K t.o an 80% reduction
in area before fracture. Earlier work b y Wain and I-Ienderson (1953)
had shown t h a t the room temperature brittleness of chromium could
be overcome by purification.
(ii) Geometrical limitations on slip

In metals with a close packed hexagonal structure, slip occurs primarily
in ~he basal plane, and this provides only two independent slip systems
(corresponding to two of the three slip directions in the basal plane).
This is not sufficient to allow a polycrystalline aggregate to deform and
maintain coherence between the grains, for yon Mises (1928) has shown
that to obtain an arbitrary plastic strain five independent slip systems
are needed. Nor does twinning appear to be sufficient to provide ~he
necessary additional degrees of freedom; for example, in magnesium
the maximum strain obtainable from twinning is 7~o for optimum
orientation and will be much less for randomly orientated grains. At
high temperatures, pyramidal and prismatic slip occur and can provide
434 A . N . Stroh o n a
the additional slip systems needed, b u t these are not operative at low
temperatures. I f then one grain yields, the surrounding grains m a y be
orientated so as to be unable to deform in a way which will accommodate
the slip in the first grain. Thus we can obtain one deforming grain in a
rigid matrix and slip in this grain can proceed until the piled up groups
of dislocations formed on its boundary are large enough to initiate a crack.
In these metals, we expect to see in the deforming grain not a single
coarse slip line but a number of fine slip lines. However, a crack can
just as well be initiated by the cooperative effect of several smaller
piled up groups at the ends of a cluster of fine slip lines, provided t h a t
the total number of dislocations is still given by eqn. (9), (Stroh 1955 a).
The relation between grain size and fracture strength is unchanged.
On this view then the brittle behaviour of hexagonal metals depends

on the limited number of slip systems available. The way in which the
operation of addition slip systems can cause the material te be ductile
is very clearly shown by Hauser et al. (1957) by observations on ~ solid
solutions of lithium in magnesium. The addition of lithium to magnesium
causes prismatic slip (10i0), [2ii0] to occur at progressively lower
temperatures and for the alloys richest in lithium (10.4 and 14.5 atm °/o)
prismatic slip was observed at temperatures as low as 4"2°K. At the same
time the ductility increased markedly, and the 14.5~o alloy necked and
broke in a typically ductile manner at all temperatures investigated.
In considering why prismatic slip should occur, it is to be noted that
the addition of lithium decreased the axial ratio from 1.624 (pure Mg)
to 1.610 (Mg, 14.5°/o Li) ; this latter value is nearer the axial ratios of
titanium (1.59), zirconium (1-59), and beryllium (1-57) in all of which
prismatic slip is normal. Hauser e t a l . suggest ~hat the Peierls-Nabarro
force needed to move the dislocations depends on both the slip plane
and the axial ratios, and, for sufficiently small axial ratios, becomes less
on the prismatic planes than on the basal plane. I f the Peierls-Nabarro
force is large the dislocations will be unable to move except with the
help of thermal fluctuations so t h a t this slip system will be observed
only at high temperatures.
§ 4. THE LOWER YI]~LD POINT

In a metal such as iron where Cottrell locking occurs, thermal
vibrations will help free the dislocations from their locking impurities
and so they will be less firmly bound at high temperatures than at
low. At a sufficiently high temperature the stresses of the piled up
group will be able to free the dislocations near it, and then slip will
continue into the next grain instead of fracture occurring and so the
metal will be brittle at low temperatures, ductile at high. The stress at
which the slip propagates through the material, if it is ductile, is the
lower yield point ; we shall consider the nature of the lower yield point
in the present section and postpone till later (§ 5) a discussion of the
conditions determining the transition from brittle to ductile behaviour.
As Petch (1953) has emphasized the lower yield and the cleavage
strength depend ir~ the same w a y on the grain size, suggesting that
similar mechanisms operate in the two cases. Figure 12, due to Cracknell
and Petch (1955), showing the relation between the lower yield point
and the grain size oi" mild steel m a y be compared with fig. 6 ; similar
results on the lower yield poi;at have been obtained by Hall (1951) and
PeLch (1953). Both cleavage and yielding m a y be interpreted as due
Fig. 12
,,-VI
/
/o,//
j /ei~ o/
16
z
J A J
/-
w 12 4 .
,m $
",,J
~
s
2
~
8 /*--/ "
/
- J |,-I
1
I
| 2 3 4 5 6 7 8
The relationship between lower yield point and grains size for (1) annealed
mild steel En 2 ; (2) E . 2 nitrided ; (3) En 2 quenched from 650°c ;
(4) En 2 quenched, aged one hour at ]50°c ; (5) En 2 quenched, aged 100
hours at 200% ; (6) annealed Swedish iron. Cracknell and Peteh (1955).
436 A.N. Stroh on a
to the stresses r o u n d a piled up group ; cleavage, to the tensile stress,

and yielding to the shear stress.
A t a distance r from the head of a piled up group the shear stress is
a - = f i % ( L / r ) 1/2 . . . . . (13)
. .
where fi is a numerical factor, a p p r o x i m a t e l y u n i t y b u t depending
on the orientation. ( E q u a t i o n (13) for the shear stresses corresponds
to eqn. (5) for the n o r m a l stresses.) r must be taken equal to the distance
f r o m t h e piled up group of the source operated. Now fig. 12 shows t h a t
a n u m b e r of quite different specimens of steel all give lines with the same
slope ; hence this slope cannot d e p e n d on a n y q u a n t i t y such as the mean
spacing of the dislocation n e t w o r k which would v a r y considerably from
one specimen to another. We therefore suppose t h a t there will occur

n u m b e r of sources of various lengths and at various diseances f r o m the
piled up group ; of these the most f a v o u r a b l y situated ones will auto-
magically be selected a n d o p e r a t e d first, and these will always be the same
distance (i.e. for different specimens b u t at a given t e m p e r a t u r e ) from
the piled up group. The reason why some sources are f a v o u r e d above
others is as follows.
F o r a source to operate, the stress on it must, firstly, be great enough
to pull it free from its locking impurities ; for this the stress m u s t be at
least equal to the u p p e r yield point e~. Secondly, w h e n t he source has
been freed t h e stress m u s t be sufficient to bow it out in the characteristic
m a n n e r of the F r a n k - R e a d source shown in fig. 13; for this a stress
Fig. 13
(a ) kb) t.c)
(.a)
The operation of a Frank-Read source.
of G b / l , where 1 is the tength of the source, is needed. I t is clear from

fig. 13 t h a t a source of length I needs an area whose linear dimensions
are at least I to operate in, and so the source cannot be nearer t h a n a
distance I to the piled up group ; in general the nearest sources of length
I will be a distance ~l from the piled up group, where x is a constant of
order, but greater than, unity. T h e n the sources v e r y near the piled up
Theory of the Frac~ture of Metals 437
group will all be short and so difficult to bow out, while the stress on
those far away will be too small to pull them free of their locking im-
purities ; hence it follows t h a t the sources at some intermediate distance
will be easiest to operate. The most favoured sources are easily seen
(Stroh 1955 b) to have length l=Gb/a~ and so to be a distance r-~Gb/au
from the piled up group.
The corresponding shear stress a 8 is found from eqn (13) to be
%..----(~GbaJLfl2) 1/2.
I f t h e specimen is tested in simple tension, the tensile stress %-~2% ;
also the grain diameter d-~ 4L, giving
a ~-~4(o:a~,Gb)l/~fl-~d-1/2 ;
or including a constant frictional stress % as in § 3.4, the lower yield

stress is given by
crv~K*d-1/2+%, . . . . . . . 14)
where
K*:4(o:a~Gb)l/2f1-1 . . . . . . . . (15)
Equation (14) is the form of the relation found experimentally. K*,
in accord with the experimental results, is the same for all specimens
(at one temperature) ; for all the factors on the right-hand side of eqn.
(15) except au are clearly the same, While a~ will have a constant maxi-
mum value provided there is enough carbon or nitrogen present to form
saturated atmospheres round the dislocations, and this be so in all but
the very purest materials.
The values of % in eqns. (12) (i.e. for cleavage) and (14) (yielding)
have been found by Perch (1953) to be the same ; this is to be expected
from the interpretation of a0 as a force opposing the motion of the disloca-
tions.
At two different temperatures T 1 and T 2 eqn. (15) gives
Kl*~K2*=(a~,G)II/2/(~,G)21/~ . . . . . . (16)
where the suffices refer to the temperatures. To test this equation we
m a y use the results of Perch (1953) at liquid air temperature and those
of Cracknell and Petch (1955) at room temperature. From their published
graphs we find that K * ~ 0 . 7 2 × 10s dynes/cm-a/2 at room temperature
and 1.9× l0 s at --196°c, the ratio being 2.6. Both the theoretical
analysis of Cottrell and Bilby (1949) and the observations of McAdam
and Mebs (1943) give a change in the upper yield point a~ by a factor of
4.1 in this temperature range ; the Change in G is about 10~o ; and hence
the right hand side of eqn. (16) is equal to 2.1, agreeing fairly well with
the value 2.6 estimated from the slope of the graphs. With
a ~ - 7 × 1 0 S d y n e s / c m 2, corresponding to room temperature, the value
of K* becomes 4.6 × 107~1/~/fl ; as ~ and fl are of order unity this agrees
with Cracknell and Perch's value in order of magnitude. The model
thus appears able to account for the principle features observed in the
behaviour of the lower yield point.
438 A . N . S t r o h on a
4.1. T h e Stress Needed to Move a Dislocation

