Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
JI'I"
.:.,":""I.T ... nr ..... 1.1:.r1 .......
o 3...lI 0 l:..lJ
.AT iff'll: r:r,AFLJ'1:\""D(lN l'l .... t:d:9
......
Oxj(}1'd Uni'i\N'sUy Press, Ely House, Landrm. W. I C.U9COW }o,"FW YORX T1tIIo:rro r..lF.T..B(':ITJRto.":R ~F.1.J.1 '1(, n ~N
~ TOWN ~Al.m'UIIY II.IADAN _~.\.IH.Oill 1)..\.11. ~ aUAM LU!:IIJU ~:rml!'l A;&.\E,!\ :HI."iJ.£B 11." [: A I. c [11"[ :.. M ... rtJI.1.5: KIo JIA c: f[ I T _,r, H (nu; UA. {".C!L
HU ......... U.':'I.D',:ll SlliC.\F{]RE HONG ll.ONC 'TOX"l"n
_.
!rfJL.!!;'l' 11! hl '1'1 (I :'I I ~J.~ 6
RE:I'lI:Th"TEI":I "LH'H1Hll'iA 1"KH:Jo.J.T,.." 1 rf ~J[KJo.'" TlJl:N' .HI'I '.T ~ UKIV"CM1TV FR~, ()X"J'CIo"fL1'l
IJ;(I;O):l; 80RBBCT.ED SHBErs Oll' '1:11][ I'I1tIi.T EDITI()~ ] 9.~ 7 ~ 1'9154) 196:;>. 1970
U X IVF. ns J TAT SBIBL10THEl( C L !\ (J s T HAL
PREFACE
A ru.On"M pre~~~~~ t.Jt.I~ for this book would bo 'Mat heJllaticnJ soiutlons of the di ff\l.'5i on c qua. tion ' ~ for i L !~ wit h t h i.'9 as pee t of t.h e rna thema tic!": nf diffusion that the hook is mainly concerned. It. deals with the de!wlcjil)~ t.jon of diffusion processes in terms of solut ions of the differential equation for diffusion. Litr le mention is made of the alternative, but 1 p.~~ well dev elo ped, .le~ eript ion in t eJ'!!J ~ of w ha t is corn m on ly called • til e random VI' alk ': nor aIT theories of the mechani ..... m of diffusion in liartlt:uhu' ~'y_"'lh=:m~ included.
'I'll ~ In .... them at i cal theorv of d iffu 810ll is (o ui «led on tha ~ of hen. t
....
conduct.ion and correspondingly t.hc early !)art of't his book has developed
from 'Conduction of heat in 8Olid~' I)}: Carslaw and rTaeger. These authors present UHU: L Y solutions of t he equation of heat Don d uotion and some of them. can he applied (() ~ 1 i iT'u sion problem s for w hich the diffusion coeff ~ ien t· is cons t an t.._ I have selec ted some of the sol u t ions which seeru most likely to hE' of interest In diffusion and they have been ev al1l.a ted n umericall y all d prcsen ted in grn phioal forn ~ ~() as to be readily usable. Several problems in whieh diffusion i~ complicated by the effects of an iuunohihzing reaction of some ~m·t H,TO also ineluded. Convenient W~L)'"S of dedviug the mathematical solutions are descr i hed .
'~tht:=:n wo come to systems in \VhJClh the (UU·u::.::i(m eocfficient i!ol. not· constant but v ~ ~ri.u. hie ~ and f01' the most part th j s means conoen tra ti on [ICpe1 idcnt ~ we finct ths t st rict.l y formal rnu th ematieal sol u tions no longer exist. T have tried to indicate the \r;LTi01.l8 methods by which numerical and graphical solutions have l~~n obtained! mostly within t he laat ten ycar~ ~ a.J HI tH prosou t, again ill gra phical form, some sol uti( ms for various Cc meen tratiou de pel] dP11 t. diffusion eoeff cion t.s. A~ well as beiug useful in them ad vef; thf.'~~ sol u tion ~ ill u~ tra te the charaotoristl c tea.tll res of a con cell tr a tion _ rle p~lld(~n t· sys t em. Conai deration is alISO given tot he d. )~el:r allied pro 1) lern of determining the diffusion cocff olen 1. .a 11 d ita tl~l enden ec on eoneen tration f to II~ .f: X perimcntal n] casu reme 1 L t.s, The diffu sion eoolf c!f'1 L ts m casurcd by difft=:n=:nt types of cxperimellt arc shown to he simply !·e.1Jl.too. The final chapter deals w j th the tC':RI pern ture ch a ng('!~ W hieh sorneti rnes aceom pi:ln:r d i ffu sion,
In ~vt=:ral instonces I have thought it h~tt.(':r to refer to an easily ncce~~ihle book or paper ra ~heI' t n an to the fiTF. t I)U bl ished ueeou n t, which the reader n I Jgh t f n d diffiu u lt to 0 bt ain: EJ.L.~~ of l'BfeI'en(!e
J. C.
vi
PREFACE
usually seemed of primary importance, particularly with regard to mathematical solutions.
I should like to express my thanks to my friend and colleague, Mr4 A. C. Newns, who read the typescript and made many valuable comments and suggestions, and also to Mrs, D. D. Whitmore, who did most of the calculations and helped to correct the proofs and compile the index. I am grateful to Miss D". Eldridge who, by patient and skilful typing. transformed an almost illegible manuscript into a very clear typescri pt for the printer. I should also like to thank the following \v ho readily gave permission to use lOR terial from various publications:
Professor It. :rtl. Barrer, l_\.Ir. ~I. u. Coyle; Dr. P. V. Danek werts, Dr. L. D. Hall, Dr. P. S. H"~ Henry, Professor J4 C. .Iaeger, r». G. S. Park, Dr. R. H. S tokes, Dr ~ c. W agner, and the pu blishers of the following journals, British Journal of Applied Physics, Journal of Chemical Pbvsics, Journal of .JJ etals, Journal oj Scientific Instruments, Philo .. sophical Af aqazine, Proceedinqs of the Physica.l ~1ocietyj Transacsions of the Faradau Society. Finally, it is a pleasure to thank those members of the staff of the Clarendon Press who have been concerned with the production of this book for the kindness and consideration they have shown to me,
Maidenhead Decem be", 1955
CONTENTS
I. THE DIFFUSION EQUATIONS
1
II. }fETHODS OF SOLUTION \VHEN THE DIFFUSION
COEFFIClENT IS CONSTANT 9
III. INFINITE AND SE:r..JI·I~FINITE MEDIA 26
IV. DIFFUSION I~ A PLAXE SHEET 42
V. DIFFUSION IN A CYLINDER 62
VI. DIFFUSION IN A SPHERE 84
vrr, DIFFUSION ,,71TH A l\fOVING BOUNDARY 99
VIII .. SII\IULTANEOUS DIFFUSIO~ AND CHEMICAL
REACTION 121
IX. SOlIE METHODS OF SOL UT!ON FOR v ARIABLE
DIFFUSION COEFFICIEXTS
x. FINITE~DIFFERENCE METHODS
XI~ THE DEFINITIO~ AND MEL~SUREMENT OF DIFFUSION CQEFFICIEXTS
147
186
219
XII. SO~IE CALCULATED RESULTS FOR \T ARIABLE
DIFF·USION COEFFICIENTS 258
XIII. SIMULTAN:EOLTS OIl~}"LTSIOX OF HEAT AND
•
MOISTURE 303
TABLES 326
INDEX 341
I
THE DIFFUSION EQUATIONS
1.1. The diffusion process
DIFFUSION is the process by which matter is transported from one part of a system to another as a result of random molecular motions. It is usually illustrated by the classical experiment in which a tall cylindrical vessel has its lower part filled with iodine solution, for example, and a column of clear water is poured on top) carefully and slowly, so that no convection currents are set up. At first the coloured part is separated from the clear by a sharp, welldefined boundary. Later it is found that the upper part becomes coloured, the colour getting fainter towards the top) while the lower part becomes correspondingly less intensely coloured. After sufficient time the whole solution appears uniformly coloured. There is evidently therefore a transfer of iodine molecules from the lower to the upper part of the vessel taking place in the absence of con vection
.
currents. The iodine is said to have diffused into the water.
If it were possible to watch individual molecules of iodine, and this can be done effectively by replacing them by particles small enough to share the molecular motions but just large enough to be visible under the microscope, it would be found that the motion of each molecule is a random one. In a dilute solution each molecule of iodine behaves independently of the others, which it seldom meets, and each is constantly undergoing collision with solvent molecules, as a result of which collisions it moves sometimes towards a region of higher ~ sometimes of lower, concentration, having no preferred direction of motion towards one or the other. The motion of a single molecule can be described in terms of the familiar 'random walk" picture, and whilst it is possible to calculate the meansquare distance travelled in a given interval of time it is not possible to say in what direction a given molecule will move in that time.
This picture of random molecular motions, in which no molecule has a preferred direction of motion, has to be reconciled with the fact that a transfer of iodine molecules from the region of higher to that of lower concentration is nevertheless observed. Consider any horizontal section in the solution and two thin, equal, elements of volume one just below and one just above the section, Though it is not possible to say which
2
THE DIFFL~SION EQUATIO~S
[§ L 1
way any parbicular iodine molecule will move in a gi yen in terval of time t it can be said that on the average a definite fraction of the molecules in the lower element of volume will cross the section from below, and the same fraction of molecules in the upper element will cross the section from above, in a given time. ThusJ simply because there are more iodine molecules in the lower element than in the upper one, there is a net transfer from the lower to the upper side of the section as a result of random molecular motions,
1.2 .. Basic hypothesis of mathematical theory
Transfer of heat by conduction is also due to random molecular motions, and there is an obvious analogy between the two processeS4 This was recognized by Fi ck [1] t who first pu t diffusion on a quantitative basis by adopting the mathematical equation of heat cond uction derived some years earlier by Fourier [2]. The mathematical theory of diffusion in isotropic substances is therefore based on the hypothesis that the rate of transfer of diffusing substance through unit area of a section is proportional to the concentration gradient measured normal to the
section, i.e.
F = DeC/ax,
(1.1)
where F is the rate of transfer per unit area of section, C the concentration of diffusing substance, x the space coordinate measured normal to the section, and D is called the diffusion coefficient. In some cases, e.g, diffusion in dilute solutions, D can reasonably be taken as constant, while in others, e.g. diffusion in high polymers 7 it depends very markedly on concentration. If F, the amount of material diffusing, and C, the concentration, are both expressed in terms of the same unit of quantity, e.g. grammes or gramme molecules, then it is clear from (I r 1) that D is independent of this unit and has dimen8ion8Iength2time~1} e.g. em. 2sec.1 The negative sign in equation (1 .l ) arises because diffusion occurs in the direction opposite to that of increasing concentration.
It must be emphasized that the statement expressed mathematically by (I. I) is in general consistent only for an isotropic medium, whose structure and diffusion properties in the neighbourhood of any point are t.he same relative to all directions. Because of this symmetry ~ the flow of diffusing substance at any point is along the normal to the surface of constant concentration through the point. As will be seen later in § 144, this need not be true in an anisotropic medium for which the diffusion properties depend on the direction in which they are measured,
§ 1.3]
THE DIFFUSION EQUATIONS
3
1.3~ Differential equation of diffusion
The fundamental differential equation of diffusion in an isotropic m.edium is derived from equation (1.1) as follows.
Consider an element of volume in the form of a rectangular parallelepiped whose aides are parallel to the axes of coordinates and are of lengths 2dx~ 2dy, 2dz. Let the centre of the element be at Pix, YJ z], where the concentration of diffusing substance is C. Let A BG D and A r B' C' D' be the faces perpendicular to the axis of x as in Fig. 1.1. Then the rate at
A"'       ~        ldx                 "A FIG. 1.1. Element of volume,
which diffusing substance enters the element through the face ABeD in the plane xdx is given by
4dYdz(F;t~ dx}
where Fx is the rate of transfer through unit area of the corresponding plane through P. Similarly the rate of'Iosa of diffusing substance through the face A' B' 0' D' is given by
4dYdz(F",+ °e;dx}
The contribution to the rate of increase of diffusing substance in the element from these two faces is thus equal to
~ 8dxdydz 8Fx ..
ax
Similarly from the other faces we obtain
aR oR
 8dxdydz ___J!_ and  8dxdydz ~~ .
By &
But the rate at which the amount of diffusing substance in the element increases is also given by
BO 8dxdydz, .ot
4
THE DIFFUSION EQUATIONS
[§ 1.3
and hence we have immediately
80 + bP~+ 81~+ ()~ == o.
at ox oy iJz
( 1.2)
If the diffusion coefficient is oonstant, Fx, F1P F:, are given by (I .1), and
(1.2) becomes ~C = D(02C + (/2.0 + 02°) (1.3)
at Bx2 cy~ OZ2'
reducing simply to
&C 82C Dr
at ~ ox2'
( 1.4)
if diffusion is onedimensional ~ i.e. if there is a gradient of concentration only along the zaxis. Expressions (I ~ 1) and (1.4) are usually referred to as Fick ~ s first and second laws of diffusion, since they were first formulated by Fick [1] by direct analogy with the equations of heat conduction.
In many systems, e.g. the interdiffusion of metals or the diffusion of organic vapours in high polymer substances, D depends on the concentration of diffusing substance, O. In this case, and also when the medium is not homogeneous so that D varies from point to point, equation (1.2) becomes
