This article was downloaded by: [Moskow State Univ Bibliote]

On: 21 October 2013, At: 05:13

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

Philosophical Magazine
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/tphm19

Steady-state diffusional creep

a b
F. R. N. Nabarro
a
Cavendish Laboratory , Cambridge
b
University of the Witwatersrand , Johannesburg
Published online: 20 Aug 2006.

To cite this article: F. R. N. Nabarro (1967) Steady-state diffusional creep, Philosophical

Magazine, 16:140, 231-237, DOI: 10.1080/14786436708229736

To link to this article: http://dx.doi.org/10.1080/14786436708229736

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information
(the “Content”) contained in the publications on our platform. However, Taylor
& Francis, our agents, and our licensors make no representations or warranties
whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and
views of the authors, and are not the views of or endorsed by Taylor & Francis. The
accuracy of the Content should not be relied upon and should be independently
verified with primary sources of information. Taylor and Francis shall not be liable
for any losses, actions, claims, proceedings, demands, costs, expenses, damages,
and other liabilities whatsoever or howsoever caused arising directly or indirectly in
connection with, in relation to or arising out of the use of the Content.

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.
Terms & Conditions of access and use can be found at http://www.tandfonline.com/
page/terms-and-conditions
 Steady-state Diffusional Creep

By F. R. N. NABARRO
Cavendish Laboratory, Cambridget

[Received 16 March 19671

ABSTRACT
The theory of steady-state diffusional creep is developed for a material in
which the grains are large and the dislocations form the principal sources and
Downloaded by [Moskow State Univ Bibliote] at 05:13 21 October 2013

sinks for vacancies. Under a given applied shear stress p there is an equili-
brium length of link in the dislocation network, only slightly longer than that
link which provides a dislocation source under the applied stress. The
dislocation density increases by the operation of Bardeen-Herring sources,
and decreases as dislocations of opposite sign climb towards one another and
annihilate. When diffusion occurs predominantly through the bulk of the
crystal, the rate of strain is proportional to p s . At lower temperatures,
where diffusion occurs predominantly along the cores of dislocations, the rate of
strain is proportional to p6.

9 1. INTRODUCTION
INthe ordinary theory of creep by the diffusion of single ions (which we take
to occur by the vacancy mechanism), we consider (Nabarro 1948, Herring
1950) a polycrystal having grains of radius L , acted on by a shearing stress p .
We take the atomic volume to be Q = b3, and the self-diffusion coefficient of
the crystal to be D. Then, as a result of the presence of the stress, the
chemical potential p of a vacancy on a grain boundary which is subjected to
normal tension exceeds that of a vacancy on a boundary which is subjected
to normal pressure by 2pQ. Vacancies are thus subjected to a gradient of
chemical potential of order ZpQIL, which produces a flux of vacancies and a
counter-flux of matter of volume about 2DpQIkTL across unit area in unit
time. This produces an equal normal velocity of the grain boundaries, or a
strain rate i of order 2DpQlkTL2. Herring’s quantitative estimate for
equiaxed grains which can slide freely over one another is :
gG = 5DpQ/kTL2. , . . . , . . (1)
Herring showed that a t low stresses the dislocations in the crystal
should not act as efficient sources or sinks for vacancies, and there is
experimental evidence (e.g. Greenough 1958) for this view. We now
examine the contribution which may arise from dislocation climb, and
which may be important in cases where the grain size is large and the stress
and dislocation density are high. Possible applications would be t o
single-crystal turbine blades or to planetary matter.
~~

t On leave from University of the Witwatersrand, Johannesburg.

 232 F. R. N. Nabarro on
DIFFUSION
$ 2 . CREEP BY HOMOGENEOUS
We consider the situation in which a network of dislocations of link
length L has been established, L being only slightly greater than the
critical length, approximately bG'/p, for which continuous dislocation climb
by the operation of Bardeen-Herring sources (Bardeen and Herring 1952)
can occur under the applied stress p . A network finer than the critical
can produce by climb only a strain of the order of the elastic strain PIG,
where G is the shear modulus. As the Bardeen-Herring sources operate,
the local dislocation density increases. Simultaneously the density de-
creases as dislocations of opposite sign climb towards one another. Under
a given applied stress, the link length is stable. If it is reduced in one region
of the crystal, it approaches the critical length, and the rate of multiplica-
Downloaded by [Moskow State Univ Bibliote] at 05:13 21 October 2013

tion is rapidly reduced, while simultaneously the rate of removal by the

annihilation of pairs increases. If the dislocations of the network lay each
entirely in one glide plane, they could multiply by glide, but we are
considering situations in which the dislocations are heavily jogged, and the
motion is controlled by diffusion.