~ ' e h a v e seen t h a t , in order to obtain a relation o f the correct form
between the cleavage strength, or the lowor yield point, a n d the grain
size, it is necessary to introduce the idea of a frictional force opposing
the m o t i o n of the dislocations. Figure 12 shows clearly t h a t this
frictional force, %, which is measured b y the int,srcept on t h e stress axis,
differs considerably for different specimens; this is in direct c o n t r a s t
to the b e h a v i o u r of the gradients K and K*, which are the same for all
Fig. 14
I0
.//
O,, "O
/
B !@
=
J
d j J8
~ 6
I
@
./0
J
/
o
o ~/
Effect of the concentration of Cq-N on a0, Cra,cknell and Petch (1955).

specimens. Cracknelt and P e t c h (!955) suggest t h a t % is due to internal
stresses a r o u n d fine precipit, ates or r a n d o m l y distributed solute atoms ;
the variation of a0 shown in fig, 12, on nitriding or quenching (which
alters the conceI,tratio~l)~ and on que~ch ageing (wtfich alters the state
of dispersion), is in accot'd with ti:is idea. T h e y base a qua~xtitative
treatmen:t of the effect, on the work of Mott aI~d N a b a r r o (1948) aJ,d
Mott (1952) which leads to the result that a 0 should vary linearly with
the concentration of carbon and nitrogen over the range of concentra-
tions used, a prediction which fig. 14 shows is confirmed by observation ;
the predicted slope, however, is rather less than t h a t observed, probably
owing to apprOximations which have had to be made in the theory.
The value of % which remains at zero concentration has been inter-
preted by Cracknell and Petch as the Peierls force on the dislocations.
I f this is correct it means that the Peierls force in ~-iron (and probably
in other body-centred cubic metals) is rather larger t h a n has hitherto
generally been supposed. Heslop and Petch (1956) have found that
the part of a 0 which is independent of the concentration is strongly
temperature dependent; this is consistent with its interpretation as a
Pe;:erls force.
§ 5. THE BRITTLE-DucTILE TRANSITION

5.1. The Probability of Brittle Fracture
In the preceding sections the brittle fracture at low temperatures,
and the general yielding at higher temperatures have been considered ;
it remains to discuss the conditions which determine the transition from
brittleness to ductilityt.
By far the greatest number of experimental investigations have been
done on iron and steels, and the experimental results which are quoted
in this section all refer to these metals. I n such cases, the locking of
the dislocations is due to impurities, and the transition will be discussed
in terms of this. However, the ideas used can be applied more generally,
and any way of locking the dislocations from which they can be released
under a stress by supplying an activation energy, would serve
equally well. For example, the discussion of § 3.6 suggests that in
hexagonal metals the brittleness is due to the locking of the dislocations
in non-basal planes by a large Peierls force ; but the transition may still
be treated as in the present section.
The transition is usually studied experimen£ally by impact test, such
as the Charpy and Izod tests, in which the energy absorbed in breaking
a specimen is measured.. At low temperatures when the specimen is
brittle this energy is small, but it is greater at temperatures for which
the specimen is ductile. In the transition region a wide range of energy
values are obtained; a number of workers (Vanderbeck et al. 1953,
Ulmo and Bastenaire 1953, Crussard et al. 1956) have shown t h a t these
values are distributed bimodally, as is apparent from the histogram shown
in fig. 15; low energies characteristic of the brittle condition and high
energies characteristic of the ductile both occur, but intermediate values
t A very full experimental investigation of this topic has recently been made
by Crussard et al. (1956). They distinguish between the initiation transition
considered here and the propagation transition corresponding to a change in
the mode in which the crack spreads.
are exceptional. Thus the results of such a test m a y be represented

ideally as in fig. 16. AB is the brittle curve and CD the ductile ; below
the temperature T 1 all points lie on AB so t h a t this is the brittle region,
while above T 2 there is the ductile region. The transitional region
extends from T 1 to Tg, and here points m a y lie on either curve according
Fig. 15
Z
l l I I I
U I'34
w 5 a
n
I
O I
-1 J
W
~r
! I I
z O I 2 3 4 S 6
kq/cm z
Histogram of mild steel specimens broken by slow bending at --140°o
(15 specimens). Crussard et al. (1956).
Fig, 16
CT~
| I
Q) J I
¢- I I
tlA I I
I i
I I
, I I
T, T2 Temperature
Variation of fracture energy with temperature typically obtained in an impact
test.
to a probability law depending on the temperature, so t h a t the two

types of fracture, ductile and brittle, m a y coexist at the same temperature.
We wish to find the probability t h a t the metal will be brittle at a n y
given temperature.
Theory of the Fracture of Me~als 441
Now according to the ideas we have developed, the dislocations are

locked b y Cottrcll impurities. At. high temperatures, t h e r m a l fluctua-
tions will be able to flee t h e dislocations r o u n d a piled up group and
so relieve the stresses t h e r e ; in this case, t h e m e t a l will be ductile.
On the other hand, at low t e m p e r a t u r e , the t h e r m a l fluctuations will be
insufficient to free the dislocations before the piled u p group has grown
large enough to generate a crack ; t h e n t h e m e t a l is brittle. We t h e r e f o r e
i d e n t i f y the p r o b a b i l i t y of brittle f r a c t u r e with the p r o b a b i l i t y t h a t the
dislocations n e a r a piled up group will not b e released.
To free a dislocation, an activation e n e r g y U(a), depending on the
stress ~, m u s t be supplied, and so the p r o b a b i l i t y per u n i t t i m e of the
dislocation being freed m a y be w r i t t e n v exp {--U(e)/lcT}, where v is
a c o n s t a n t of the dimensions of a frequency. H e n c e t h e m e a n time~

elapsing before a dislocation is released is
r = ( 1 / v ) e x p {U(a)/kT} . . . . . . . (17)
T h e release of a dislocation will be a p u r e l y r a n d o m event, subject only
to the restriction t h a t it occurs on t h e average after time r. Such
r a n d o m processes are familiar in m a n y branches of physics, e.g. radio-
active decay, a n d the p r o b a b i l i t y t h a t t h e event, here the release of t h e
dislocation, will not h a v e o c c u r r e d at time t is
p= exp (--tt~),
or substituting r from eqn. (7),
p = exp I--at exp {-- U(cr)/kT}] . . . . . (18)
As the piled up group is formed the stress ~ which it produces on a n y
dislocation will increase from zero to some m a x i m u m value. T h e
p r o b a b i l i t y of the dislocations being released while the stress is still
small is negligible, and so in eqn. (18) ~ m u s t be t a k e n as the m a x i m u m
value of the stress ; t will be a measure of t h e time during which the
stress is near t h i s value. T h e n eqn. (18) will give the p r o b a b i l i t y o f
brittle fracture at a n y t e m p e r a t u r e . Because o f the double exponential,
p wilt change fairly a b r u p t l y from 1 to 0 as T increases, so t h a t the
model reproduces the sharp transition observed. The transitions will
occur at a t e m p e r a t u r e
T~--U(cr)/k log vt . . . . . . . . (19)
The a c t i v a t i o n energy U(e), needed to free a dislocation depends on
the stress ~ acting on it, and ~his in turn depends on the distance of t h e
dislocation from the piled up group. Thus in general there will be a
wide range of activation energies. H o w e v e r the consideration of
t Equation (17) gives the relation between tile time to release the dislocations
and the stress. The existence of such a ~relation is just the interpretation
in dislocation theory of the experimental result (e.g. Taylor 1946) that the
observed yield stress depends on the duration of loading.
§ 4 led to the conclusion t h a t the F r a n k - R e ~ l sources which were

most easily o p e r a t e d would occur at some definite distance from the
piled u p group, and it is the operation of these sources which will deter-
mine whether the metal is ductile or brittle. Accordingly the value of
the activation energy which is of i m p o r t a n c e is t h a t needed to free one
of these most easily o p e r a t e d sources when the piled up group has grown
just large enough to initiate a crack. This a c t i v a t i o n e n e r g y has a
uniquely d e t e r m i n e d value, and it is to this value t h a t we shall refer
in speaking of the activation energy.
Stroh (1955 b) has a t t e m p t e d to estimate T~ b y using Cottrell and
Bilby's calculations (1949) of U ( a ) ; t h o u g h a transition ~emperature
of the correct order was obtained, there are so m a n y sources of u n c e r t a i n t y
involved t h a t little weight can be a t t a c h e d to the final result.