Be == !_(D 6C) 0 (n ~C) ~A(D 80)
~t ~ ~ + r; 5i + r. ,,'
() (.IX ox oy l'Y oz oz
(1.5)
where D may be a function of x~ y~ Z, and C.
If D depends on the time during which diffusion has been taking place but not on any of the other variables, i.e.
D == f(t)t
then on introducing a new timescale T such that dT ~ j(t) dt~
the diffusion eq ua tion becomes
&0 62C 820 i/l.C
(}11 = ox2 + cy2 + 'BZ2 '
(1.6)
which is the same as (1.3) for a constant di ffusion caeffi cient equal to unity.
1.31. Diffusion. in a cylinder and sphere
Other forms of the above equations follow by tra.nsformation of coordinates, or by considering elements of volume of different shape. Thus by putting
x == r cos B,
y ~ r sin e',
§ La]
THE DIFFUSION EQUATIONS
or by considering an element of volume of a cylinder of sides dr ~ r dB, dz, we obtain the equation for diffusion in a cylinder,
.00 I{O(DO~ B(DoC) 8(nOC)}
at = r or r or) + 80 r. 00 + 8z r az '
( 1.7)
in terms of the cylindrical coordinates T, OJ z, The corresponding equation for a sphere in terms of spherical polar coordinates r, 0, rp is obtained
by writing x == r sin 0 cos t$,
y == r sin 8 sin q" z == rcos 0,
or by considering an element of volume of a sphere of sides dr, rdO, r sin e d4. It is
eo ~ _!_{!_(nr2 cO) + I !!_(D sin 8 8e) + D 82C) ~ (1.8 )
Bt r2 or 8r sin 8 EO he sin28 i:J¢2
The simplified forms of (1.7) and (1.8) for purely radial diffusion t e.g in a long cylinder where end effects are negligible or in a spherically symmetrical system ~ are given in Chapters V and l'I where some sol utions of the differential equations are to be found. All these diffusion equations can be expressed in terms of the nomenclature of vector analysis as
eo
ot == div(D grad C).
1 .. 4. Anisotropic media
Anisotropic media have differen t diffusion properties in different directions. Some common examples are crystals, textile fibres, and polymer films in which the molecules have a preferential direction of orientation. For such media it is not always true, as was stated in § 1.2 for isotropic media, that the direction of flow of diffusing substance at any point is normal to the surface of constant concentration through the point, This means that (1.1) must be replaced in general by the assumptiona
ec (}C 8e
Fx ::::;:::! D11+ Dl'2~+D13;"
ox By oz
aO ao BO
_. s, = D21ax + D22 ey + DZ3fiZ r'
80 ae ae
~ === D31+Ds2+DssJ
ax oy bz
( 1.9)
80 that Px, for example, depends not only on Be/ax but also on oG/ay and ao/ez. The D'8 have the significance that D13 Be joz} for example;
6
THE DIFFUSION EQUATIONS
[§ L4
is the contribution to the rate of transfer in the xdirection due to the component of concentration gradient in the zdirection. Substituting from (1.9) for the E'« in (1.2) we obtain
eo 82C »c 820 82C
fit = Dll ox2+ D2'l oy2+ D33 az:/' + (D2s+Daz) oyoz +
820 vo
+(Da1+D13) Bz8x + (D12+ Du) axvy' (1.10)
if the D's are taken as constant. The extension to nonconstant D's is obvious from (1.5)4 A transformation to rectangular coordinates " 7}t , can be found which reduces (1.10) to
80 820 0"0 820 ~ = Dl~+D ~+D3
at a,2 2 aTJ2 8,,, .
This is the same transformation as that by which the ellipsoid DII X2+ D22 y2 + DS3 z2+ (D23 + D32)yZ +
+ (DSl + D13)zx + (D12 + D21)xy :::=: constant (1.12) o, ~2+ D2 7}2+ o, ~2 == constant. ( I .13)
(1.11)
is reduced to
The new axes may be called the principal axes of diffusion and Dl~ D2, Da the principal diffusion coefficients. If we make the further transformation
(1.14)
where D may be chosen arbitrarily, (1.II) becomes
eo (020 82C 02Q'\
at = D o~f + B'7~+ Btlr
(l.15)
This has the same form as equation (143) for isotropic media, and hence certain problems in anisotropic media can be reduced to corresponding problems in isotropic media, Whether or not this can be done in a given case depends on the boundary conditions, Thus it is possible when the medium is infinite, or when it is bounded by planes perpendicular to the principal axes of diffusion 80 that the boundary conditions are of the familiar form C = constant, g == O~ , ~ it t > 0, for example, and similarly for TJ and {. The problem of diffusion into an anisotropic cylinder which has its axis along ~ and is bounded by planes perpen~ dicular to ~ reduces to the corresponding problem in an isotropic cylinder provided D2 == Da.
Certain properties deduced by Carslaw and Jaeger [3) P: 29] indicate the physical significance of the ellipsoid and also of the principal axes
§ 1.4]
THE DIFFUSION ~:QU ATI0NS
7
of diffusion. 'rhus it can be shown that the square of the radius vector of the ellipsoid in any direction is inversely proportional to the diffusion coefficient normal. to the surfaces of constant concentration at points where their normals are in that direction ~ Hence the di ffusion coefficient ~ DM, at right ang les to surfaees '''" hose normals ha ve direction cosi nes l ~ m, n relative to the principal axes of diffusion is given hy
D,~ == 12 », +m21)~ +n2jJa' (] .16)
Carslaw and Jaeger further show that if there is symmetry about the planes g == 0 and '1"J == 0, then the general relationships (I. H) for the F's reduce to
( 1.17)
This simplification also occurs for other types of crystallographic S.:vrnmetry. It means that the flow through a surface perpendicular to a pri noi pal axis of diffusion is proportional simply to the coneen tra t ion gradient normal t.o the surface as is the case for isotropic media.
1.41. Significance of measurements in anisotropic media
Since in the majority of experiments designed to measure a. diffusion coefficient the flow is arranged to be onedimensional, it is worth while to see how such measurements are affected by anisotropy. If the diffusion is onedimensional in the sense that a concentration gradient exists only along the direction of x~ it is clear from (1.10), since both C and 8e/ex are ovorywhore independent of!1 and z, that the diffusion is governed by the simple equation
he! (;2C: _ .. == Dll 
r·t .... 2'
C ox
{I.IB}
and D1l is the diffusion coefficient measured. If the direction of diffusion is chosen to he that of a principal. axis, then 1)11 is equal to one or other of the principal diffusion coefficients 1)1' J)2~ or /)3. Otherwise the cocfficien t D} t == I)n' related to D; t D2, D3~ by (1.1 f)) is measured. This would be measured, for example, by an observation of the rate of flow through a plane sheet of a crystal cut, so that its normal has direction cosines (I! Rl1 n) relati ve to the princi pal axes of diffusion of the crystal. Similar remarks apply to a high polymer sheet in which there is both uni planar and unidirectional orientation, i .e, the molecules a 1'0 arranged with their long axes lying main ly paral lc I to the pl anc of the sheet and all parallel to one direction in that plane. The principal axes of diffusion of such a sheet will be normal to the plane sheet., and along and
.'~ . . '~.'
'~."
8
THE DIFFUSION EQTJATIONS
u 1.4
perpendicular to the preferred direction of orientation in that plane. Even if a concentration gradient exists in one direction only, it is clear from (1.9) and (1.15) that the diffusion flow is not along this direction unless it coincides with a principal axis of diffusion .
•
REFERENCES
14 Fick, A.j Ann. PhY84 Lpz: 170 (1855), 59.
2. Fourier, J. B~I' Theorie analytique de la chaleur. <Euvros de Fourier (1822).
3. Carslaw, H. S.~ and Jaegor, J. C., Conduction of Heat in Solide (Oxford" 19(7)'t
.. , .
.. i:"
."1"
·'~.
·:7:'
18. ~ar on
II
METHODS OF SOLUTION WHEN THE DIFFUSION COEFFICIENT IS CONSTANT
2~1~ Types of solution
GENERAL solutions of the diffusion equation can be obtained for a
7)~ variety of initial and boundary conditions provided the diffusion
coefficient is constant. Such a solution usually has one of two standard forms. Either it is comprised of a series of error functions or related integrals, in which case it is most suitable for numerical evaluation at small times, i.e. in the early stages of diffusion, or it is in the form of a trigonometricaJ series which converges most satisfactorily for large values of time. When diffusion occurs in a cylinder the trigonometrical series is replaced by a series of Bessel functions. Of the three methods of solution described in this chapter, the first two illustrate the physical significance of the two standard types of solution. The third, employing the Laplace transform} is essentially an operator method by which both types of solution may be obtained. It is the most powerful of the three, particularly for more complicated problems, The methods are presented here as simply as possible" The fuller treatments necessary to make the discussion mathematically rigorous are to be found in works on heat conduction, e.g. Carslaw and Jaeger [1].
,I:
.. , ",
} :' ,'J.'
2.2~ Method of reflection and superposition l.21~ Plane source
It is easy to see by differentiation that
C _ A xl/WI
~ tl e ,
where A is an arbitrary constant) is a solution of
(2.1)
(2.2)
which is the equation for diffusion in one dimension when D is constant. The expression (2.1) is symmetrical with respect to x = 0, tends to zero as x approaches infinity positively or negatively for t > 0, and for t = 0 it vanishes everywhere except at x == 0, where it becomes infinite. The
B
I.. t
,: sat,.!
, h ....
t·lS elM
. 0 ~
10
METHODS OF SOLUTION 'VHEN THE
[§ 2.2
total amount of substance, MJ diffusing in a cylinder of infinite length and unit erosesection is given by
00
M = f Cdx,
and if the concentration distribution is that of expression (2.1) we see,
on writing
that
QO
M = 2ADi f e£' tiS = 2A(1i'D)1.
00
(2.5)
]·0
L ~
1 \
•
\i
i..
L: ~
It ,
11\
L
~ t:1 ~ ~ ~ 0·,
·1
o
2
]
FIG. 241. Concentret.ion distance curves for an instanteneoua plane source.
Numbers on curves are values of Dt.
Expression (2~5) shows that, the amount of substance diffusing remains constant and equal to the amount originally deposited in the plane x == o. Thus, on substituting for A from (2.5) in equation (241), we obtain
a _ M ~xt/'Dt
~ 2(1TI!t)l e ~
(2.6)
and this is therefore the solution which describes the spreading by diffusion of an amount of substance .ill deposited at time t = 0 in the plane x === o. Fig, 2~1 shows typical distributions at three successive times.
2
1
f 2.2]
DIFFUSION COEFFICIENT IS CONSTANT
J I
2.22~ Reflection at a boundary
Expression (2.6) carr be used to build up solutions of other problems in linear .flow by introducing the concept of reflection at a boundary. Thus, in the problem just considered, half the diffusing substance moves in the direction of positive x and the other half along negative x. If, however, we have. a· semiinfinite cylinder extending over the region x > 0 and with an impermeable boundary at x === OJ all the diffusion occurs in the direction of positive x. We can consider the solution for negative ~ to be reflected in the plane x = 0 and superposed on the original distribution in the region x > o. Since the original solution was symmetrical about x = 0 the concentration distribution for the semiinfinite cylinder is given by
C  M xl/4Dl
 (,"Dt)le l>
(2.7)
This procedure of reflection and superposition is mathematically sound, for reflection at x == 0 means the adding of two solutions of the diffusion equation. Since this equation is linear the sum of the two solutions is itself a solution, and we see that (2.7) satisfies the condition that the total amount of diffusing substance remains constant at M. Furthermore, the condition to be satisfied at the impermeable boundary is
20 lox == O~ x = 0, (248)
since this is the mathematical condition for zero flow across a boundary. As ae / ax is zero at x = 0 in the original solution (2.6), it is clearly still zero after reflection and superposition.
2~23. Extended initial distributions
So far we have considered only cases in which all the diffusing substance is concentrated initially in a plane. More frequently in practice, howevert the initial distribution occupies a finite region and we have an initial state such as that defined by
C ~ GOt X < 0, G :=;;;;; 0, x > OJ t = O. (2.9)
This is the initial distribution, for example, when a long column of clear water rests on a long column of solution, or· when two long metal bars are placed in contact end to end. The solution to such a problem is readily d~duced by considering the extended distribution to be composed ?f an infinite num her of'line sources and by superposing the corresponding mfinite number of elementary solutions, With reference to Fig. 2.2t consider the diffusing substance in an element of width d€ to be a line
METHODS OF SOLUTION WHEN THE
[§ 2.2
12
source of strength Code4 Then, from (2.6) the concentration at point ]J) distance € from the element, at time t is
00 dg ef'/4Dt
2 (1TDt). t
~c
(~
.0