Fig. 1

t-- 1, !-+ I-+ J-

J. J. J.
iP
A tlvo-dimensional array of edge dislocations, showing the direction of climb of
each dislocation under the stresses indicated by the arrows.
 Xteady -state Diffusional Creep 233
To an adequate approximation, we may treat the diffusion problem by
considering unit length of a dislocation link as equivalent to unit length
oE an infinite straight dislocation which is climbing in an array of infinite
straight dislocations such as that shown in fig. 1. The domain of each
dislocation is taken to be a cylinder of radius L/rr112. Under steady-state
conditions, there is no accumulation of vacancies except on the line of the
dislocation, and the inward flux of vacancies a t a distance r from the dis-
location line (fig. 2 ) must be roughly of the form Fir. If the chemical
potential of a vacancy in this domain is p ( r ) ,we must have :

or
Downloaded by [Moskow State Univ Bibliote] at 05:13 21 October 2013

This chemical potential, produced by the applied stress, provides the

driving force for vacancy migration through most of the domain of the
dislocation. However, within a distance r e from the dislocation line,
vacancies are swept into the dislocation by a short-range attraction. The
energy density at a, distance r from the dislocation line is about b2G/8n2r2,
arid the presence of a vacancy of volume Q reduces the energy by
about Qb2G/89r2, corresponding to an inward force on the vacancy of
Qb2G/4n2r3. The distance r c is determined by equating this force to the
force QkTFIDr produced by the applied stress.
Fig. 2
-

f - - - L---+

A section of the array of fig. 1. The full arrows show the direction of climb of
the dislocations, and the dotted arrows the direction of vacancy flow.
'We obtain:
rc=
b
-
2n
(-)LTB
GD
. . . . . . . . (3)

I n the portion of the domain of this dislocation lying between r e and

L / d 2 ,the power dissipated in unit volume is ( 6 / r ) grad p. The power
P.M. Q
 234 F. R . N. Nabarro on
W , dissipated in unit length of the domain is easily found to be :
GDb2 ) .
47rkTFL2
w,= 7rQkTFa1n(
D
. . . . . (4)
We determine F by equating W , to the power W , generated in the domain
by the applied stress p . The rate of removal of matter from unit length
of the domain is 2nFQ. A part bGIL of the applied stress does work in
producing new dislocations, and only the part
p ' = p - bG/L . . . . . . . . (5)
is available to dissipate energy a t a rate :
W , = 2nFQp'. . . . . . . . . (6)
Downloaded by [Moskow State Univ Bibliote] at 05:13 21 October 2013

Equating (4) and (6), we find that

L1n(
4rrkTPL2
GDb2 )2Dp'
= r. e
. . . . (7)
This equation may be solved by successive approximations, taking as
zero approximation F = 2Dp'IkT. The first approximation is then :
2Dp'
F = - In
kT
(b2C)
8vLZp'
. . . . . . .

Fig. 3

The shear stress represented by the single-headed arrows is resolved into the
tension and compression represented by the double-headed arrows.
The interstitial loop PRQ and the vacancy loop PSQ form adjacent cells
of the dislocation network. Under the applied stress, these loops both
grow, the line PQ acting as a source for the multiplication of each loop.

We must now determine the equilibrium value of L. The dislocation

array is not in fact two-dimensional, as we have assumed, but three-
 Steady-state Diffusio.na1 Creep 235

dimensional. A typical element in the array may contain the adjacent

vacancy and interstitial loops illustrated in fig. 3. Under the applied
stress, both loops expand, their junction acting as a Bardeen-Herring
source for both. There are L-3 sources in unit volume. Each dislocation
climbs at a speed 2rB’Q/h, and each source operates 2rFQ/Lh times a
second, producing a length r L of new dislocation each time it operates. If
the rate of increase of dislocation density by this mechanism is (dp/dt),,,we
have :
(dp/dt)lH= 2 r 2FQ/L3b. . . . . . . . (9)
This must be equated to the rate - (dp/dt),, at which dislocations are
destroyed by climbing towards one another. A symmetrical array such as
that of fig. 1 is metastable against climb, but in general the adjoining
Downloaded by [Moskow State Univ Bibliote] at 05:13 21 October 2013

dislocations of opposite sign at a distance 21i2Lapart attract with a force of

order h2G/2% L per unit length. As in the case of climb under an extern-
ally applied stress, the flux of vacancies to or from a dislocation is of order
F / r , and the power of the applied forces, in this case (2nF‘Q/b)(h2G/23I2rL)
per unit length, may be equated to the power dissipated between an inner
cut -off radius rc‘ and an outer radius L/r112. We find :
4nkTF‘LZ bGD
F’ln( GDb2 ) = 2112nLkT ’
. . . .
leading to the first approximation :