I t seems b e t t e r to compare equation directly with experiment. I f a
n u m b e r of specimens ara broken at the same t e m p e r a t u r e , t h e y m a y be
classified into ewo groups according as t h e y fail in a brittle or a ductile
m a n n e r ; the energy absorbed in fracture provides a useful criterion for
this (fig. 15). If this is done p m a y be identified with the fraction of the
total n u m b e r of specimens which break brittly at each t e m p e r a t u r e .
E q u a t i o n (18) m a y be rewritten
log l o g ( 1 / p ) = l o g v t - - U / k T , . . . . (20)
or log log (I/p) is a linear function of I / T ; the results of V a n d e r b e c k
etal. (1953)~ shown in fig. t 7 are seen to follow this law within the experi-
m e n t a l scatter. F r o m the slope of the line, we obtain t h e a c t i v a t i o n
energy U = 0 . 2 8 ev. This value has been entered in table 3, t o g e t h e r
w i t h estimates of the a c t i v a t i o n energy o b t a i n e d b y t r e a t i n g the results
of B a e y e r t z et al. {1949) and ,of Ulmo and Bastenaire (1953) in the same
way. These values will be discussed later (§ 5.4) after we h a v e considered
other properties of the transition also leading to an estimate of the
a c t i v a t ; o n energy. Meanwhile we m a y note t h a t ~n order of m a g n i t u d e
the values obtained are not inconsistent with those e x p e c t e d from the
theoretical calculations of Cottrell and Bilby (1949).
5.2. Effect of Strain Rate

In eqn. (18) describing the b r i t t l e - d u c t i l e transition, the only q u a n t i t y
whh, h will depend on strain rate is t, the time for which the stress on a
dislocation is near the vMue a. W e e x p e c t t to be given a p a r t from a
numerical factor, b y
t~-o/g,, . . . . . . . . . (21)
where a dot over a letter denotes differentiation with respect to time.
The stress a which the piled up group produces will increase in p r o p o r t i o n
t Vanderbeck et al. have fitted a Galton curve, obtained by integrating a
Gaussian distribution, to their experimental points ; however, this is not based
m~ ~ny definite atomic mecharfism,
Theory of the Fraeture of Metal8 443
to the stress pihng up the dislocations ; t h a t is t o the part a l - - a 0 of

the applied stress (in the n o t a t i o n of § 3.4) which is not used in over-
coming the frictional resistance to the motion of the dislocations. Hence
eqn. (21) may be written
t-(~r-%)/a~,
or using eqn. (12)
t~__K/dl/~b].
--.....
Fi~. 17
cY"
I I I I I I
o
0.1
-0. I
• \
-0.5 4- "!.
4-
-0-7
-I-I
, I I I I I ,a'~
5.0 5.z 5.4 5.5 5.e 5.0 6.z 5.4
t000/T
Variation of log~o log~0 (l/p) with 1/T for steel. Experimental points due to
~Vanderbeek et al. (1953).
As long as the material is brittle and general yielding does not occur.
the applied stress a/will be related to the strain by Hooke's law, so t h a t
finally we have
t~_.K/Edl/~ . . . . . . . . . (22)
where ~ is the strain rate.
P.M,~. 2 H
Inserting this value of t into eqn. (19), we find t h a t the relation between
transition temperature and strain rate is given by
1/To=--(k/U) log ~ + C . . . . . . . (2a)

whore C is independent of To and i.
Such a linear relation between 1~To and log i has been observed by
W i t t m a n and Stepanov (1939) whose results are reproduced in fig. 18,
Fig. 18
7 I I I I I I
\
6 \
\
0-
\
\
-2 I I I I I I
5.5 4.5 5.5 6.5 7.5 8.5 9.5 I0.5
I000/T
Dependence of transition temperature on strain rate for steel. Wittma~
and Stepanov (1939).
and similar results have been obtained by MacGregor and Grossman

(1952). Tim slopes of these lines also lead to an estimate of the activa-
tion energy U, and the values obtained are shown in table 3
Table 3. Values of the Activation Energy needed to Free a Dislocation

in Steel.
Activation
Type of Observation energy U Authority
eY
Probability of brittle 0.28 Vanderbeck et al. (1953)

fracture 0"19 Ulmo and Bastenaire (1953)
0"36 Baeyertz et a/. (1949)
Strain rate 0.24 Wittman and Stepanov (1939)

0.44 MacGregor and Grossman (1952)
Variation of grain (a) 0.30"~

size (b) 0-22f Hedge et~l. (1949)
5.3. Effect of Grain Size

It is generally agreed that a fine grained specimen has a lower transition
temperature than a coarse grained one. On our model, there are several
reasons .why t h a t should be so. Firstly it follows immediately from the
considerations of § 5.2 that t, the effective time for which the F r a n k - R e a d
sources round the piled up group are stressed, depends on the grain size
d ; eqn. (22) shows that t is proportional to d -1/2. This in turn leads
to a dependance of the transition temperature on d ; though only about
15 °/o of the observed variation can be accounted for in this way.
A more important contribution arises because the grains are not all
identical but differ in size and orientation, so that some yield more easily
than others. Slip will, on the model, propagate by a piled up group
in one grain operating sources, and so starting slip, in the next ; a grain
is ductile if it yields before the piled up group in the previous grain to
slip has grown large enough to initiate a crack. Now suppose that the
temperature and strain rate are such t h a t the average grain is just ductile.
At any stage in the deformation there will be a number of grains which
may slip n e x t ; the one to do so will be t h a t one which yields most
easily. As long as there is at least one grain ready t,o slip, the deformation
can continue and the material as a whole will be ductile. Occasionally,
however, it will chance to happen t h a t all the grains through which
slip can propagate yield only with difficulty; then the applied stress
must increase slightly and the piled up groups will grow larger; this
may be sufficient to initiate a crack so that the material, as a whole, is
brittle. Now the probability of finding a grain yielding at any time is
the product of the number of grains considered and the probability
v exp (--U/kT) of a given grain yielding. I f specimens of the same
geometrical size and shape are considered the number of grains in any
region is inversely proportional to the volume of a grain and so propor-

tional to d -a. Thus the frequency v must be multiplied by a factor
proportional to d-a; it is convenient to redefine v so that it includes
this factor; the transition temperature is then still given by an
expression of the form (19).
The two effects we have been considering give the product vt propor
tional to d-~/2; eqn. (19) then gives the relation between transition
temperature and grain size, as
1/T,=-- 7 ( k / U ) log d + C . . . . . . (24)
where C is independent of T o and d. Since the conventional grain-size

numbers (ASTM or Jernkontoret) form a logarithimie scale, an increase

of unity in the grain-size number N doubling the number of grains per
unit area, eqn. (24) m a y be written
1/T¢=I.21(k/U)N-~C'. . . . . . . (25)
Hodge et al. (1949) have concluded that there is a linear relation between
transition temperature and grain-size n u m b e r ; however their results,
which are replotted in fig. 19 agree equally well with eqn. (25).
Fig. 19
5.5 I I I I ~ /
6.0 - / / ~ e C b )ei ~ / / ~ C
5.5-
0o5.0-
j /4)
4.5- e * -
4.0-
3.5 ~ I I I I R
2 3 4 5 6 7
ASTM qrain size number
Dependence of transition temperature oil grain size. Steel with (a) 0.03%
Ni ,and (b) 3"64% Ni. Experimental points of Hodge et al. (1949).
The slope of the lines give other estimates of the activation energy U,
which have again been entered in table 3.
Theory of the Fracture of Metal~ 447
5.4. The Activation Energy

In the preceding sections we have seen how the activation energy U
may be found from a number of quite different types of observation.
The estimates of U which have been obtained are collected in table 3 ;
there is a marked tendency for values near 0-3 ev to occur.
P a r t of the variation in values obtained is doubtless due to experi-
mental uncertainties which in some o f the cases may be important.
I t is unlikely, however, t h a t this accounts for the whole of the variation,
and indeed there seems no justification tbr expecting all the entries in the
table to be identical ; t h e y refer to experiments m~de on different s~eels
having different compositions. The presence of small amounts of
alloying elements is known to produce changes in the transition temper-

ature (cf. for example I~inebolt and Harris 1951) ; though the mechanism
by which this is brought about is not as yet understood, the effect may
well be due, in part, to a change in the activation energy needed to
free the F r a n k - R e a d sources, which is caused by such impurity atoms.
For this reason it would be of great interest if the different types of
observation we have been considering were to be repeated on specimens
o f the same composition so that the activation energies deduced would
be strictly comparable.
l~earing in mind the variation in composition, we see t h a t the various
activation energies obtained agree as well as may be e x p e c t e d ; in
particular there is no evidence for a systematic difference in the activation
energies obtained from observations of different type. This general
agreement, and also the fact t h a t the value obtained (about 0.3 ev) is
consistent with Cottrell and l~ilby's theoretical curves (1949), is evidence
in favour of the model of fracture which has been developed.
5.5. Notch Brittlene6"s

The presence of a notch in the specimen has an important effect oil
fracture and produces a large increase m the transition temperature. In
considering how this is brought about, we note the ibllowing properties
of the notch.
(i) I f yielding has not yet occurred and the material is elastic, the
notch will act as a stress concentrator so t h a t large stresses occur near
the notch. The smaller tile radius of curvature of the root of the notch.
i.e. the sharper the notch, the greater is the stress magnification produced,
but these large stresses occur only in a relatively small region near the
notch root.
(ii) Consider a notched specimen, subjected to a uniaxial tension
(fig. 20). Yielding in the notch tends to produce a contraction in the
cross section there, which is opposed by the unyielding material on either
side; thus a transverse tension will be produced. The effect of the
notch then is to change the stress system from a uniaxial to a biaxial, or
triaxial, one. Since the maximum shear stress is equal to h.~lf the
448 A . N . S¢roh on a
difference of the principal stresses, the shear stress will be reduced by

the presence of the notch, while the maximum normal stress is unchanged.
Plastic flow depends on the shear stresses, and so the axial stress needed
to produce yield is increased by the presence of the notch ; according to
Prandtl (1923), Orowan (1945) and Hill (1949), the axial yield stress is
increased by a factor of about 3~
Fig. 20
Stress system near a notch.
(iii) Near the notch the strain rate will be enhanced. For elastic
strains this follows from (i) on account of the proportionality between
stress and strain. Wtmn yielding occurs the flow will be such as to
relieve the stresses present; thus the elastic stress concentration near
the root of the notch is converted into a region of plastic strain concentra-
tion, and hence there will also be a high plastic strain rate near the
notch.
The classical theory of notch brittleness which originates from the
work of Mesnager (1906) and Ludwik and Scheu (1923) and which has
subsequently been elaborated by Orowan et al. (1945) and Orowan
(1945), takes (ii) as the principal effect in providing brittleness. Fracture
is assumed to depend on the tensile stress, so the triaxial stress system
produced by the notch will tend to favour fracture rather than yielding.
On the present model this will not be the case, for both fracture and
yielding depend on the shear stress and so will be affected in the same
way by a change in the nature of the stress system. There is some
experimental evidence t h a t this is the case. McAdarfl et al. (1947)
slightly deformed notched specimens at room temperature (where they
were ductile) so t h a t some plastic flow took place round the root of the
notch and the stress concentration considered in (i) above was eliminated ;
the triaxial stress system (ii) of course, still remained. On reloading
the specimens at a lower temperature, it was found that the fracture
stress was increased b y the n o t c h in the same m a n n e r as t h e yield stress.