  .   ~ .   ~ t   ~       .  i
O{ :1;==0 P X
FlO. 2~2. Extended initial distribution.
and the complete solution due to the initial distribution (2.9) is given by summing over successive clements df, i.e. by
where '1 == ~/2~(Dt).
A standard mathematical function, of which extensive tables are available, is the error function, usually written as erfz, where
z
erfz = :. J e7/' d 1j. o
(2411 )
This function has the properties
erf(z) = erfz, erf(O) == 0,
erf(oo) == 1)
(2.12)
and hence, since
~ 00 z
J e'l' dT} = J e1J' d7J  f e'" d1j = lerfz = erfc e,
e 0 0
(2.13)
where erfc is referred to as the error ... function complement, the solution (2.10) of the diffusion problem is usually written in the form
ot», t) = !Co erfe 2.J~Dt) ·
(2.14)
,
DIFFUSION COEFFICIENT IS CONSTANT
13
§ 2.2)
C venienf tables of the error function are those of the Works Project
on 1 .
Association [2J and shorter tables arc to be found, for examp e, In
Milne~Thomson and Comrie [3]. Table 2.1, taken from Carslaw and Jaeger (I], is sufficient for many practical purposes. The form of the concentration distri bu tion is shown in Fig. 2.3. J t is clear from (2 .14) that C ~ lOo at x = Ofo~ all t > o.
~ r.
"\ 1\
\ ~
\ ,
r"\.
<; ~ 06
o 2
1 xl(4Dt)1i
2
1
o
FlO. 2.!t Concenfrat.iozrdistance curve for an extended source of infinite extent.
The error function therefore enters into the solution of a diffusion problem as a consequence of summing the effect of a series of line sources ~ each yielding an exponential type of distribution.
In the same way, we can study the diffusion of a substance initially confined in the region ~h < x < +h as in Fig. 2.4. Here the integration is from x~h to x+h instead of from x to 00 as in (2.10), leading immediately to the result
( h~x h+X}
G = tOo erf+erf ~
2.J{Dt) 2~(Dt)
The concentration distribution at successive times is shown in Fig. 2.4. It is clear that the system can be cut in half by a plane at o: = 0 without affecting the distribution, which is symmetrical about x = O. Therefore expression (2~15) also gives the distribution in a semiinfinite system. Such a system is realized in practice in the classical experiment hi which
(2.15)
METHODS OF SOLUTION WHEN TH.~
(§ 2.2
a cylinder contains a layer of solution having on top of it an infinitelylong column of water, initially clear, In practice, this means that concentration changes do not reach the top of the column during the time of the experiment.
l'O~~~~~~~~~'
etc,
O. g.t+t.......,_ ............ ....,.__.._,_t"""""r....+.......__.,_.+r
o
o~~~~~~~~~~~~
4
3 x/h 4
o
)
1
3
2
FIG. 2.4. Concentrationdistance curves for an extended source of limited extent, Nwnbers on curves are values of (Dt/ht.)·~
2~24. Finite systems
If the column of water, referred to above, is of finite length, l, the condition that the concentration tends to zero as x approaches infinity is to be replaced by the condition that there is no flow of diffusing substance through the top surface, i.e,
ao/ax == 0,
x ~ I.
(2.16)
We have seen that this condition is satisfied if the concentration curve is considered to be reflected at the boundary and the reflected curve superposed on the original one. In the finite system we are considering now, the curve reflected at x == l is reflected again at x = 0, and then at x = l and so on, the result of each successive reflection being super ... posed on the original curve (2.15). Since the original solution is the sum of two error functions, the complete expression for the concentration in the finite system is an infinite series of error functions or errorfunction
I 2.2]
DIFFUSION COE}1FICIENT IS CONSTANT
15
complements so that
{ xh x+h 2lhx
C = !OO erfc 2,.j(Dt)  erfc 2.j(Dt) + erfc 2.J(Dt) 
2l+hx 2l~h+x 2l+h+x
erfc 2,J(Dt) +erfe 2.,j(Dt) erfe 2.j{Dt) +
. 4l~h~x 4l~x+h }
lerfo 2.,j(Dt) erfc 2,.j(Dt) + ...
~ { h+2nlx . h~2nl+x}
= 100 n?_~ erf 2,.j(Dt) +erf 2,.j(Dt) · (2.17)
A solution of this kind can be obtained for most problems in diffusion by use of the Laplace transform, to be discussed below in § 2.4, or otherwise. Such solutions are most useful for calculating the concentration distribution in the early stages of diffusion, for then the series converges rapidly and tW9 or three terms give sufficient accuracy for most practical purposes.
. In all cases the successive terms in the series can be regarded as arising from successive reflections at the boundaries. The nature of the reflection depends on the condition to be satisfied. For the impermeable boundary already considered a simple reflection ensures tha.t Be I ox = 0 as required. Another boundary condition which occurs frequently is of the type G = 0, in which case it is necessary to change the sign of the concentration when it is reflected at the boundary. A further example of the use of this method is given by Jost [4, p. 38]. For more complicated problems, however, the reflection and superposition method soon becomes unwieldy and results are more readily obtained by other methods.
2.3., Method of separation ofvariables
A standard method of obtaining a solution of a partial differential equation is to assume that the variables are separable. Thus we may attempt to find a solution of (2.2) by putting
C = X(x}T(t), (2.18)
Where X and T are functions of z and t respectively. Substitution in (242) yields
X'!!_ == DTd2X
dt dx'2. '
which may be rewritten
I dT D dZX
T (jj = X dx2'
(2419)
16
METHODS OF SOLUTION WHEN THE
[§ 2.3
so that we have on the lefthand side an expression depending on t only, while the righthand side depends on x only. Both sides must therefore be equal to the same constant which, for the sake of the subsequent alge bra, is conveniently taken as  AiD. We have, therefore, two ordinary
differential equations dT
I ~D 2)
.r s :  , (2~~O
and
(2.21 )
of which solutions are
(2.22) (2.23 )
and
x = AsinAx+BcosAX,
leading to a solution of (2.2) of the form
C == (A sin Ax+ B cos :tr)e~VDtJ
(2.24)
where A and B are constants of integration. Since (2.2) is a linear equation, the most general solution is obtained by summing solutions of type (2.24), so that we have
eo
C = I (Am sin Amx+Bm cos~x)e~!'Dl,
m=l
(~.25)
where Am} Bm, and Am are determined by the initial and boundary conditions for any particular problem 4 Thus if we are interested in diffusion out of a plane sheet of thickness I, through which the diffusing substance is initially uniformly distributed and the surfaces of which are kept at zero concentration, the conditions are
o == 00' 0 < x < l, t === 0,
C === 0, x == 0,. x === l, t > o.
The boundary conditions (2.27) demand that
Bm = 0, ~ == m1T/lt
and hence the initial condition (2.26) becomes
(2.28)
co
Go == I Amsin(m1Tx/l)J 0 < x < l. 1
By multiplying both sides of (2.29) hy sin(p1Tx/l) and integrating from o to l using the relationshi pa
I
f · p'TTX · m1TX d _ {O, m =F p,
SIll l Sln·l x ~ l[ _
'I, m  p~
o
(2.30)
§ 2.3]
DIFFUSION COEFFICIENT IS CONSTANT
17
we find that terms for which m is even vanish, and Am == 4Go/m1r, m == 1, 3., 5,411.
The final solution is therefore
C 400 L:w I D(2Il+1}t1f~t/la. (2n + I }7TX
==  ~ e SIll l '
tt 2n+ 1
n=O
(2.31)
where 2n+l has been substituted for m for convenience so that n takes values 0, 1., 21 .... This trigonometriealseries type of solution converges satisfactorily for moderate and large times, and it is then used for numerical evaluation in preference to the errorfunction type of solution discussed earlier in § 2.24.
In (2429) the initial distribution is expressed as a sum of sine functions.
This reveals the physical significance of the trigonometrical series in (2.31)~ each term of which corresponds to a term in the Fourier series (2429) by which the initial distribution can be represented.
2.4. Method of the Laplace transform
The Laplace transformation is a mathematical device which is useful for the solution of various problems in mathematical physics. Application of the Laplace transform to the diffusion equation removes the time variable, leaving an ordinary differential equation the solution of which yields the transform of the concentration as a function of the space variables x, Y, %1 This is then interpreted, according to certain rules, to give an expression for the concentration in terms ofz', y, z and time, satisfying the initial and boundary conditions. Historically the method may be regarded as derived from the operational methods introduced by Heaviside. Full accounts. of the Laplace transform and its application have been given by Doetsch [5],. Carslaw and Jaeger [6]) Churchill [7], and others. Shorter accounts by Jaeger [8] and Tranter [9J are also available, Here we shall deal only with its application to the diffusion equation, the aim being to describe rather than to justify the pro eedure.
The solution of many problems in diffusion by this method calls for no mathematics beyond ordinary calculus. No attempt is made here to explain i ts application to the more difficult problems for which the theory of funct.ions of a complex variable must be used, though solutions to problelns of this kind are quoted in later chapters. The fuller accounts shou ld be consulted for t.he derivation of such solutions,
)8
METHODS OF SOLUTION \VHEN THE
2.41. Definition. of the Laplace transform
Suppose J{ t) to be a known function of t for positi ve values of t, Then the Laplace transform !(p) of f(t) is defined as
co
J(p) = f eJJtj(t) dt, (2.32)
o
where p is a number sufficiently large to make the integral (2~32) converge. It may be a complex number whose real part is sufficiently large, but in the present discussion it suffices to think of it simply as a real positive number. For example, if I(t) === e2t, p must exceed 2. Unless it is necessary to emphasize thatJis a function of p, just as/is a function of t, we shall usually denote the Laplace transform of f by J.
Laplace transforms of common functions are readily constructed by carrying out the integration in (2~32) as in the following examples:
co
J(p) = J e_pi dt = lip,
o
1(1) == 1,
(2.33)
oc: IX'
J{p) == fePie", dt = r e(pa)l dt === 1 , (2.34)
~ pa
o 0
co
J(p) == f e~pl sin wt dt ::::=;: w ·
p'2+w2
o
Extensive tables or dictionaries of Laplace transforms are available, s~me in the works referred to above and others by Doetsch [10] and by l'tlcLa.chlan and Humbert [Ill A short table of transforms occurring frequently in diffusion problems is reproduced from Carslaw and Jaeger's book [IJ in Table 2.2.
/(t) = sin wt,
(2.35)
2 .. 42" Semiinfinite medium.
As an example of the application of the Laplace transform, consider the problem of diffusion in a semiinfinite medium, x > 0, when the boundary is kept at a constant concentration, Co, the initial concentration being zero throughout the medium. We need a solution of
satisfying the boundary condition
C == ('01
t > O~
x=o ~
(2.37 )
12•4]
DIFFUSION COE}"FICIENT IS CONSTANT
19
and the initial condition
C = 0, x > O~ t === O. (2438)
On multiplying both side~ of (2.36) by eP/ and integrating with respect to t from 0 to 00 we obtain
00 00
J a2C 1 f  taO
en•pt_ dt~  e P  dt ~ o.
8x2 D ct
o 0
If we assume that the orders of differentiation and integration can be interchanged, and this can be justified for the functions in which we are
interested, then
(2.39)
co 00 _
r a2C B2 J 02C
eP' dt =  Oepl dt = ~.
8x2 oxt ox2
•
U . 0
Also, integrating by parts we have
(2.40)
00 W
I eptaa dt = [GePt]!;O + P J Cept de = pOt
et 0
o 0
since the term in the square bracket vanishes at t == 0 by virtue of the initial condition (2.38) and at t = 00 through the exponential factor. Thus (2.36) reduces to
(2.41 )
n82C = c. ex' P
(2.42)
By treating the boundary condition (2~37) in the same way we obtain
rJ)
 J G.
a = aoepi dt = ;, x = O. (2,(3)
o
Thus the Laplace transform reduces the partial differential equation (2.36) to the ordinary differential equation (2.42). The solution of {2.42) satisfying (2443)~ and for which 0 remains finite as x approaches infinity is
C == 00 eilX, (2.44)
p
where q2 == pID. Reference to Table 2.2, item 8, shows that the function whose transform is given by (2.44) is
x
G = Goerfe 2J(Dt)' (2.45)
where, as before, erfc z = 1 erf z, (2.46)
~t ~ easy to verify that (2.45) satisfies (2.36), (2437), and (2.38) and that It is therefore the required solution of the diffusion problem.
20
METHODS OF SOLUTION WHEN THE
[§ 2A~
2,,43. Plane sheet
In the problem just considered the transform solution could be inter ... preted immediately by reference to the table of transforms, Consider now, as an example of a slightly more difficult problem in which this is not so, a plane sheet of thickness 2l, whose surfaces are maintained at constant concentration, 00' and with zero concentration of diffusing substance throughout the sheet initially, Let the sheet occupy the region ~l ~ x ~ L, so that there is symmetry about x === 0 and the boundary conditions may be written
G ::= CO~ x =: 1, t ~ 0,
aCj8x==O, x~O, t~O.
Equation {2.48} expresses the condition that there is no diffusion across the central plane of the sheet. It is often more convenient to use this condition and to consider only half the sheet, 0 ~ x ~ 1, instead of'using the condition a = CO) x == ~l.
The equations for the Laplace transform C arc
d2C 20 ~ 0
dx2Q ,
o < x < I,
(2.49)
with
x === 0,
(2.50) (2~51 )
and C = Co/p, x = 1,
where q2 == plD as before. The solution of these is
o = °ocoshqx.
pcoshql
There are two methods of dealing with this transform solution, leading to the two standard types of solution we have already met. \Ve shall first obtain a solution useful for small values of the time.
(2.52)
(i) Expan.sion in negative exponentials
We express the hyperbolic functions in (2.52) in terms of negative exponentials and expand in a series by the binomial theorem. TllUS we obtain from (2.52}t