The two dislocations, initially separated by W2L,move towards one

another with a speed 4rF’Q/b, and the time to collision is Lb/23/2rF’Cl. The
collision destroys both dislocations.
Hence :
(dp/dt),,= -23/2nFrQ/L3b. . . . . . (12)
The sum of (9) and (12) must vanish. Inserting the values of F and F’
from (8) and ( 1 1 ) respectively, we find:
bG 111(8nL2p’/b2G)
L= - . . . . . (13)
2~213’ In ( 2 3 / 2 q b ) ‘
Again we insert the zero approximation L = bG/2n2p‘ in (13) to obtain the
first approximation :
bG In ( 2G/n3p’) bG
L= - %- . . . . (14)
2 d p ‘ In (21/2G/r2p‘) 2 r 2 p f’
Inserting the value ofp’ from ( 5 ) ,we find :

or
L = (1+ l/2r2)(bG/p). . . . . . . . (15)
The equilibrium link length thus exceeds the critical length by only
ame part in twenty, while the effective stress p‘ is given by :
p’=p/(27?+ 1). . . . . . . . (16)
Q 2
 23 6 F. R. N. Nabarro on
The flux of vacancies to each dislocation is now found by inserting
p' =p/2n2 and L = bGlp in ( 8 ) to obtain :
2lcT
F=*/ln($). . . . . . . (17)
If we neglect the logarithm and insert this value of F in (3), we obtain
r , = +b(G/p)1/2, and the conditions for the validity of our calculations,
namely b <rc <L, are satisfied.
Finally, the strain rate tHis equal to n P Q / L 2 ,and so :
Dbp3
tH=-/1n($). . . . . . .
nkTG2
5 3. CREEP BY SHORT-CIRCUIT
DIFFUSIOK
Downloaded by [Moskow State Univ Bibliote] at 05:13 21 October 2013

At sufficiently low temperatures, short-circuit diffusion along the cores

of the dislocations will predominate over homogeneous diffusion through
the perfect crystal. We take a dislocation core to be a channel of cross-
sectional area b2 within which there is an enhanced self-diffusion coefficient
D c + D . Then about 4Dcb2pt/nLkTvacancies pass each node in unit time,
and the dislocation advances at a speed 16Dcb4p'ln2L2kT. The number
of times the source operates in a second is 32D,b4p'/n2L3kT,and the rate
of increase of dislocation density is :
(dp/dt),,=32Dcb4p'lnL5kT. . . . . . (19)
Dislocations also climb towards one another by core diffusion and
annihilate one another under the influence of their mutual attraction.
The stresses between dislocations are of order bGIBnL, and, by the argument
just developed, two dislocations approach with relative speed
16D,b5G/n3L3kT. Their initial separation is L, so the time to collision
is n3L4kT/16D,b5G. Each collision destroys a length 2L of dislocation,
and the number of neighbouring pairs of dislocation links of opposite
sign in unit volume is about ;L3.
Thus :
(dpldt),, = - 16D,b5G/n3L6kT. . . . . (20)
From ( 1 9 )and (20)we again obtain (15)and (16). Knowing the density
and the speed of the dislocation array we at once find the strain rate :
i, = 4D,bp5/n4kTG4. . . . . . . . ( 2 1 )
Q 4. DISCUSSION
The models we have considered bear some resemblance to those discussed
by Weertman (1955, 1957) who found a strain rate proportional to p9l2.
However, Weertman's strain rate depends explicitly on an assumed
dislocation link length L , whereas we have shown that under a given
applied stress an equilibrium link length is established which is only
slightly greater than the critical length of a dislocation source operating
under the applied stress.
A more exact treatment of the influence of elastic interactions between
a dislocation and a vacancy on the rate of climb has recently been published
 Steady -state Diffusional Creep 237

by Gibbs and Turnbull (1967). This affects only the logarithmic factors in
our formulae.
ACKNOWLEDGMENTS
It is a pleasure to thank Dr. E. Hart, who revived my interest in the
theory of diffusional creep, and Dr. J. A. Venables, who suggested the
possible role of short-circuit diffusion.

REFERENCES
BARDEEN,J., and HERRING,C., 1952, Imperfections in Nearly Perfect Crystals
(New York: John Wiley), p. 261.
GIBBS,G. B., TURNBUL BULL, J. A,, 1967, MetalSci. J.,1,25.
Downloaded by [Moskow State Univ Bibliote] at 05:13 21 October 2013

GREENOUGH, A. P., 1958, Phil. Mag., 3, 1032.

HERRING, C., 1950, J. appl. Phys., 21,437.
NABARRO, F. R. N., 1948, Strength of SoZids (London: The Physical Society),
p. 75.
WEERTMAN, J., 1955, J . appl. Ph,ys., 26, 1213; 1957, Ihid., 28, 362.