F r o m their experiments on the p r o p a g a t i o n o f brittle cracks, to be dis-
cussed more fully in § 6.3, Felbeck and Orowan (1955) conclude t h a t the
triaxial n a t u r e of the stress is u n i m p o r t a n t and t h a t the decisive factor
in determining whether brittle or ductile f r a c t u r e will occur is the velocity
dependance of the yield stress. Other evidence supporting the view
t h a t fracture depends on the shear stress r a t h e r t h a n the tensile stress
has been reviewed by Zener (1948).
The view t a k e n here is t h a t the increase in st~'ain rate mentioned in
(iii) is the principal effect of the notch in causing brittleness. I t was
see/~ in § 5.2 t h a t an increase in strain rate would increase the transition
t e m p e r a t u r e ; while Castro and Gueussier (1949) and Crussard et al.
(1956) have emphasized t h a t the effects of t e m p e r a t u r e , strain rate,

and triaxiality of the stress are equivalent. The importance of the
e n h a n c e d strain rate near a n o t c h has been emphasized by t t o l l o m a n
(1944) who estimates that the elastic strain rate in a n o t c h bend test
m a y be as much as l0 v times the strain rate in a normM tensile t e s t ;
this alone is sutficient to produce a change in the transition temper-
ature of the order of 100~c. When plastic flow occurs at the n o t c h the strain
rate will be even greater so t h a t a f u r t h e r increase in the transition temper-
ature will occur. T h o u g h a n u m b e r of studies of the initial yielding in
n o t c h e d specimens have been made (Green and H a n d y 1956, Crussard et al.
1956) and good agreement found between theoretical and e x p e r i m e n t a l slip
line fields (figs. 21 and 22, P1. 15) it has not as y e t p r o v e d possible to
m a k e a n y q u a n t i t a t i v e estimate of the m a x i m u m plastic strain rate
Fig. 21
p P
RI IO
S R R S
Theoretical slip line field for pure bending with a V notch. The central pivot
OPQTQPO remains rigid and the parts uf the bat" on either side rotate
about this (compare fig. 22, P1. 15). Green and Hundy (1956).
450 A. N Stroh on a
near a notch. There can be little doubt, however, on the general results,
firstly that near a notch the strain rate will be greatly increased, and
secondly that such an increased strain rate will be effective in increasing
the transition teml)erature.
§ 6. THE PROPAGATION OF CRACKS

6.1. Gril~th' 8 T h e o r y Applied, to Metals
]n a completely brittle material,, the condition that, a crack, once
formed, should extend is given by the Griffith's criterion, eqn. (3).
Orowan (1949, 1955) has considered the application of this to the brittle
fracture of metals. As it stailds eqn. (3) cannot be applied to a metal

since it assumes t h a t no plastic deformation at all has occurred ; in a
metal a thin layer adjacent to the fracture surface is found on examina-
tion to have deformed plastically eren in brittle fracture.
Plastic deformation mllst occur in polycrystals as the cleavage planes
in neighbouring crystals will not in general intersect along a grail
b o u n d a r y ; thus the cleavage surt~ces in different grains can link up
only by a t)recess of teari~lg at the grain boundaries which will involve
plastic deformation. Sn('h tearing has also been observed in single
crystals by Tipper and tIall (1953) (fig. 23, Pt. 15), when the path of
the crack changes fl'om one cleavage plane to a neighbouril~g one.
Orowan 1)oints out that when the plastic deformation is confined to a
region whose thickness is small compared to the length of the crack, the
plastic work done is t)roI)ortional to the area of the surface and its value
/J per ulfit area may be added to the surfa(~ energy ~. in G r i ~ t h ' s equation.
The intensity of the plastic distortion a t the fracture surface can be
estimaled from the diffuseness of the spots in back reflection x-ray
photographs; the eitective thickness of the deformed layer can be
found from ttm r~t¢ at which the spots in successive photographs become
sharper as more and more material is removed by etching away the
surface. In this way, ()rowan (1945) and Felbeck and Orowan (1955),
have found that the plastic work p in the brittle, fracture surface of a
low carbon st~cel was aboul 2 × l0 s ergs/cm* ; the effective thickness of
the deformed layer was 0.3 or 0.4 ram. Since this value is large com-
pared with y (of order l0 a ergs/cm2), y may be neglected and eqn. (3)
for the growth of the (;rack becomes
a - - ( E p / c ) 1t~. . . . . . . . . (26)
Felbeck a~l(t ()rowan have tested this equation by introducing cracks
ilito specimens of ship-I~late and measuring the stress needed to extend
them. Their results, shown in fig. 24 verify that the stress needed is
illversely proportional to the square root of the initial crack length, as is
l~redi(;te(l by eqn. (26). The value of p deduced from these results is
4.9x ll) (~ergs/cm 2, agreeing in order of magnitude with t h a t estimated
from x-ray observations.
Theory of the Fra~ure of Metals 451
In § 2.2 it was seen t h a t the Griitith condition for the growth of a

crack could also be obtained by requiring that the stress at about an atomic
distance from the tip of the crack should equal the molecular cohesion
of the material. Orowan (1955) has also considered what must be the
corresponding interpretation in the present case. We must now consider
macroscopic quantities only, since clearly the treatment will no longer
apply at distances lass than the thickness t of the plastically deformed
layer, i.e. 0.2 to 0-4 ram. At a distance t from the tip of the crack the
stress will be magnified by a factor o f about (c/t)1/~ ; for She crack of
fig. 24, this gives a stress of about 6000 kg/em ~, or about 40 tons/in.",
F~.
~
4000
4 . 9 - 106 e r 9 1 cm 2
3000 0
Mean
Stress
kg/cm 2
2OOO
Asymmetrical specimens of
plate PAD at room temoera?ure
I000
0
0 ! 2 3
Crock L e n g t h - cm
Variation of stress needed to propagate a crack, with crack lcllgth.

Felbeck and Orowan (1955).
which is of the same order as the observed cleavage strength of steel.

An indication as to the mechanism operating is afforded by Tipper's
observation (1945, 1957) t h a t the crack does not propagate continuously ;
ratlmr a number of small unconnected cracks are formed in grains ahead
of the main crack, and these subsequently join up (fig. 25, P1. 16). For
the crack to propagate by such a process of repeated reqnitiation of
fracture, it is understandable t h a t the macroscopic stress should have to
attain locally the observed cleavage strength.
The cracks in the individual grains might be produced by piled up
groups of dislocations at the ends of slip lines as has been considered in
§ 3; alternatively a crack in a previously fractured grain could serve
instead of the piled up group, for, as was remarked before, the stresses
round a crack and a slip line are very similar. In view of the high
speed at which a brittle crack propagates the latter seems more likely
as requiring less time to operate ; but in either case the same macroscopic
stress is needed_
Our picture of the propagation of a crack is then as follows. The
main crack produces locally, but on a macroscopic scale, a region of
stress concentration at its tip. In this region there are microeraeks,
due to cleavage within grains, ahead of the main crack. The stress
acting on these microeracks, will be magnified again to produce, now on
an atomic scale, a stress equal to the atomic cohesion of the material ;
in this way, the stress is raised sufficiently to initiate cleavage in other
grains. Finally the mierocracks will link together and join up with the
main crack by tearing, and it is during this process that most of the
plastic work deformation associated with the fraeture is done.
In addition to producing plastic deformation, the propagation of the
crack may also induce twinning as, for example, the Neumann lamellae
observed in iron (Tipper and Hall 1953). The stresses associated with a
moving crack have been studied by Bilby and Bullough (1954) with
special reference to the formation of twins ; they found that they could
give a consistent explanation of the formation of the twins and kinkbands
which Deruyttgre and Greenough (1954) had observed in the cleavage
of zinc single crystals. All such processes will result in ~he absorption of
some energy, and this must be added to the plastic work p introduced
above.
6.2. T h e Velocity of a Crack