{j = q,(eqx+eOX) = CO{e!iIXI+eq(I+X}} i (_1)ne2nol
pet1l( 1 + e2ql) p 11.=0
= 00 i ( I ) "eQ{('h'+1)1.?:} + q~ ~ ( I)neq{(·n"_lJll.?:}. (2.53)
P n=.·O P 11=0
" ".:"
§~.4] DIFFUSION COEF}~ICIE~T IS CONSTANT 21
Thus, using item 8 of the table of transforms (Table 2.2) we obtain
ee (2n+ l)lx ~ n (2n+ l)l+x
0=0. 2: (I)"erfc 2.j{Dt) +OO~f'o (1) erfo 2..j(Dt) ·
.0 (2454)
This is a aeries of error functions such as we obtained by the method of reflection and superposition. Successive terms are in fact the concentrations at depths Ix, l+x, 3lx, 3l+x,. .. in the semiinfinite medium. The series converges quite rapidly for an except large values of Dt/fl.. For example, we have for the concentration at the centre of the sheet
(x = 0) when Dt/l2 === 1
(2.59)
CIOo = 095900·0678+0·0008 = 0·8920, and when Dtjl2 = 0·25
C/Go === 0·31460·0001 ==:; 0·3145.
(ii) Expre8sion in partial fractions
It can be shown that if a transform 9 has the form
Y = f{p},
g(p)
(2.55)
(2.56)
(2457)
whereJ(p) and g(p) are polynomials in p which have no common factor, the degree off(p) being lower than that of yep), and if
g(p) === (pal)(p~a2)40.(pan)' (2.58)
where at) a2,. .. ~an. are constants which may be real or complex but must all be different) then the function y(t) whose transform is y(p) is given by
y(t) = .i: I,(ar) ,p.t. t"= 1 g (ar)
Here gl(ar) denotes the value of dg(p}/dp when p ~ a.. A proof of this by Jaeger [8] is reproduced in the Appendix to this chapter. It is derived by expressing (2.57) in partial fractions. Since the hyperbolic functions cosh z and sinhz can be represented by the following infinite products (see, for example, L 12], p. 275)
( 4Z2) ( 4z'2 ) ( 4z'2 )
cosh z = 1 + 172 1 + 321T2 1 + 52".2 ••• t
. (22) ( Z2 ) ( Z2 )
sinh z = z 1 + 7T2 I + 221T '2 I + 32.,,2 • •• ,
(2.60)
(2.61 )
a quotient of these functions such as in (2.52) may still be regarded as being of the type (2.57) except that now f(p) and (J(p) have an infinite
22
I'.:lETHODS OF SOLUTION \VHEN THE
[§ 2.4
number of factors. The av a2 J' •• are the zeros of g(p)" i.e. solutions of the equation} g{p) = 0, and if these are all different it is plausible to assume that (2.59) still holds with 'TIt == 00. The justification of this assumption involves the theory of functions of a complex variable in order to carry out a contour integration and is to be found in the fuller accounts of the subject. There is, in fact, a rigorous mathematical argument by which the use of (2~59) with n == 00, can be justified in diffusion problems in a finite region only. It must not be applied to (2.44), for example, for the semiinfinite region. The above refers to al, a2"" all different. The extension of (2.59) to cases in which g(p) has repeated zeros, e.g. one of its factors is square, is given in the Appendix. Its application to an infinite number of factors is still justifiable.
We may now consider the application of (2.59) to (2.52). First the zeros of the denominator must be found. Clearly, p = 0 is a zero, and the other zeros are given by the values of q for which cosh ql == O, i.e,
(2n+ 1)1Ti
q = ± 2l ; n == 0, I, 2,.~. (2.62)
D( 2n+ 1 )2.,rt
P ==  412 ~ n = 0, I, 2 p •• •
and hence
To a.pply (2.59) to (2.52) we need
g'(p) = ~(p cosh ql) = cosh ql+ iql sinh ql.
For the zero p = 0, gJ(p) = 1. For the other zeros, given by (2.62) and (2.63), coshql = 0, and
 h 1 · h (2n+ l)1Ti •• (2n+ 1)7T .( l)n
SIn q = sm ;::::::;:;: 'I, Sill ::::=:= 1. ~ .,
2 . 2
(2.65)
so that for these zeros, by substituting in (2.64) we obtain
, (2n+ I )1T( 1 yr.+l
g (p) == .
4
Hence finally by inserting the zeros into (2.59) we obtain
C ~ n 40.0 2:m (~l)n D{2n+l)II1T'f141~ (2n+ I )1TX
~ Ve)  ~ e cos  •
7T 2n+ I 2l
n=O
This is the trigonometricalseries type of solution obtained previously by the method of separation of the variables. The series converges rapidly for large values of t. Thus for the concentration at the centre of the sheet (x == 0) when Dt/l? . 1 ~
c leu ~ 1 ~ o 1080 == 08920, (2.68)
(2.66)
(2.67)
§ 2.4:J
D IFJi"USIO N CO E F F rot E NT I S CO NST ANT
23
and when Dt/l2 ~ 0·25
C IGo == I ~O·6872 + 0'0017 ==:; 0"3145.
(2.69)
2.5. Solutions expressed as the product of the solutions of simpler problems
Consider the equation of diffusion
62c 82c B'2c 1 oe
, .. +~+ .~ = ~, 8x~ axl oxl D at
in the rectangular parallelepiped
al < Xl < bl, a2 < X'2 < b2~ u's < Xa < b3~
(2.70)
(2.71 )
For certain types of initial and boundary conditions, the solution of (2.70) is the product of the solutions of the three onevariable problems, and thus can be written down immediately if these are known. The following proof is given by Carslawand Jaeger [I, p. 22].
Suppose cr(xr) t)~ r =::;; 1 ~ 2, 3, is the solution of
(2.72)
with boundary conditions
oc.r a 0
ex,. ~ ~ flr C,. = ,
aXr
I > 0,
(2.73)
t > 0,
(2.74)
where the aJ" and J3r are constants, either of which may be zero (so that the cases of zero surface concentra tion and no .flow of heat at the surface are included) and with initial conditions
Cr(Xr, t} == Cr(xr), t == 0, Ur < XT < br
Then the solution (2 ~ 7 0) in the region defined by (2. 71) with
c == C1 (Xt)C2(X2)Ca(Xs), t == 0,
a.nd with boundary conditions
(2.75)
(2.76)
t > 0, r == I, 2, 3,
(2.77)
•
18
(2.78) (2.79)
r == I, 2~ 3,
t > 0,
24
METHODS OF SOLUTION WHEN THE
[§ 2.5
For substituting (2.79) in (2.70) gives
o'Scl B2c2 a2c.3 1 (OCI oC2 8cs )
C2iC3~+ CSC! ~+Cl CZ2~D C2C3 ;+ cac1+ c1 c2  = 0,
ax 1 vX2 OXa ut at Be
(2480)
using (2.72) t Clearly the initial and boundary conditions (2.76), (2.77)) and (2.78) are satisfied.
Similar results hold for combined radial and axial flow in a solid or hollow cylinder, and for other regions such as a rectangular corner [IJ.
APPENDIX TO CHAPTER II
To deduce the function y(t) whose Laplece transform y(p) is given by
~() f(P)
y p ~ g(p)1
we first put y(p) into partial fractions in the usual way by assuming
(1)
n
f{p) = ~ Ar = Al + A2 + A3 + h. + An . (2)
g(p) ~ pur pa1 pa, pa3 jJan. r=l
n.
f(p) == ! Ar(pa1)u.(p a,l)(PG'+l)··'{P an),
,=1
(3)
Then
and putting p = a; in this gives
/(ar) == Arf a,.  at}4 •• ( a;  Qrd(ar ar+l ).u(a,  all) (r == 1, 2,~ •. , n). (4)
Substituting for A, from (4) in (2) gives
n
y(p) =" I f(a,.) )" (5)
~ p Ur (ar  a I}' ~.( a,  a,_d( ar ar+1 }u.( aT  an
rl
Now since g{p) = (pal)(p a~)41.(pan)t (6)
we have} on differentiating by the ordinary rule for differentiation of a product
'R
g'(p) ~ L (p~al)·u(pa,_l){par+l}···(p~an).
r""'l
(7)
Putting P ~ ar in this, gives
g~(a,.,.) ~ (a, atl·~·(aT~a1'"_l)(ar ar+tl~ .. (aran) and using (8) in (5) gives a further form for y(p) namely
(8)
n
 ""' f(ar)
y(p) = ~ (pa )g'{a ).
r=l r r
(9)
On applying item 3 of Table 2.2 to successive terms of (9) we obtain immodiately
( 10)
l)IFFUSION COEFFICIENT IS CONSTANT
ThiR result applies only to the case to which g(p) has no repeated zeros, but it can read ily be genera] i zed for th e C Rae of repeated rae torli4 Th us (10) j m plies that to e.a.c h lin ear f actor p  Ur of t.he denom j na tor of fj(p) th ere correspond 8 a term
f (ar) e~" t ( 11 )
g'(a,)
in the solut.ion. The generalization is that, to each squared factor (pb)= of the denominator of y(p} there corresponds a term
rp Wf(l'l] tei>t+ [!!_ (fp~b)~(p)}] ebl (12)
g(P) p=b dp g{p) v;c;b
in the solution. To each multiple factor (p c}m of tho denominator of fj(p) there corresponds a term
mt
" I d3 {(P ~ C}"1(p)}] tm,l
~ dp~ 9(P) p~.c8! (1flsl)! ed
".:.....0
(13)
In tho solution.
RE~"ERENCES
1. Cars}a\v, II. S., and Jaeger, J. C.t Conduction oj Heat in Solids (Oxford, 194') . . 24 Tables of the probability function (Works Project Association, New York, 1941 ).
3. Milne Thomson, L. M.t and. Comrie, L. J. ~ Standard Fourfigure Mathem.atical T obles (1\1 acmillan, 1944) .
4. Iost, \V., Diffu.sion in Solids, Liquids, Gases (Academic Press, 1952).
5. Doetsch, G.t Theone und Anwendung der Laplace Tran.a/orm.ation (Springer, 1937).
6. Carslaw, H. S.~ and Jaoger, J. C., Operational Methtxl8 in Applied Mathematica
[Oxford, 1941). .
7. Churchill, R. V., Modern Operational Mathematics in Engineering (McGrawHill, 1944).
B. Jaeger, J4 C.~ An Introduct·ion to the Laplace TransJort1Ultion (Methuent 1949).
9. Tranter, C. ~J.~ Integral Tronsforms in klathematical Ph:yBicJJ (~Iethuen~ J951).
10. Doet.sch, G., Tabellen zur Laplace Trans!offlUltwn und Anleitung zum Gebraf1.Ck (Springer, 1947).
11. McLachlan~ N. W., and Humbert, r., Pormulaire pour 1e. Calcul Sym.bolique (Oauth icrVillars, 1941).
12. Carslaw, H. S.~ Plane Trigonometry (Macmillnn, 1909).
o
III
INFINITE Ai\D SE~II .. IN~"tINITE ~IEDIA
3.1. Introduction
IN this and the following three chapters sol utions of the diffusion equation are presented for different initial and boundary conditions, In all cases the diffusion coefficient is taken as constant. In many cases the solutions are readily evaluated numerically with the help of tables of standard mathematical functions. Where this is not 80, and where numerical evaluation is tedious, as many graphical and tabulated solutions as space permits are given.
3.2.. Instantaneous sources
Under this heading are included all problems in which an amount of diffusing substance is deposited within a certain restricted region at t.ime t = 0 and left to diffuse throughout the surrounding medium .. For example, it may be located initially at a point) or in a planet or within a sphere, when we have an instantaneous point, planet or spherical source as the case may be.
The solution for an instantaneous plane source in an infinite medium has already been gi ven in Chapter II j equation (2. 6 ) ~ The corresponding solution for all instantaneous point source on an infinite plane surface is obtained in the same way by recognizing that
(3.1)
is a solution of
(3 .. 2)
which is the equation for diffusion in two dimensions when the diffusion coefficient is constant. Tho arbitrary constant A is expressed in terms of M ~ the total amount of substance diffusing, by performing the
integration
ct) 00
M = J Ie dxdy = 47rDA,
IX) OJ
the concentration G being expressed in this problem as the amount of diffusing substance per unit area of surface. The concentration at a
,
§ 3.2)
INFINITE AND SEJ\.II~I~FINITE lfEDIA
27
distance r from a point source on an infinite plane surface is thus given by C _ ~f r:ll/4rH (3.4)
 471Dte ·
The corresponding expression for a point source in an infinite volume is
o == _ M er:J/4Dt. (3.5)
8CrrDt)t
By integrating the appropriate solution for a point source with respect to the relevant space variables, solutions may be obtained for line, surface, and volume sources. Thus for surface diffusion in the x, y plane due to a line source along the yaxis we have
co
o == J M e(x'ty'J/4Dt dy = Jf eX'/1Dl (3.6)
4TrDt 2(1T Dt)~ ,
where now .. JI is the amount of diffusing substance deposited initially per unit length of the line source. Th is is the same as expression (2.6) of Chapter II for a plane source of strength ~ll per unit area in an infinite volume. The corresponding result for a line source of strength M per unit length in an infinite volume, obtained by integrating (3~5), is
c ~ ..J..lf e  r~/4r.Jt
47TDt '
(3.7)
which is the same expression as (3.4) for a point source OJl an infinite plane surface, though .1lf has a different significance in the two cases. Results for a variety of sources are derived by Carslaw and Jaeger [1, p. 216]. The spherical and cylindrical sour~es are likely to be of practical interest. If t.he diffusing substance is initially distributed uniformly through a sphere of radius a, the concentration C· at radius r, and time t is given by
C = ~G. {erf a+r + erf __ a~r } _ OoJ(Dt){e(arr~I4.Dt_e(a+r)~/4DI}
o 2, .. ./(Dt) 2,/(Dt) r 17 '
{3.8)
where Co is the uniform concentration in the sphere initially. Expression (3.8) may easily be written in terms of the total amount of diffusing .8ubstance~ M, since
M ~ !1ra:l0o• (3~9)
The corresponding result for a cylinder of radius a may be written in the form
a
C =:::: Co erl[4Dt f e~r'llWt 1 ( rr' ) r' dr'
2Dt o 2Dt t
o
(3.10)
28
INFINITE AND SEMI·INFINITE 1\["~DIA
f§ 3.2
where 10 is the modified Bessel function of the first kind of order zero. Tables of 10 are available [2]. The integral in (3~IO) has to be evaluated
numerically except on the axis r == OJ where (3.10) becomes .
(3.11)
These expressions may be applied, for example} to the diffusion of a sphere or cylinder of solute into a large volume of solvent. Curves showing the concentration distribution at successive times are given in Figs. (3.1) and (3.2).
The expression (3.10) and the curves of Fig. 3.2 also apply to a circular disksource, of radius a, on an infinite plane surface, if Co denotes the uniform concentration of the diffusing substance over the region o < r < a initially. ....~n alternative solution given by Rideal and
Tadayon [3] is 00
G = aGo J J1{ua,)Jo(ur}eDiU' du, (3.10a)
o
where Jo and J1 are Bessel functions of the first kind and of order zero and one "respectively. Tables of Jo and J1 are available [2]. Rideal and Tadayon [3] also give an expression for the total amount of diffusing substance, Q, remaining on the disk after time t, which is
(3.10 b)
For small values of t, (3.10h) becomes
Q = WU!Co{ 1 :~:)i},
and when t is large we have
Q == ?Ta'Oo/( 4Dt).
(3.10c}
(3.10d)
Cases of an extended source in an infinite medium t where the diffusing substance initially occupies the semiinfinite region x < 0 or is confined to the region h < x < h, have been considered (see equations (2.14) and (2.15) in Chapter II). The solution to the corresponding problem in which the region h < x < h is at zero concentration and x < h at a uniform concentration Co initially is
(3.12)
J '0 Cleo o·g
0'2
r. 0 I
1
4