Mort (1948) has considered the speed at which a Griffith crack will
sI)read. He points out that when the crack is moving the kinetic energy
of the material disturbed by the crack must be taken into account;
for a crack of length c moving with speed V through a material of density p,
the kinetic eflergy may be written as kpc~V2(~2/E 2, where k is a numerical
constant. Adding this term to the Griffith eqn. (2), the conditiofi
d W/dc~-O for energy balance gives the velocity of the crack to be
V-~(~r/4k)U~(E/p)l/2(1--ca/c) . . . . . . (27)
where c a is the Griftlth length
Co=4yE/~ra~ . . . . . . . . (28)
When the length of the crack is large compared with the length co,
cqn. (27) shows that the velocity will tend towards the limiting value
(Tr/4k)U2(E/p) 1/~ ; (E/p) u2 is here the velocity of the longitudinal waves in a
thin rod of the material. Roberts and Wells (1954) have evaluated the
constant k and so obtained the value of 0.38 (E/p) u2 for the limiting
velocity.
The treatment just considered strictly holds only when the velocity
is small as relati~dstic' effects which occur when the velocity of the
crack approaches that of sound have been ignored. An exact treatment
T a b l e 4. Measured Velocities o f Cracks
Velocity Vo-- v/(E/p) Velocity V

Investigator Material of longitudinal waves of crack V/Vo
ft/sec. ft/sec.
Sehardin and Struth (1938) Glass 17000 4650-5150 0.27-0-30

Fused Quartz 17000 7200 0.42
Hudson and Greenfield (1947) Steel 16500 3370 0.20
Barstow and Edgerton (1939, 1941) Glass 18000 5000 0-28
Smith and Ferguson (1950) Cellulose acetate 3700 1000-1375 0.27-0.37
Kennedy (1945) Steel 16500 4500 0'27 -.¢
Boodberg et al. (1948) Steel 16500 4600~600 0.28-0.40
Robertson (1953) Steel 165OO 6000 0.36
Bueche and White (1956) Silicone Rubber I 11-26 0.16.-0.38
Silicone Rubber I I
67
150 I 26-53 0-18-0.35
Irradiated polyethylene 380 150-300 0"38-0"78
Plexiglas 1600 330 0'21
Glass 19000 5000 0.26
Fused silica 24000 7300 0"32
O0
4t14 A.N. S t r o h on a
in this case presents a rather intractable proi~lem ; but assuming the

result t h a t the velocity of the crack tends to a limiting value independent
of the Griflith length, and hence of the surface energy, this limiting
velocity is easily seen to be just the velocity of the Rayleigh surface
waves.
For suppose the crack at any instant of time extends from A to B
(fig. 26) and consider the stress acting across AA 1 just ahead of the crack.
Fig. 26
CI
hi BI
To illustrate the spread of a crack.
When the crack moves from A to A 1 an amount of work is done which is

given by the product of the stress on AA1, and the displacement AA~ ;
this work must be equal to the surface energy of the new surface A~A 2.
But the limiting velocity of the crack is independent of the surface
energy, and so this surface energy may be taken to be zero. Hence,
since AA~ is not zero, the stress on AA t must be zero. Thus the propaga-
tion of the crack involves the surface A ~A C B B 1 deforming to A 1 A ~Cl B t,
the surface being unstressed throughout, and so is just a disturbance
moving on a stress-free surface. The velocity with which such a
disturbance moves is the velocity of the Rayleigh waves.
I f Poisson's ratio is taken to be ¼, the velocity of the Rayleigh w;~ves
(Love 1927) is 0.919~/(G/p)---O.58v/(E/p). This value is somewhat
higher than t h a t obtained by Wells and R,oberts quoted a b o v e ; the
difference is probably due to the neglect of relativistic effects by t h e ~
authors.
A number of measured values are collected in table 4. The values
are seen to be all a little less t h a n the theoretical va[ue ; this is not alto-
gether unexpected as any dissipative process must reduce the crack
velocity. Moreover, Tipper (1957) suggests t h a t cracks p r o p a g a ~ in
steel in a jerky manner, coming to rest and then moving forward again ;
such a process would give a mean crack velocity rather less than the
limiting velocity.
6.3. The Propagation Transition

A crack can propagate in a brit, tle manner only if the dislocation
sources near which it passes are locked so that plastic flow"is inhibited.
The looking of' the sources m a y be due to impurity atoms. At higher
temperatures thermal fluctuations will be able to free the dislocations
so that the material is ductile, aad we obtain a tra~lsiti(m temperature
for the propagation of the crack in tile same way as we obtained one
ibr the initiation. The only cha~lge needed i~a ~he trcatment of § 5.1 is
t h a t t is now the time the crack takes to pass near a F r a n k - R e a d source :

this will be of order r/V, where r is the distance of closest approach of the
crack. Such transition temperatures have been measured by Robertson
(1951, 1953). Specimens of ship plate were subjected to a uniform
tensile stress, and a temperature gradient in a direction perpendicular
to the stress axis was maintained by heating at one end and cooling
with liquid nitrogen at the other. A cleavage crack was started from a
saw cut in the cold end by a small explosion. The crack travelled some
distance in the direction of increasing temperature, and the temperature
at the point where it came to rest was noted.
I f any plastic flow occurs, this will involve the expenditure of energy,
and so will tend to reduce the speed of the crack. On the other hand if
the crack slows down, the disloco,tions will be subjected to the stress field
of the crack for a longer time and so more plastic flow will occur t. The
process is thus a cumulative one, and if plastic flow starts it m a y build
up rapidly- until the crack is no longer able to propagate ; in this case the
material is ductile.
The way in which the velocity of the crack determines whether it
shall propagate in a brittle or a ductile manner is very clearly shown in
Felbeck and Orowan's experiments (1955) in which a crack which had
been introduced artificially in a low carbon steel was caused to propagate.
In fig. 27, P1. 16, 1 is the initial position of the crack. When a load was
applied at first considerable plastic deformation occurred at the tip of
the crack which started to propagate as a ductile crack 2. After a short
run, this changed into a brittle crack 3 which spread at high velocity.
Owing to an inability of the testing machine to maintain the load during
the rapid extension of the specimen, the load dropped drastically (to
about half its initial value) and the crack stopped in the material. On
increasing the load again the process could be repeated, the crack starting
as a ductile crack 2 and then becoming brittle 3. The contraction of
the thickness of the plate due to plastic deformation when the crack is
ductile can be seen in the photograph. The only simple explanation
of the behaviour of the crack is that brittle propagation cannot occur
until the velocity of the crack has attained some minimum value.
A simple model which exhibits some of the features observed has been
proposed by Gilman (1956). Equation (27) for the velocity V of a crack
may be rewritten
r = V0( 1 -- Us/UB) 1/2 . . . . . . (28)
where V0 is the limiting velocity of a truly brittle crack, U s is the surface
energy and U E the elastic energy of the crack. I f there is a n y plastic
deformation at the fracture surfaces, then U s must, as in § 6.1 include
the plastic work done. We now assume that the amount of plastic work
The relation between the amount of plastic work done and the velocity
of the crack has been considered by Hall {1953}.
456 A . N . Stroll an a
is proportional to the time for which the stress is applied, and so inversely
to the velocity V of the crack ; t h a t is, we write
U , ~ U ',/ V ,
eqn. (28) then becomes
V = Vo(1--U',/UEV)I/2 . . . . . . . (29)
I f U's~(2/3~c/3)U~Vo, eqn. (29) is not satisfied by any real value of V ;
that is, if plm,tic flow occurs too copiously, the propagation of a crack
in a quasi-brittle manner is impossible. By assuming U', to be an
increasing function of the temperature, we thus obtain a critical tem-
perature above which the brittle propagation of a crack does not occur.
For U ' , ~ ( 2 / 3 ~ / 3 ) U E V o, eqn. (29) is satisfied by a vMue of V between V0

and Vo/~/3, depending on the value of U ' , / U E ; this represents the high
velocity brittle crack which is obtained when plastic flow is restricted.
Thus far the model reproduces fairly well the observed features of crack
propagation. However under the conditions in which brittle propagation
is possible, eqn. (29) is also satisfied by a second and smaller value of V ;
for small values of U',, this value of V is approximately U'8[UE, and
becomes very small if plastic flow is much restricted. Such a low velocity
brittle crack is not observed.
I t should be noted t h a t the assumption on which the model is based,
t h a t the amount of plastic work done is proportional to the duration of
loading, is certainly not true in general. This is shown by the experi-
ments of Clark and Wood (1949) on delayed yielding in iron. I f a
fixed stress is applied to the material, at first little plastic flow occurs,
but after a certain time catastrophic yield sets i n ; this delay time is
just the time needed for thermal fluctuations to release the dislocations
from their locking impurities.
6.4. The Effect of Transverse Stress

Robertson (1953) has found t h a t the temperature at which his cracks
were arrested was independent-of the stress transverse to the crack
provided this stress was above some minimum value. For stresses
below this critical value, the transition temperature dropped rapidly
with decreasing stress (fig. 28).
A suggestion of Fisher's (1954) indicates a reason why such a transverse
stress may be needed. The cleavage plane will be threaded by dislocations,
and each time the crack crosses a screw dislocation a step in the cleavage
plane will be produced ;~ this will increase the area of the fracture surface
and so of the surface energy which must be supplied. I f there are p
dislocations per unit area, the surface energy will be increased by an
amount ~pc~b? ; adding this term to the right of eqn. (2), we see t h a t the
This is probably the origin of the ' river pattern ' markings seen on the
cleavage surface (fig. 29, P1. 17).
stress needed to spread the crack does not become vanishingly small for
large crack, b u t tends to the limiting value o f
~= {~Cpb~/~(l--0}- ~ . . . . . . (a0)
W i t h p ~ 1 0 ~° cm -~, this stress becomes 6 × 108 dynes/cm 2 or 4 tons/in. ~,
which is the order o f the critical stresses observed b y Robertson. I f
Vhe stress is only slightly above the value (30), t h e n the crack is likely
Fig. 28
16
,o
,Z12
w
r~
~IC
r~
z
I
~6
0..--- I
4
-30
-20 IO O IO 20
ARREST TEMPERATURE,°C.
Dependence of the arrest temperature of a crack on transverse stress.
Robertson (1953).
to be held up whenever it meets a density of dislocations greater than

the a v e r a g e ; for example, when the crack crosses a low angle mosaic
b o u n d a r y . W h e n this happens the delay m a y give an o p p o r t u n i t y for
plastic flow to occupy and so to p r e v e n t f u r t h e r spread of t h e crack.
This plastic flow can be avoided only if the t e m p e r a t u r e is sufiiciently low ;
hence a decrease in stress will be accompanied b y a fall in the t e m p e r a t u r e
of arrest.
§ 7. DUCTILE FRACTURE
I f a ductile specimen is tested in tension three stages in its fldlure
are observed.
(i) The specimen necks. F r o m this point on further deformation is
confined to the neighbourhood of the neck.
(ii) A crack is formed in the centre of the neck normal to the stress
axis (fig. 30, PI. 17).
(iii) Final separation occurs b y shear at an angle of 45 ° to the stress
axis,
458 A.N. S t r o h on a
The maximum load occurs when necking first starts ; the corresponding
stress is the ultimate tensile strength ' of the material. This, however,
has no direct relationship with fracture but is the point at which uniform
elongation becomes unstable. The strain at which this happens can be
determined by a simple geometrical construction from the stress-strain
curve as was shown by Consid~re as early as 1885; good accounts of
this topic (e.g. Orowan 1949) are available, and it will not be discussed
hero.
The final stage in failure (iii), is also not a true fracture but is due to
localized slip. The explanation of this was first given b y Zener (1948)
who has also studied a number of other examples of the localization of
slip. In general if the strain rate is sufficiently high the deformation
will be adiabatic and the resulting rise in temperature produces a softening

of the material which acts in opposition to the work hardening. Thus,
as Zener and HoUoman (1944) have noted, the stress-strain curve has a
maximum. For strains greater than that corresponding to the maximum
stress, a homogeneous deformation becomes unstable; the shear strain
is localized to the vicinity of a surface which becomes very hot and thereby
loses most of its resistance to furtber deformation. Zener (1948) has
demonstrated this in experiments on punching a steel plate (fig. 31, PI. 17).
In ' s t a t i c ' punching deformation occurs generally, but in dynamic
punching it is confined to the immediate vicinity of the punching surface.
An indication of the high temperatures which are obtained in such
dynamic experiments is given by the formation of martensite at the
surface.
To see how the cone of a cup and cone fracture may be produced in
this way, it must be remembered that, in the final stage of failure, plastic
deformation is confined to a small region in the neck ; the strain rate
here must be large to maintain the average strain rate in the specimen,
and so the conditions for adiabatic strain will prevail. When an internal
crack has formed plastic strain will be favoured in the two shaded regions
in fig. 32; under adiabatic straining the plastic strain will be entirel~
confined to these two regions, and the specimen will separate by slipping
apart here.
The formation of the internal crack (stage (ii)) which is the only stage
involving a true fracture, has still to be considered. Because the forma-
tion of the crack has been preceded b y necking, the stress system
prevailing is rather complicated. In most experimental investigations
the true stress has not been measured but rather the 'conventional
stress ' (i.e. the load divided by the original cross sectional area) ; though
the latter may be sufficient for technological applications it does not
give much information on the fundamental proces~s involved. One
set o f experiments in which necking has been fully allowed for and the
true stress determined, is due to Petch (1956 a) ; he finds that ~the maxi-
mum stress in the ductile fracture of iron has the same linear dependance
on d -1/~, where d is the graiv diameter, as was found for brittle fracture
(fig. 6) ; and ~lso the slopes of the lines are the same in both eases. This
seems a clear indication t h a t the same basic mechanism must operate,
namely t h a t cracks arise from dislocations piled up on the grain boundary.
Mott (1953) had previously suggested that this would be the case; in
a ductile material, however, the cracks would be stopped by plastic
flow from growing to any size, but as the deformation proceeded the
number of cracks would increase ~ ultimately these cracks would link
up to form a macroscopic crack.
Fig. 32
Localization of slip near an internal crack. Zener (1948).
Stroh (1957) has suggested t h a t the changes in density and electrical

resistance observed by Clareborough et al. (1955, 1956) i~1 nickel after
heavy cold work can be attributed to microcracks formed in this way.
However Seeger (1956) disagrees with this suggestion and maintains that
dislocations are sufficient to account for the effects, but to do so he finds it
necessary to attribute a rather large electrical resistance to a sta~king fault.
The fact t h a t the fracture strength is directly related to the grain size
iinplies that the grain boundaries are the only obstacles strong enough
to withstand the large piled,up groups needed for fracture. It is note-
worthy that single crystals of metals with a face-centred cubic strucfure
do not fracture but separate by drawing down to a double wedge. (Single
crystals of other metals do, however, fracture ; at present the mechanism
by which this occurs has not been established.) In accord with this
Stroh (1956) has estimated that one type of obstacle considered important
in work hardening--the sessile dislocation of Lomer (1951) and Cottrell

(1952)--is t o o weak by an order of magnitude to withstand the large
piled up group needed for fracture.
Figure 6 shows that the intercept a 0 varies considerably with the
conditions under which fracture occurs. (The change shown is probably
due not only to the different temperatures, but also to the fact that
different amounts of plastic strain preceded fracture in each case.) % has
been interpreted in § 4.2 as the stress needed to move a dislocation,
and as such it will be affected b y all the factors determining the flow
stress in a work hardened material t. No theoretical treatment of % in
the ductile case has as yet been given, while experimentally only the two
values shown in fig. 6 have been determined; very little can be said,
therefore, about the variation of %. It m a y be surmized, however, that

most processes effecting the fracture strength (for example the effect
of prior plastic deformation which has been studied by Ripling and
Baldwin (1951) under the name of rheotropie embrittlement) will do so
by producing a change in the value of %.
ACKI~OWLEDGMENTS
My thanks are specially due to Professor N. F. Mott, F.R.S., for his
encouragement in the writing of this article, and for valuable discussions
and comments. I also wish to thank Dr. A. R. Entwisle, Dr. E. 0. Hall
and Dr. P. Pratt for most helpful discussions, and Dr. J. Nutting for
permission to reproduce fig. 4, P1. 14.
APPENDIX A
T h e E n e r g y o f a Crack
The change in the elastic energy which occurs when a crack is formed
near a piled up group (i.e. along the plane OP of fig. 2) is considered here.
Equation (5) for the stress on OP m a y be written
a = A / r 1/~. . . . . . . . . (A1)
where A =:L1/2f(O)=~a L 1/2 sin ~ cos 1~. . . . . (A2)
when the value off(0) is inserted. If a crack of length r is formed along OP,
but the stresses are not allowed to relax, (A1) gives the normal stress on
the surface of the crack. Now take as origin of coordinates the centre
of the crack, y axis in the plane of the crack, and unit of length, one half
the length of the crack. The stress (A1) on the surface of the crack
becomes a ( y ) : A ( 1 - - y ) -1/~ . . . . . . . . (A3)
I f the stress (A3) is relaxed the change in the elastic energy (Stroh 1955 a)
is I"1
W~= .! d, . . . . . (A,)
J0
t Recent accounts of the theory of work hardening have been given by
Friedel (1957) and Seeger et al. (1957).
where r '=~--1 ~(/x sin O) d O , . . . . . . . (A5)

a-½n
and ~
T"--~ -1 ~(/~ sin O) sin 0 dO. . . . . . (A6)
Evaluating these integrals, we obtain

W~= --8( I--v)A2/~G,
or in ordinary units of length
W~=--4(1--u)A2cffrG . . . . . . . (A7)
The value of A is a maximum giving the maximum change in energy when
cos0=½ (or 0=70.5 °) then A = ( 2 / X / 3 ) % L 1/2 and (AT) becomes
W,==--16(1--v)%ULc/3rrG . . . . . . . (AS)
For this value of 0 it may be shown that the shear stress on OP is zero
so that the shear stress contributes nothing to the energy, and (AS) gives
the total change in the elastic energy on forming the crack.
Adding the surface energy 2cy the total change of energy on forming
the crack is
W = -- 16( 1--v)%'~Lc/3~rG@ 2cy
and equating this to zero we obtain the condition for the ibrmation
of the crack to be given by eqn. (7).
APPENDIX B
The Initial Growth of a Crack"
We shall consider here whether, after the crack has been initiated,
it will spread catastrophically through the material or come to rest at a
finite length. The material is assumed brittle so t h a t plastic flow will
not prevent the growth of the crack.
Suppose when the crack has formed n' dislocations enter it so that
there is a wedge n'b sln 0, where 0 is the angle between the crack and the
slip plane (fig. 33,) tending to open the crack. The component of
applied stress normal to the crack is a t and that parallel to the crack %.
I f the crack has length c, its energy.is
n'~b~G , 4 R ,
W-- ~-]----~) tog ~ + 2 y c - - ~Tr(l--~,)(a,.21+%2)c2/G--½n'bal sin O, (BI)
the terms represent (i) the elastic energy associated with the wedge
deformation (Stroh 1954), (ii) the surface energy, (iii) the elastic energy
of the crack in the applied stress field, and (iv) the energy due to the
increase in volume on opening tim crack. I f lengths c 1 and cu are defined
by
cl:n'%2G/87r(l--u)~ . . . . . . . (B2)
and c2~8),G/(al2 +%u)rr(1--u) . . . . . . . (B3)
the eqn. (Bl) becomes
W : 2 7 { c 1 log (4R/c)+c--c2/2e,, -2(cl/c2)l/2al(ct12-JCaz2) -112 G sin 0}. (B4)
462 A . N . S t r o h on a
The s t a t i o n a r y values o f W occur when