"
....._
<, K
~
1 ~ ~
~ 0'6
o
]
')

3
ria
4
FiG. :t I. Concentration distributlons for a. spherical source. Numbers on curves are values of (Dtja!)t.
l'O~~~~~~~~~~~'~~'
Cleo
O'8~~~+~~4~~~~r~~+~~
1
2
3
ria
4
:FrG+ 3 . .2. Concentration distributions for & cylindrical source. ~mnber8 on curves are values of (Dt/al)' ~
IN.FINITE AND SEMI~I)lFINITE ~l}:DIA
[§ 3.2
30
3.21. Coniinuoue sources
A solution for a continuous source, from which diffusing substance is liberated continuously at a certain rate, is deduced from the solution for the corresponding instantaneous source by integrating with respect to thne t Thus if diffusing substance is liberated continuously from a point in an infinite volume at the rate cfo per second, the concentration at a point distant r from the source at time t is obtained by integrating (345) and is
(3.5 a)
If cP is constant and equal to q, then
o = q erfc r .
47TDr 2~(Dt)
(3.5 b)
Solutions for other continuous sources are obtained similarly and examples are given by Carslaw and Jaeger [1, p. 220].
3113. Semiinfinite media
The solution for a plane source deposited initially at the surface, x === 0, of a semiinfinite medium was given in Chapter II, equation (2. 7)~ and that for the initial distribution C == Go, 0 < x < h, C = 0, x > h, was seen to be given by equation (2415) for x positive.
The problem of the semiinfinite medium whose surface is maintained at a constant concentration OO~ and throughout which the concentration is initially zero, was handled by the method of the Laplace transform in Chapter II, § 2~42 (see equation 2.45). Other results of practical importance which may be obtained in the same way are given below.
(i) The concentration is Co throughout, initially, and the surface is maintained at a constant concentration 014
001 f x
 == er t
G~Cl 2~(Dt)
(3.13)
The special case of zero surface concentration is immediately obvious. The rate of loss of diffusing substance from the semiinfinite medium when the surface concentration is zero, is given by
(3.14)
§ 3.3]
INFI N ITF. A, ND S E1'.IIAINFINITE ltEDIA
31
so that the total amount .J.'i, of diffusing substance which has left the )l1ediuln at time t is gi ven by integrating (3414) with respect to t and is
" r (Dt)!
il~ == 200 ; •
(3,15 )
The same expression with Co replaced by C1 gives the total amount taken up by the medium in time t if the initial concentration Co is zero. If the initial concentration is zero throughout the semiinfinite medium, and the surface concentration varies with time, solutions are still obtainable by the Laplace transform. Cases of practical interest are gi ven below, Here .l~ is used throughout to denote the total. amount of diffusing substance which has entered the medium at time t.
(ii) G~.o == ki, where l~ is a constant.
C ~'::': u {(I + ~:~)erfc __ ~,_~_ _ x~ __ e ~X~/4Dl}
tDt 2,/(Dt) ,. . ./(rrDt)
Akt '2 . X
::=: I eric ~
2,,1 (Dt)
Jl[, = ~t(~t)t.
The function i2:erfc is defined and tabulated in Table 2.1~ so that values of 0 may be written down immediately. The effect of an increasing surface concentration is shown in Fig. 3~3.
(iii) Cx:.:o ~ ktl, where k is a constant.
C ==== ktt {eXi/4Df ~ X71t erfc x \
2~(Dt) 2,/(Dt))
= k( 7ft)! ierfc "j x .
2 (Dt)
jJft =::;: 1 kt ( 1r D) l.
(3.18)
(3,19)
The function ierfe is defined and tabulated in Table 2.1. In this case M is directly proportional to t and so the rate of uptake of diffusing substance is constant.
(iv) Cx=o ~ kt&1\ where k is a constant and n is any positive integer, eVen or odd,
o === k r(ln+ 1)(4t)ini"erfc ..; X _.
"2" ' 2 (Dt)
M, = .!";{Dt)(4t)ln r(ln+ I) .
t 2n f(in+ I)
(3.20)
32
INFINITE AND SEMIINFINITE MEDIA
[§ 3.3
The function i=erfo is defined and tabulated up to n == 6 in Table 2.1. r(n) is the gamma function defined and tabulated, for example, in ref, 4, p. 210. If n is even, so that in == N, an integer, then
r(}n+l) == r(N+l) == N!. (3.22)
6 ~~~~.
MtD/k or MI/~
5 ~~~++~~~~
4 ~
1/
c:
.~
"oi
iJ1.
w...,;
3 §
€
(j
~
~
~
2 0·4
1·2
FIG. a.:I. Sorp tion curves for varia.ble and constant aurf ace coneen trBtiona in a. semi infinite modi urn.
If n is odd, so that tn = M ~t, then
r(!n+ 1) = r(M +l} ~ 1.3.5 .... (2M3)(2M 1)17'/21l1. (3423) Other properties of gamma functions are
r(n+l) == nr(n); r(i) == 1T.r
(3.24)
A polynomial in ti may sometimes be a useful way of representing a given surface concentration empirically. In such a case the complete expression for the concentration at any point is the sum of a number
§ 3.2J
IN1"INITE AND SE~[I·IN:FINITE MEDIA
33
of terms of type (3.20) corresponding to successive terms in the polynomial.
(v] These solutions can be extended to cover other initial and boundary concli tions by using the fact that for a linear differential equation the BU m of two sol utions is itself a solution. For exam ple, if
(3.25)
and the semiinfinite medium is initially at zero concentration throughout, the solution is given by
o = Coerfe ./D + 4kti2orfc ./n '
2 ( t) 2 ( t)
(3.26)
since the first term on the righthand side of (3.26) is the solution satisfying the conditions
o = Co~ C == 0,
x == 0 }~
x>O
(3.27)
and the second term satisfies
C = kl
,
C == 0,
(3.28)
In general the solution to the problem of the semiinfinite medium in which the surface concentration is given by F(t) and in which the initial distribution is f(x), is given by
C == c1+c2:'
where C1 is a solution of the diffusion equation which satisfies
c1 = 0, t == 0 ),
c1 == F(t), x == 0
and c2 is another solution satisfying
(3.29)
(3.30)
C2 == !(x),
t == O}.
Cz ==== 0, x ==== 0
Clearly, with C1 and C2 so defined, the diffusion equation and the initial and boundary conditions are satisfied.' Consider, as an example, the problem of desorption from a semiinfinite medium having a uniform ~ni tial coneen tra tion CO) and a surface concen tration decreasing accord 1ng to (3.25), with k negative. The solution is
(3.31 )
c == Co+ 4kt i2erfc .~~,.
• '2, ... '/( Dt) ,.
I N FIN IT E AND S ElII INFI N ITE .:\1 E DIA
[§ 3.3
34
which is obtained by adding to (3.26) the solution satisfying
C~ == Co, x > U, t=:=:o ,
G~ == 0, x == 0, t > 0 J'
i.e, by adding 00 erf{x/2 .. ./(Dt)}.
(3.33)
3.31. Sur face evaporation condition.
In some cases the boundary condition relates to the rate of transfer of diffusing substance across the surface of the medium. Thus, if a stream of dry air passes over the surface of a solid containing moisture, loss of moisture occurs by surface evaporation. Similarly if the solid is initially dry and the air contains water vapour, the solid takes up moisture. In each case the rate of exchange of moisture at any instant depends on the relative humidity of the air and the moisture concentration in the surface of the solid. The simplest reasonable assumption is that the rate of exchange is directly proportional to the difference between the actual concentration, Os, in the surface at any time and the concentration, CO~ which would be in equilibrium with the vapour pressure in the atmosphere remote from the surface. Mathematically this means that the boundary condition at the surface is
80
D· == a_(CoCs)~ x ~ 0, (3.34)
ax
where a: is a constant of proportionality.
If the concentration in a semiinfinite medium is initially C2, throughout, and the surface exchange is determined by (3.34), the solution is
O~2/~ erfc. ,:_"_ehX+h'Dlerfc'9 IX +h.,J(Dt)}, (3.35)
Co  C2 2~' (Dt) t ,,' (Dt)
where h == Ci./D. The special cases of zero concentration in the medium initially (02 == 0), and evaporation into an atmosphere of zero relative humidity (Co := 0), are immediately obvious from (:3.35)~ The rate a.t which the total amount, ~I,) of diffusing substance in the semiinfinite medium per unit crosssectional area changes is given by
dll~ (De~ C
_. = . ~ === ~(Co ~)~
dt ox X~O
(3.36)
and, on substituting for G.Q the value obtained from (3.350) by putting x == 0, after integration with respect to t we obtain for the total quantity of diffusing su bstance having crossed unit area of the surface,
(3.37)
If 06 is greater than O2 this amount is taken up by the medi urn; if Co, is less than Cz this amount is lost by evaporation from the surface. The expression (3.35) can be written in terms of any two of the dimensionless
parameters x
.1 3.81
0·8
0·4
0·2
INFINITE AND SEJ\.II·INFINITE :PtIEDIA
h~(Dt)j
0['
hx.
(3.3S)
06
1\
1\
1\
\
1 ,.
~
, \
\ 1\
,.'S \ \
~
\ '\
1'0. ~
\ v
1\ J
.\ . ... :
i\ r
1\ \ \
~. \00
,
~ \
rY
" f\
.:\_ v . 4'
:~
1\
1\ 1"\ ['I;; "'\ "'
~ I\.. 2'1ro.. :\ 1'1"
"Ii
r\. '\
"'\ "\ 1 I'\: \ "
I'\. r\. .
I\.. N I\.
" 1"0.
i"'o r... ~t 1'1" ~
I' ...... r...
r, ...... fo"
!'o., {' r.... r.. r.
""" r.. r, t... .....  .
r. .... .... ['ii; .... ...:;
""" r;:;; ""'" ~ .... r.
.... 1'. . .... .... .... ~ 1""'"1'.
'' r,, ~I. ~&~~ T
I o·
08
16
04
F,G. 3.4. Concen tration diet ribution for l) surface evaporation condition in a semi ~ infinite medium. Numbers on curves are values of h{Dt}'.
In Fig. 3.4 the ratio (C C2)/(COC2) is plotted as a function of x/2,/(Dt) for various values of h ... .,/(Dt)4 In order to evaluate hl1~/((}oC:?) from (3.37), only one dimensionless parameter 1l\!(Dt) is needed. The relationship is readily evaluated from standard functions and is Sl10'Vll graphically in Fig. 3~5. The evaluation for large h, . ./(Dt) is made easier by using the asymptotic formula
{ 2 f. 1 (I I I . 3 )
exp z )er e z === .   ~3 + r •• r •
1T~ Z 2z 22 . z.l
( ~l.39)
3t32. Squareroot relationship
. Expression (2.45) shows t hat the solution of the problem of diffusion lnto a semi~infinite medium having zero initial concent.ration and the
36
INI1'INITE A~D SEIUI~IXFI~ITE ~rEDIA
[§ 3.3
6
'r
~ 1 I
 "
, ,,
 ~
 "