C2~- (1-- 2(CI/C2)l/20"l(q12-~-q22)--l/2 sin O}c~c--t-clc2-~O. (B5)
E q u a t i o n (B5) has either two positive real roots, or no real roots, i f
there are two real roots, the smaller gives a m i n i m u m value o f the e n e r g y
W, and so a Crack of this length will remain in stable equilibrium in the
material. I f there are no real roots no position o f equilibrium o f the
crack is possible, and so it will grow indefinitely. The critical, case
occurs w h e n the roots of (BS) are equal ; this will h a p p e n if
2(Cl/%) (1 + al((r19"-4- au~)-112 sin 0)= 1,
or using (B2) a n d (B3), if
n'b{(al2A- a2")l/2A- al sin 0 } = 4 ~ . . . . . . (B6)

T h e condition for the initiation of the crack, eqn. (9), rib%= ~7r8~, m u s t
also be satisfied, where n is the t o t a l n u m b e r of dislocations in the piled
up group and %, the resolved shear stress o n the slip plane. H e n c e
the f o r m a t i o n of the crack will result in i m m e d i a t e fractions of the
material if
n'/n-~32%/3rr~{(%~--}-%~)l/2+a 1 sin 0)} . . . . (B7)
Fig. 33
Slip line and initial crack in a tensile specimen.
Now suppose the applied stress is a uniaxial tension a. The most

f a v o u r e d slip lines are those at an angle of 45 ° to the tensile axis, so
. . . . . . . . . . (Bs)
T h e calculations of S t r o h (1954) indicate t h a t the crack should f o r m a t
an angle of a b o u t 0-=cos -1 ½--~70.5 ° to the slip plane ; in this case we find
al:½(1--4/9V/2)~:=0.186~, and a2:l~s-a. (B9)
Itence combining (B7), (B8) and (B9),

n'/n=O.89 ; . . . . . . . . (Bt0)
this means that there are sufficient dislocations in the piled up group to
cause the cr~ck to spread right through the material.
i n the above treatment we have neglected the stress due to those
dislocation of tile piled up group which h~ve not entered the crack;
since these will tend to spread the crack, consideration of their effect
can only reinforce our conclusion that the crack will be unable to find
an equilibrium position and so will spread catastrophically.
[~EFERENCES
ALLEN, N. P., HOPKINS, B. E., and McLEN~o~, J. E., 1955, Proc. roy. Soe. A,
234, 221.
AaDLEY, G. A., and COTTRELL, A. [-[.,1953, Proc. roy. Soc. A, 219, 328.
BAEYERTZ, M., CRAIO, W. F., and BuMPs, E. S., 1949 Trans. Amer. Inst. rain.
(metall.) E~grs, 185, 48l.
BARSTOW, F. E., and EDGEI~TON, H. E., 1939, J. Amer. ceram. See., 22, 302 ;
1941, Ibid., 24, 131.
BEC~rI'OLD, J. H., 1953, Trans. Amer. Inst. rain. (metall.) Enffrs, 197, 1469;
1954, Trans. Amer. Soc. Metals, 46, 1449 ; 1955, Acta Met., 3, 249.
BEOIITOLD, J. H., and SHEWMON, P. G., 1954, Trans. Amer. Soc. Metals, 46,
349.
BILBY, B. A., and BtrLLOUOR, R., 1954, Phil. Mag.o 45, 631.
BILBY, B. A., and ENTWISLE, i . R., 1956, Acta Met., 4, 257.
BoODBERG, A., DAVIS, H. E., and PA~K~R, E. 1~., 1948, Weld. J., 27, 186 s.
BeaN, M., and FURTIt, R., 1940~ Prec. Camb. phil. See., 36, 454.
B u E c ~ , A. M., and WHITE, A. V., 1956, J. appl. Phys., 27, 980.
CASTRO, 1%., and GUEUSSIE~, A., 1949, Rev. Met., 46, 517.
C~VROHMAN, A. T., 1955, Acta Met., 3, 22.
CLARK, D. S., and Woo]), D. S., 1949, Prec. Amer. Soc. Test. Mater., 49, 717.
CLAaEBROUOK,L. M., HAaGR~AV]~S,M. W., and WEST, G. W., 1955, Prec. roy.
See. A, 232, 252; 1956, Phil. Mug, 1, 528.
CONSID~.RE, A., 1885, Ann. des ponts et chaussdes, 6, 9.
COTTI~ELL, A. H., 1948 Report of Bristol Conference on the Strength of Solids
(London : Physical Society), p. 39 ; 1949, Progress in Metal Physics I
(London: Pergamon Press), p. 77 ; 1952 Phil. Mag., 43, 645; 1953
Dislocations a ~ Plastic Flow on Crystals (Oxford: University Press).
COTTRELL,A. H., and BILBY, B. A., 1949, Prec. phys. Soc. Lend. A, 62, 49.
CRAOK~TELL,A., and PETC~, N. J., 1955, Acta Met., 3, 186.
CRUSSARD, C., BORIONE, R., PLATEAU, J., IV[ORRILLON, ~IT.,and MARATRAY, F.,
1956 J. Iron St. Inst., 183, 146.
DERUYTT~IRE, A., and GREE~O~rGH, G. B., 1954, Phil. Muff., 45, 624 ; 1955, J.
Inst. J~let., 84, 337.
ESHELB¥, J. D., FRANK, F. C., and NABARRO, F. R. N., 1951, Phil. Mug.,
4,2, 351.
FELBECK, D. K., and OROWAN, E., 1955, Weld. J., 34, 570 s.
FISREa,.J.C., 1954, Unpublished communication to Birmingham Conference;
1955, Acta Met., 3, 109.
FRANK, F. C., 1950, Symposiwm on the plastic deformation of crystalline solids
(Pittsburgh: Carnegie Institute of Technology and Office of Naval
Research), p. 89.
FEIEDEL, J., 1956, Les Dislocations (Paris: Gauthier-Villars) ; 1957, Proc.
roy. Soc. A. To be published.
GIBBONS, D. F., and COTTRELL, A. H., 1947, Nature, Lond., 162, 488.
GmMAN, J. J., 1956, J. appl. Phys., 27, 1262.
GREEN, A. P., and HUNDY, B. B., 1956, J. ~lech. phys. Solids, 4, 128.
GREENWOOD, G. W., and QUARREL,A. G., 1954, J. Inst. Met., 82, 551.
GR~FEIT~, A. A., 1920, Phil. Trans. A, 220, 163.
HALL, E. O., 1951, Proc. phys. Soc. Lond. B, 64, 747 ; 1953, J. Mech. phys.
Solids, 1, 227.
HAUSER, F. E., LANDON, P. R., and DORN, J. E., 1956, Trans. Amer. Inst.
rain. (metall.) Engrs, 206, 589 ; 1957, Trans. Amer. Soc. Metals (to be
published).
HESLOP, J., and PETCtf, N. J., 1956, Phil. Mag., 1, 866.
HILL, R., 1949, Quart. J. Mech. appl. Math., 2, 40.
HODGE, J. M., MANNING, R. D., and REICHOLD, H. M., 1949, Trans. Amer.
Inst. rain. (metall.) Engrs, 185, 233.
HOLLOMAN, J. H., 1944, Trans. Amer. Inst. rain (metaU.) Engrs, 158, 298.
HUDSO~~, G., and GREENFIELD, M., 1947, J. appl. Phys., 18, 405.
INGLIS, C. E., 1913, Trans. Instn nav. Archit., Lond., 55, 219.
JACQUET, P. A., 1954, Acta Met., 2, 752.
KENNEDY, H. E., 1945, Weld. J., 10, 597 s.
KOC~ENDLRFER, A., 1954, Arch. Eisenh~tenw., 25, 351.
LOMER, W. M., 1951, Phil. Mag., 42, 1327.
LOVE, A. E. H., 1927, Theory of Elasticity (Cambridge : University Press).
Low, J. l~.~ 1954, Trans. Amer. Soc. Metals, 46 a, 163.
LUDWIK, P., 1926, Z. MetaUk., 18, 269.
LUDWIK, P., and SCHEV, R., 1923, Stahl u. Eisen, Di~seldorf, 43, 999.
McADAM, D. J., GEXL, G. W., and MEBS, R. W., 1947, Trans. Amer. Inst. rain.
(metall.) Engrs, 172, 323.
McADAM, D. J., and MEBS, R. W., 1943, Proc. Amer. Soc. Test. Mater., 43,
661.
MACGREGOR,C. W., and GROSSMA~,N., 1952, Weld. J., 31, 20.
MESNAGER. /~., 1906, Intern. Ass. Testing Materials, Brussels Cong., Raport
non.o]fic. No. A. 6f., p. 1.
MOTT, N..F., 1948, Engineering, Lond., 165, 16 ; 1952, Imperfections in Nearly.
perfect Crystals, ed: W. Shockley (New York: John Wiley), p. 173:
1953, Proc. roy. Soc. A, 220, 1 ; 1956, J. Iron St. Inst., 183, 233.
MOTT, N. F., and NABARI~O, F. R. N., 1948, Report of Bristol Conference on
Strength of Solids (London : Physical Society), p. 1.
OROWA~, E., 1933, Z. Phys., 86, 195 ; 1945, Trans. In~tn Engr~ Shipb. Scot., 89,
165 ; 1949, Rept. Progr. Phys., 12, 185 ; 1955, Weld. J., 34, 157 s.
OROWAN, E., NYE, J. F., and C~R~S, W. J., 1945, Theoret. Res. Rep. No. 16/45,
Arm. Res. Dep., M.O.S., London.
PETCn, N. J., 1953, J. Iron St. Inst., 174, 25 ; 1954, Progress in Metal Physics V
(London: Pergamon Press), p. 1 ; 1956 a, Phil. Mag., 1, 186; 1956 b,
Ibid., 1, 331.
PETCH, N. J., and STABLES, P., 1952, Nature, Lond., 169, 842.
PRANDTL, L., 1923, Z. angew. Math. Mech., 3, 401.
READ, W. T., 1953, Dislocations in Crystals, (New York: McGraw Hill).
RINEBOLT, J. A., and HARRIS, W. J., 1951, Trans. Amer. Soc. Metals, 43, 1175.
R~LING, E. J., and BALDWIN,W. M., 1951, Trans. Amer. Soc. Metals, 43, 778.
ROBERTS, D. K., and WELLS, A. A., 1954, Engineering, Lond., 178, 820.
I~OBERTSON, T. S., 1951, Engineering, Lond., 172, 445 ; 1953, J. Iron St. Inst.,
175, 361.
SACK, 1¢. A., 1946, Proe. phys. Soc. Loud., 58, 729.
SCHARDI~, H., and STRUTII,W., 1938, Glastech. Ber, 16, 219.
SEEGER, A., 1956, Canad. J. Phys., 34, 1219.
SE~.(~ER, A., DIEHL, J., MADE•, S., and REESTOCK, H., 1957, Phil. Mag., 2, 323.
SMI~, H. L., and FEEGtTSO]~, W. J , 1950, N.R.L. Progress Report.
SMITH, R. L., and RUTHERFORD,J. L., 1957, Trans. Am(~r. Inst. rain. (m~.tall.)
Engrs., 209, 478.
STROH, A. N., 1954, Proc. roy. Soc. A, 223, 404 ; 1955 a, lbid, 232, 548 ; 1955 b,
Phil. May., 46, 198 ; 1956, Ibid., 1, 489 ; 1957, Ibid., 2, 1.
TAYLOR, G. I., 1946, J. Instn ciw Engrs, 26, 486.
TAYLOR, J. W., 1954, Metallurgia, 59, 161.
TIPPER, C. F., 1945, Report of Conf. on Brittle Fracture in Steel Plates, Cambridge
(British Iron and Steel Res. Assoc.), p. 24; 1957, J. Iron. St. Inst.,
185, 4.
TI-eP~'.E, C. F., and H,~L, E. O., 1953, J. Iron St. Inst., 175, 9.
TE~Y, R., and KRAVS, S., 1936, Nature, Lond, 138, 331 ; 1937, Ibid., 139, 30.
UL~O, J., and BAST~.NAIRE, F., 1953, Rev. Star. appl. 1, 45.
VANDERBECK, R. W., WILDE, H. D., LII~DSA¥, R. W., DANIEL, C., 1953,
Weld. J., 32, 325 s.
vo~ MIs~.s, R., 1928, Z. angew. Math. Mech., 8, 161.
W ~ , H. L., and COTTRELL, A. H., 1950, Proc. phys. Soe. Lond. B, 63, 339.
W ~ , H. L., and HENDERSON, F., 1953, Proc. phys. Soc. Lond. B, 66, 515.
Wrm~A~, F., and STEPANOV, W , 1939, J. Tech. phys. U.S.S.R., 9, 1070.
ZE~ER, C., 1948, Fracturing of Metals (Amer. Soc. Metals), p. 3.
ZENER, C., and tfOLLOMA~, J. H., 1944, J. appl. Phys., 15, 22.
ZWmKY, F., 1923, Z. Phys., 24, 131.
A. N. STROH Phil. Hag. Suppl. Vol. 6, No. 24, PI. 14.
Fig. 3
Piled up groups of dislocations in polycrystalline a-brass. Jacquet (1954).
Fig. 4
Sllp lines in iron, Direct carbon replica. × 25,000. Nutting and Brandon
(private communication).
A. N. STROH Phil. Mag. Suppl. Vol. 6, No. 24, PI. 15.
F~. 22
Deformation pattern observed after pure bending. Green and Hundy (1956).
Fig. 23
Section of a cleavage plane in an iron ery~tal )<200. Tipper and Hall (1953).
A. N. STROH Phil. Hag. Suppl. VoL 6, No. 24, PI. 16.
Fig. 25
Typical groups of microcracks ahead of the main crack in mild steel plate
× 700. Tipper (1957).
l~ig. 27
The propagation of a crack in ship.plate showing charges from ductile to