"
Wi'
"J
1,1
r,.ill
:.J
• .111
.
II"
!o !o I ,,
f. ... ~ V
II'

. 10 r &I
 LJI
1..1
." 1.111
t...
_.
.~
v
.,
v
l.I
~
l.#I
l..,j
I~
.III ,  
..
I
~
IIf"
,
~ I
If'
II .
"
~
,
"J r
~ v
~ I • 1.
....
r 1 1

~
~
II"
I.0Il ,
1..01 5
4
J
2
J
o
3
4
1
5
6
2
h{Dl)1 PIGr 3.5. Sorption curve for a surface evaporation condition in a semiinfinite medium.
surface of which is maintained constant, involves only the single dimensionless parameter
x
( 3.40)
It follows from this that
(i) The distance of penetration of any given concentration is proportional to, the square root of time.
(ii) The time required for any point to reach a given concentration is proportional to the square of its distance from the surface and varies inversely as the diffusion coefficient.
(iii) The amount of diffusing substance entering the medium through
unit area of its surface varies as the square root of time.
These fundamental properties hold ill general in semi infinite media, provided the initial concentration is uniform and the surface concentration remains constant. They also hold for point and line sources on
§ 3.31
I~FINITE AND SE}II~I~FINITE l\IEDIA
37
infini tc surfaces or in infinite modi a, and also for the case of diffusion in an infinite medium where the diffusing substance is initially confined to the region x < O. Clearly they do not apply to cases where parameters other than x /21>,./( i)t) are in vol ved t such as the width of an extended source or the rate of change of surface concentration, etc.
3.4. The infinite composite medium
Here we consider diffusion in systems in which two media. are present.
Suppos~ the region x > 0 is of one substance in which the diffusion coefficient is D1, and in the region x < 0 the diffusion coefficient is D24 In the simplest case, the initial condi tions are tha t the region x > 0 is at a uniform concentration Co, and in x < 0 the concentration is zero initially. If we write C1 for the concentration in x > 0 and c2 in x < 0 the boundary conditions at the interface x ==== 0 ma~y be written
x===o
,
(3.41) (3.42)
where k is the ratio of the uniform concentration in the region x < 0 to that in x > 0 when final equilibrium is attained. The condition (3.42) expresses the fact that there is no accumulation of diffusing substance at the boundary. A solution to this problem is easily obtained by combining solutions for the semiinfinite medium so as to satisfy the initial and boundary conditions. We seek solutions of the type
x > 0,
x < 0,
which are known to satisfy the diffusion equations in the two regions. By choosing the constants AI' B l' A 2' B2 t.o sa tisfy the initial condi tiona and (3.41)~ (3.42) we obtain
(3.45)
(3.46)
Fig. 3~6 shows a typical concentration diatribut.ion for the case where D2 ~ 4Dl and k ;:;::;:~. Graphs for other cases are shown by Jost l5, P: 26]
INFINITE AND SE MI· IN'FIN ITE MEDIA
if 3.4
3A
and by Burrer [6, p. 10]. \·Ve may note that, as diffusion proceeds} the concentrations at the interface, x == 0, remain constant at the values
00 kCo
c1 == 1+k(D2/D1)i' c2 = 1+k(D.JD1)i·
(3.47)
I{}
! V
V
:
~ V
 CIC
I 0
04
0·2
4 X
6
2
o
2
6
4
FIG. 3+6. Concentration distribution in 0. compoaite medium.
VI t :;;::: 1 ~ D= = 4D1' k = i.
3.41 ~ Interface resistance
If we have the same problem as in § 3.4 except that there is a contact resistance at x === 0, then (3.41) is to be replaced by
Dr eCl/cx+h(c2~Cl) == 0, X = Ot . (3.48)
while (3.42) still holds. The expressions for the concentrations in this case arc
c1 == ,!l_o·~rl+(D2)I{erf~ _.+
1 + (D2.! DI)! D} 2 .. ../(D1 t)
+ e" :t:+ftfDli erfc(. _x~ + k !(D t))}] (3.49)
~ 2~./ { D, t) 1 \J 1 ~
C2 === .. COl l{erfc!.~.~ehIX+hiDt.terfc(!XI_+h2,./(D2t))}'
1+ (D2,.' Dl)· 2~./(D2t) 2~(D2 t)
(3~50)
where
(3.51 )
§ 3.4J
IN.FINITE AND SE1\.1I·J~FINITE MEDI~\
39
The concent.ra tiona on ci ther side of the in terface are no longer constant but each approaches the equilibrium value, leo) relatively slowly. This} and the general distribution at suceessive times, is illustrated in Fig, 3.7 for the case in which Dl = D2 and k == it,Dj•
J +0 Cleo
0·8
o~ v ~
~
25
/
V

l~
.i> V ~o
__... 0·6
0·2
40
 30
20
]0
)0
o
20
30 x 40
FlO. 3.7. Concentration distribution in a composite medium with a reeistenoe at the interface. Numbers on curves are values of D. t, Dl = DI~ Ie = ..fuD i 
3.5. The semiinfinite composite medium
This is the case of a semiinfinite medium which has a skin or surface layer having diffusion properties different from those of the rest of the medium. Thus, suppose in the semiinfinite region l < x < OOt the diffusion coefficient is D, in the region ~l < x < 0, and that the concen tration is denoted by C1 there) w hile the corresponding q uan titles in x > 0 are Dz and Cz. If we assume the conditions at the interface to be
C1 ~ c2' X == 0, (a.52)
Dl ccll/ex = D2 2c2/?xt X == 0) (3~53)
the solut ion to the pro blem of zero initial concentration and the surface x = l maintained at constant concentration Co is given by Carala W· and Jaeger [I, p. 2631t and is
2:00 n{ (2n+ 1 )l +x (2n + I )lxi
c1 === Co a erfc y ..  •  Ct: erfc ...... _ ~
n=o ~\/(Dlt) 2\/(Dlt) }~
(3 . .'54)
(3.!J5)
40
INFINITE AND SE1\IIINFI~ITE MEDIA
lk
0::== •
l+k
(3.56)
where
The total quantity entering the medium through unit area of the surface x == ~ l in time t is llli~ where 11ft is given by
_~~ = 2(·DJt)I{I+2 ~ ll:"en"'ID.I}~4 ~ na"erfc nl . (3.57)
to, 7Tl2 is is ~(Dlt)
6 r;· . ; . T :., ~.I~
. . .! ,;  ~ : .: : 1 ; .. 1 _, ! ~ D CD ,. . ... , .,.~
M ~/1" . ; . ~ 7;' j . .·1 ~:..'  :. . ~ ., ~ . ! . .  i : : :' . ~ , f:.
~ ""tJ. ., ' ..: •.. _~. . :.  ,  ., _ .~:', I , 1 ••••. : .~.~
r:.: ' . : .;. . .: .: +, + .
.• ~ 1 . ••• 1 • i . !   . . :j , r·  ...:___
j... I • ~ ! _~. I.' : • •• I . _. • i .' !.. . .  ;
~.~ ..  . i· i ·of '.' •• •• , ! _r •• ,.
S I : : • ••••• r
 i I· r . ... T""'.:.. .!.; ; .. .,. i. .; . ~
..  1+..+. . ,.411
,.......:...,.'1., ~+I .+~+.. . . 1. . 7
f7. ~ . t·· :, _. ' +.~+. , " ... ,~ 7 . . •
~+~~ ~++' _. . IZI. ,., .; .."/
 1  .+'........ I. .. ~~ ,.. . " ++.:'
_. ::. , ' .' ~ ./ ... ::
, . . ~ ,  .: . ··1
~ ., . : .. . ...;. I' I .. ~ .,: :~  . ;  ... r ..:.._. I
4 :; :; "  !  . ... i ; . ~.  !.: ":7 /. ~  , 
rr : .!. . i . ~ . I .. ~ : .J., .' . /+ .
fi..  ~ : .: . ,.  , ~ .  !. ~I , .'  , 1/ '. ._. ,  .. _;_
_+. _' _.;..... +. t· _. I  I I I " ~ ...
~. .• • ._. I _.r ..._
f.~ , ! _ .. _ :: ." f· _ . :. __ . : . ~.~_.~V:.. . : . : t· :>'""'I
..o+...;......+~ .. ~.~. . . I . . . ~ /. .. r . ,. :.. ~
3 ~~.. .,' . ..  ~ . .  ., . ~ : I .~. ...., ~...
I .. . ....... ... •• L. .$ ... '. _;..1 I • I _/
.; ., ~  ~!  :. . . : ;. /: "/; ..  "', _. ,.~. ~
;i:. :to • ,. :. • :~ :, • :; ,. /' • • ~ • • !.
:. i . .• .. ," . _/. ,. ~ t ~. .__.;... 'I
.. y' """""'_1
~ , _.  1 .++T"""""'".,.  . ..i'~ ~ ./".: ,.  •
r;. ,. ., . .1 . ., /;. ., /." , ~ _;_
111. ,. I .. y . . . 10
~.. .. : ; .. + '.../ ' :.;  ' ...... I. '" : ~ .... 
2 " . .. <  ! • • ' / .~ '~....,....'...... ...~
, .: ~. ; ,J. .~ .. .,'
 . I. : :.. . . ...,.. /.> :..:~ ~ .., ~.t:I
. . f· ': ..: ./ ~: .•.  ~  .: •. ., .• ~
1· . / 1 .. _..I' _. ., . ; ~... II~
~ ..  . .;. . . •.. ~. i/./ .:;.. .. , _... _ .••• _ ............... _+ ... ~
~~,+_ . ,.: _. / .::: ~~ ,. __ . :. . .L~
r+ .; ; . ../ / ..r . ,.., .. n ;:: 0 +fo..
1· ... ~ 7~· . ! ..  !. ~. ++.... ...
f. .. ~ 1"':.  ,. _ . _.. ,I _. '  ••. , __
~. '. .. ., .....:iI,.....___.., I I _ 'I. 1_ ~ ....... ~ ..
: . .   .. 7""'" 1 '  : . ~..  . _.........,..I!+. .+.I
., I    ". •• • • 'I I _. I ~ . •. "I _.
05
1'0
15
2·0
FIG. 3J~t Sorption curves for a. composite semiinfinite medium. Numbers on curves are values of Dl/DI,4
Following Carslaw and Jaeger [1, P4 263]~ for very large times the exponentials in (3.57) may all be replaced by unity. This is true also of tho errorfunction complements in (3.57) and so for large times we have approximately
_~~ ~ 2(Dlt)l(I+~~.) __ 4ct ,
lCo '1Tl2 1 ~ct (1 ~ (};)2
provided Ct"2 < 1. Fig. 3.8 shows M,/IOo as a function of (Dl t/12)l.
Whipple [7] has given formulae for the concentration in a semiinfinite region of low diffusion coefficient bisected by a thin welldiffusing slab, at different times after the boundary of the semiinfinite region has been raised suddenly from zero to unit concentration. 'This is of interest in grain boundary diffusion.
(3.58)
D t1 7
r N _l4" I ~l TE A:!'J D S E~I I· I NFL X ITE l\lEDI A
41
REFER.EXCES
1. Carsla\v~ II. S., and Jaeger, .r. C., Conduction of Heat 1~n Solid« (Oxford, 19(7).
2. British Association. M athcmai ical Tables, voL 6.
34 Ridesl) E. K., and Tadayon, J~, Proe, R01j4 Soc, 225 A (1954) 357.
4. Milne Thomson J L4 I\J ~t and Comrie, L4 L, Standard Fourfigure Jl athematical
Tables [Macmillan, 1944).
5. .Iost, ""T.:! Diffusion in Solids, Liquids, Gase« {Academic Press, 1952}~
6. Barret" R. ::,\:1., Diffusion in. and through Solids (Cambridge, 19(1)4
7. Whipple, R. rr, P., Phil. J.l1.ag~ (7) 45 (1954) t225t
D
IV
DIF~~USION IN A PLJ.~NE· SHEET
4.1" Introduction
IN this chapter we consider various cases of onedimensional diffusion in a medium bounded by two parallel planes, e.g. the planes at x == 0, x = l, These will apply in practice to diffusion into a plane sheet of material so thin that effectively all the diffusing substance enters through the p1ane faces and a negligible amount through the edges.
4.2. Steady state
Consider the case of diffusion through a plane sheet or membrane of thickness l and diffusion coefficient D, whose surfaces, x == O~ x == l, are maintained at constant concentrations 01, C2 respectively. After a time, a steady state is reached in which the concentration remains constant at all points of the sheet. The diffusion equation in one dimension then reduces to
d2C/dx2 == 0,
(4.1 )
provided the diffusion coefficient, D) is constant. On integrating with respect to x we have dC/dx == constant, (4.2)
and by a further integration we have, on introducing the conditions at x = 0, x = If
001 X
O201 ~ [4
(4.3)
Both (4.2) and (4.3) show that the concentration changes linearly from 01 to O2 through the sheet. AIso~ the rate of transfer of diffusing substance is the same across all sections of the membrane and is given by
(4.4)
If the thickness, 1, and the surface concentrations Cl~ O2 are known, D can be deduced from an observed value of F by using (4.4). Experimental arrangements for measuring D in this and other ways have been reviewed by Newns [1].
If the surface x ;;::;;: 0 is maintained at a constant concentration Cv and at x = l there is evaporation into an atmosphere for which the equilibrium concentration just within the surface is 02' so that
x == I ,
§ 4,2]
DIFFUSION IN A PLANE SHEET
43
then we find
C01 hx
=  ... ~.,
02C1 l+hl
F ::= Dh(Ol ~C2) ~ l+hl
and
If the surface conditions are
aC/cx+h1(C1C) == O, X = 0; 8C/ox+h2(CC2) ~ 0, x == l,
then 0 == hI GI!l +h2(Ix)}~: h2 O2(1 +hl xl, (4.3 b)
h1+h2+h1h2l
and
(4,4 b)
4.21. IJermeability constant
In some practical systems, the surface concentrations 01, C2 may not be known but on1y the gas or vapour pressures PI' P2 on the two sides of the membrane. The rate of transfer in the steady state is then some
times written
(4,5)
and the constant P is referred to as the permeability constant, Here P is expressed, for example, as e.c. gas at some standard temperature and pressure passing per second through I em. 2 of the surface of a membrane 1 em. thick when the pressure difference across the membrane is 1 em, of mercury. The permeability constant is a much less fundamental constant than the diffusion coefficient which is expressed in units such as cm4 '2 [eec., particularly as different investiga tors use clifIeren t units and even different definitions of P,
If the diffusion coefficient is constant, and if the sorption isotherm is linear I i.e, if there is a linear relationship between the external vapour pressure and the corresponding equilibrium concentration within the membranet then equations (4,4) and (4.5) are equivalent, but not otherwise. The linear isotherm may be written
o ~.:= J..9p, (4.6)
where C is the concentration within the material of the membrane in equilibrium with an external vapour pressure p, and /3 is the solubility. Since al~ PI and 02' P2. in (4.4) and (4.5) are connected by (4.6) it follows that, with due regard to units,
P = Dl9. (4.7)
4.22" 0 oncentrationdependent d·iff usion. coefficient
If the diffusion coeffieien t varies wi th coneen tra t.ion it is clear that the simple value of D deduced from a measurement of the steady rate
DIFFUSION IN A PLANE SHEET
[§ 4.2
44
of flow is some kind of mean value over the range of concentration involved. 'rhus, if D is a function of G (4.1) is to be replaced by
!!_ (D '!SJ_) === 0,
dx dx
and hence the relationship
F ~ DdCldx == constant
J
(4.9)
still holds, as of course it. must in the steady state. Integrating between 01 and 02" the t\VO surface concentrations, we have
C2
F === : f D de = DACIC2}/l~
(4.10)
where
(4.11 )
and this is the mean value deduced from a measurement of F. It follows from (4Lfi) that if D depends on C the concentration no longer depends linearly 011 distance through the membrane, Concentration distributions for D depending on C in a 'number of ways are given in Chapter XII.
4.23. Composite membrane
I f we have a  composite membrane composed of n sheets of th icknesses 11, l2'H~' t.. and diffusion coefficients D}, D2"~" o; the fall in concentration through the whole membrane is the sum of the falls through the com.ponent sheets. Since the rate of transfer, F, is the same across each section, the total drop in coneen tration is
n, Fl2 n, R R R F
n+]f+···+ 15 == ( 1+ 2+·'·+ n) ,
1 2; n
where R} === 11/ D], etc., may be termed formally the resistance to diffusion of each sheet. Thus the resistance to diffusion of the whole membrane is simply the sum of the resistances of the separate layers, assuming that there are no barriers to diffusion between them.
(4.12)
4.3. Nonsteady state
All the solutions presented here can be obtained either by the method of separation of the variables or by the Laplace transform as described in § 2.4. Many of the results are quoted by Barrer [2], Carslaw and Jaeger [3]t Jost [4], and others. The emphasis here is on numerical evaluation ..
§ 4.3]
DIFFUSION IN A PLANE SH~EET
4.31. Surface concentrations constant. Initial distribution f(x)
If
C === 01, X = 0, t ~ O~ ( 4.13)
C == C2, X :::::::;: l~ .. t ~ 0, (4.14 )
C == j(x), o < x < 1,. t ~ ()~ (4~ 15) the solution in the form of a trigonometrioal series is
r « _ 0 +(0 C )x+2 ~ C2cos·n1TC1 ,~ n7TX Dn'1T~tll'.i+
\J ~ 1 2 1 7 7T ~ ~n·· sm . t ' I
1
0') 1
+ 2" . n".x lht~llln';J J f( '). U7TX' d r
 L.., Sln  e I x Sln ._ .r L
l 1 l l
o
In the cases of most common occurrence !(x) is either zero or constant so that the integral in (4~ 16) is readily eval uated. Very often the pro blem is symmetrical about the central plane of the sheet, and the formulae are then most convenient if this is taken as x = 0 and the surfaces at x === ±l.
4.32 . Uniform initial distrib ution, Surface concentrations equal
This is the case of sorption and desorption by a membrane, If the region ~l < x < l is initially at a uniform concentration Co' and the surfaces are kept at a constant concentration Cit the solution (4~ 16)
becomes .
OOf! = 1 _ 4 ~ ~_l )" elJ(!"tl).".'f/4l" cos (2n+ I )1TX. (4.17)
01"'00 1T L, 2n+ 1 21
n=O
lf .il~ denotes the total amount of diffusing substance which has entered t·he sheet at time t, and .il1c;tJ the corresponding quantity after infinite time, then
J.I, == 1 _ ~ 8 eD(2n+l)~1t1114l •
. M~ L.., (2n+l)2172 •
71=0
The corresponding solutions useful for small times are
001} = ~ (I)"erfc (2n+l)lx _ ~ (1)" erfc (2n+I)l+x
C1CO L.. '2.J(Dt) t L., 2"./(Dt) I
n~O n=O
( 4.19)
11ft (Dt) i{ ~ 2:OC: . nl }
 = 2  11  + 2 ( ~ 1)1~ ierfc _ .
M; 12 n== I J{Dt)
The modifications to these expressions for Co == 0 or 91 = 0 are obvious.
and
(4.20)
DIFFlJSIOK IN A PL ... \NR SHEET
u 4.3
46
It is cl ear 1 h at e x pre ssi 0 n s (4. I 7 ), (4 + I H), (4 ~ 1 U), (4. 20) c an be wri tte n in terms of the dimensionless parameters
v  x Il
.... 1..  I,
(4.2l)
1·0 _ = .... _.~~. _ _. ~ ,_ ~ I J l
... '"
cc
o
. _._
C .. Co
t
0'8
0·2
1·0
o
FIG. 4,1. Concentration distributions at various times in the sheet ~l < x < l with initial uniform concentration Co and. surface concentratiun. Cp Numbers on curves are values of Dtfl'2+
so that the solutions for all values of I), 1, I, and z can be obtained from graphs or tabulated values covering these two pararnetere. Graphs of (C' CO)/'(C1CO) are shown for various times in Fig. 4.1. These are reproduced with change of nomenclature from Carslaw and Jaeger's book (3~ p. S3]. Tabulated values of (C'CO)/·(G11 ~Co) and of .J.'/,/JJXJ are given by Henry [5J v allies of J~/.Jl'J) have also been ta bu1ated by l\lcKay [6] and extensive numerical values for the concentration at the centre of the sheet, x ==' 0, are given by Olson and Schulz [7]. The curve labelled zero fractional uptake in Fig. 4.0 shows how ~11d.l~1':,C varies with t he square root. of time in a sheet of t.h i cknesa 2n w hen the concentration at each surface remains constant.
§ 4.3]
DIFFUSION IN A PLANE SHEET
47
4.33. Umijorni initial distrilnuion, Surface concentrations different This is the case of flow t.hrough a mern brane, If one face, x = 0, of a nlclubranc is kept at a constant concentration 01 and the other) x = l, at C2, and the mem brane is initially at it uniform concentration COJ there is 'a finite interval of time during which the steadystate condition previotlsly discussed in § 4.2 is set up. During this time the concentration changes according to
') ce· C C
(T _ rr + (C c ) x + .J 2: 2 cos n'TT 1 • nn_x_ Dn!17~IIJI+
.  VI 2 1   _. .  Sln .. _. e
l w n I
n :··1
+ 4l?1! ~ 1 sin (2m+ I )1TX ell(2m+ 1)'".'1/1'. (4.22)
1t L 21U+l l
·m=O
As t approaches infinity the terms involving the exponentials vanish and VlC have simply the linear concentration distribution obtained in § 4.2. If .J~ denotes the total amount of diffusing substance which enters the sheet during time t, and _JJ~ the corresponding amount during infinite time, then
(4.23)
In this case u; = l(OJ ~C2_CO) and the total content of the membrane at time t is given by M, + lCo. The ex pression (4. 23) is similar to (4·.18) and is readily evaluated from the tabulated values in references [5] and L 6] or from the curve labelled zero fractional uptake in Fig. 4.6, with the proviso that in (4.2·3) l signifies the whole thickness of the membrane but in (4.18) it denotes the half thickness.
The rate at which the gas or other diffusing substance emerges from unit area of the face x ~ 0 of the membrane is given by D{c'(}/ex)..r=_o, which is easily deduced from (4.22). By integrating then with respect to t, we obtain the total amount of diffusing substance, Qf) which has pasRed throng h the morn brane in time t, w here
o, === D(C~~Cl} 11+~! , ... 92 cosn21TC1 (l_e~Dn:;1Tttll~) +
'7T ~ n
1
+ 4C(ll ~ I (lell(2m i1l'","//'). (4.U)
1T2 ~ (2'1n+ 1)2
m=f)
~n the commonest. experimental arrangement both Co and 01 are zero, i.e, ~he tuembrane is initially at zero concentration and the concentration
DIFFUSION IN _A PLANE SHEET
[§ 4.3
48
at the face through which diffusing substance emerges is maintained effectively at zero concentration. In this case (4~24) reduces to
Q, == Dt~~~! ~ (l)n elh''''7fltJ~,
lC'J, l2 6 772. L n2
1
(4.25)
o
~
  ~  ...
10
i"~  ..
J
+
f.j... 1.1