brittle behaviour, Felbeck and Orowan (1955).
A. N. STKOH Phil. Mag. Suppl. Vol. 6, No. 24, PI. 17.
Fig. 29
' River pattern ' marking on pearhte free steel fractured at --196°c.
Crussard et al. (1956).
Fig. 30
Section of a specimen showing internal crack formed during a tensile test.

Ludwik (1926).
Fig. 31
Experiments on punching. Zener "(1948).

Stroh1957 PDF

Загружено:

Сведения о документе

Исходное описание:

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Stroh1957 PDF

Загружено:

Авторское право:

Доступные форматы

This article was downloaded by: [University of Pennsylvania]

On: 12 June 2013, At: 07:14

A theory of the fracture of metals

To link to this article: http://dx.doi.org/10.1080/00018735700101406

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-

A Theory of the Fracture of Metals

§ 3, THE BRITTLE FRACTURE OF METALS.

§ 4. T ~ E LOWER YIELD POINT.

§ 6. THE PROPAGATION OF'CRACKS.

APPENDIX A : T h e energy of a crack.

observed in practice. The ideal strength is easily estimated by a simple

Typical values of the constants for a metal are E : 1 0 1 2 dynes/cms,

2.2. Gri~th' s Theory

§ 3. THE BRITTLE FRACTURE OF ~ETALS

3.1. Piled Up Group~ of Dislocations

_l_ _l_ _i_ _a_ I

Piled up group of disloca.tions.

N e a r a piled up g r o u p t h e stresses will be large : their n a t u r e is m o s t

3.2. The Prod~ctio.n of n Crac]c

g r o u p of dislocations will be sufficient to generate a crack. T h e v a l u e

As an indication of the magnitude of n suppose fracture occurs under a

ductility is very slight, there is some evidence to show t h a t fracture is

3.3. The Growth of the Crack

dislocation represents a half plane of atoms wedged between two complete

A dislocation with Burgers vector 3b, showing an incipient crack.

3.4. C o m p a r i s o n with E x p e r i m e n t a l Results

II,/'d, iN. "~

The experimentally determined values o f the slope K o f the line

Table 1. Observed and Calculated Values of c o n s t a n t K

Fe 1600 1.9 × l0 s 2-1 × iO s

~ N o dircct determination of 7 for molybdmmm is available, and the

3.5. Hydrogen Embrittlement

concentration of hydrogen will be greatest near the grain boundaries;

Table 2. The Effect of Duration of Loading on the Hydrogen Embrittle-

I f the whole of the effect of the hydrogen can be attributed to a change

I f the adsorption takes place at temperature T from a diatomic gas

h y d r o g e n m a y be c a l c u l a t e d ; for a grain size corresponding to

Lowering of surface energy of iron by hydrogen adsorption. Petck (1956 b).

3.6. Slip near a Piled Up Group

ature is unlikely to play any appreciable part in its production; it is

(i) Cottrell locking of dislocations

is to occur. The Cottrell locking will also help produce fracture in

(ii) Geometrical limitations on slip

On this view then the brittle behaviour of hexagonal metals depends

§ 4. THE LOWER YI]~LD POINT

to the stresses r o u n d a piled up group ; cleavage, to the tensile stress,

one specimen to another. We therefore suppose t h a t there will occur

of G b / l , where 1 is the tength of the source, is needed. I t is clear from

or including a constant frictional stress % as in § 3.4, the lower yield

4.1. T h e Stress Needed to Move a Dislocation

Effect of the concentration of Cq-N on a0, Cra,cknell and Petch (1955).

§ 5. THE BRITTLE-DucTILE TRANSITION

are exceptional. Thus the results of such a test m a y be represented

to a probability law depending on the temperature, so t h a t the two

Now according to the ideas we have developed, the dislocations are

a c o n s t a n t of the dimensions of a frequency. H e n c e t h e m e a n time~

§ 4 led to the conclusion t h a t the F r a n k - R e ~ l sources which were

involved t h a t little weight can be a t t a c h e d to the final result.

5.2. Effect of Strain Rate

to the stress pihng up the dislocations ; t h a t is t o the part a l - - a 0 of

1/To=--(k/U) log ~ + C . . . . . . . (2a)