~
f"' f. r' l
I
~
U'

•
~. 02
01
o
01
02
0)
04
05
06
DIN2 }'IG~ 4~2. Approach toateadyat.ate flow through a plane sheet.
which, as t ;:.. 00, approaches the line
Q, = DC2(t~~).
l 6D
(4.26)
This has an intercept, L, on the t axis given by L = l2/6D.
( 4.27)
Following Daynes [8], Barrer [2, p. 19] has used (4.27) as the basis of a method of obtaining the diffusion constant, the permeability constant, and the solubility of a gas by analysing stationary and nonstationary flow through a membrane. Thus from an observation of the intercept, L~ Dis deduced by (4.27); from the steadystate flow rate the permeability constant P is deduced by using (4.5), and S follows from (4.7)4
DIFFUSION IN A PLANE SHEET
49
A graph of QdlCz as a function Dtll" is shown fOT the case Co == Ct ==== 0 in Fig. 4.2. To within the accuracy of plotting the steady state is achieved v,rhen Dt/'l2 == 0·45 approximately.
4.34. ~T ariable surface concentration
The solution to the general problem of diffusion in the region 0 < x < I wit}l the surfaces at' concentrations q;1 (t) and 7>2( t} and the initial coneen ~ trat,ion f(x} is given by Carslaw and Jaeger [3, p. 86]. For empirical values of rpl (t), cfo2(t), and f(x)J three integrals arise which have to be evaluated graphicalJy or numerically, In certain cases, however, where the surface concentration can be represented by a mathematical expression, the solution can be considerably simplified.
(i) One case of practical interest is that of a sheet in which the concentration is initially zero and each surface of which approaches an equilibri urn concentration, Co, exponentially ~ i.e.
rpl(t) == rp2(t) = Co(le~'}. (4.28)
This can represent a surface concentration which is changed rapidly but not instantaneously, a situation which usually arises when an instantaneous change is attempted in an experiment. For the sheet whose surfaces are at ±l the solution is
C == ]_e_f*cosx(~/D)t_
G~o cos l(pjD)&
] 6{3l2 ~ (~l )"neD(2n+l):!17~ii4J2 (2n+ I )1TX
~ 7T L.. (2n + 1)[4,8[2 D7T2(2n+ 1 )2{OS 2l ,( 4.29)
n.=o
provided f3 is not equal to any of the values D (2n + 1 )27T2/ 4l2• The sorptiontime curve, i.e. the curve showing the total amount, 1J~, of diffusing substance in the sheet as a function of t ime It is obtained by integrating (.t~2U) with respect to x between the limits l and +1 and is
_AI.
2lG~ = le"~[lt(DIf3l2)l tan (JJ/f3l2)l ~
8 ~ e~{2n+l):l:7T"2D1I41:a
 n2 .6 (2n+ 1)2[1':_(2n+ 1)2{D112(4,Bl2)}]" (4.30)
Fig. 443 shows uptake curves for different values of the parameter f312/D pJotted against (Dt/l?)'. \Vhen f3 === oo, the surface concentration rises
.
Instantaneously to Co and the curve of Fig. 4.3 'has the characteristic
initial linear portion followed by the approach to the equili brium value, _J~ ~ 21(}o' The uptake eUTYCS for finite values of jJ['2//D, for which the
1·0
0·8
06
0·4
02
o
50
DIF}~t:'SION IN A PLANE SHEET
(§ 4.3
surface concentrat.ion rises at a finite rate, all show points of inflexion. At first the rate of uptake increases as sorption proceeds but later
I .. j
r.'  r   . ~.~ ·In
 ..   .. I •
l r
. • :
 ~ L ,..
5·0 ·r ; ..

J ·0
 ·,Ii :
)·5 ~, ~I'" 

_j ~~~

Ir..


4
·t·
...:..
 ~f:
).}
~~

r .. ~
. i ~ .

.40 f.o. """ 1' r'  .
. .,
. ~ ,
"'7
0·01 2·0
3·0
4·0
Fro. 4.3. Calculated sorption curves for surface concentra.tion gi ven hy Co{ I ~ eP~ ,.
N umbers on curves are values of fil2f D.
decreases as the final equilibrium is approached, Curves of this kind are often referred to as sigmoid sorption curves. 'I'hey Ina}" arise in practice because surf ace equili bri um conditions are not esta blished instantaneously, but they Illay also result from other causes (see § ]:.? .6:! below),
(Ii) If the surface concentrations vary linearly with time, I.e.
(4.31 )
the solution is
pC ~ ~~ + ~(:r2 _ I) + If! ~ ( __ l)n crl(2"lj'1T'I,'~I' cos (2n+ 1 )7TX.
kl2 1'J. :? [2 113 (:1 n + I ):J il
n I)
(4.32)
.1 II
I'
2
}
I ~
r l.... i"" f
., L..o i'"
Lao i'"
r,... .....
._ Ir "'"
I.' Ir
~ """
1
.....011.'
l .... ~
L,... ~
I,.rol"' L r r
~
.........
l I'"

rr f l. I"
q·i I. ~
~
..... l I
........ t" . o·~1
... I~ ....
1.0
. 1.0 i'
L.
.... ~o· ]
,
1 DC

kl2
o
0,2
04
0·6
0,8
1·0
x/I
FfO~ 4.4. Concentration distributions in a plane sheet for surface concentration kt, Numbers OU CUf\"eS are values of /Jt/12.
4
v
.t/ ~f
/
[7
/
'7
1/
1/
[;i"
I'
._
~
. J
I'~
LI
I ....
V 
V
J
i.""
......
l,.. I 
.... .... , 3
.,.
....
1
o
0·5
t 0
1 5
20
DriP
FlU. 4, 5~ Sorption curve for plane sheet with 8U l'fnC"~ coueen t.rn tion kt:
DIFFL'"SION IN A PLANE SlIEET
[§ 4.3
62
The corresponding expression for .i.'~ is
Some numerical results are given by Williamson and Adam~ [9] and Gurney and Lurie [101. Fig. 4.4 shows DC/kl2 plotted as a function of x/I for various 'values of D(/12. Fig. 4.5 shows DJII/kl3 as a function of .the single variable DtJl2•
These solutions can be extended to cover modified surface conditions such as tPl(t) !;;;;; <P2(t) = Co+kt~ and a nonzero initial concentration by superposing solutions as in § 3.3 (v) 4
4.35. n·iffusion from a stirred solution of limited volume
If a plane sheet is suspended in a volume of solution so large that the amount of solute taken up by the sheet is a negligible fraction of the whole, and the solution is well stirred, then the concentration in the solution remains constant. If, however, there is only a limited volume of solution, the concentration of solute in the solution falls as solute enters the sheet. If the solution is well stirred the concentration in the solution depends only on time, and is determined essentially by the condition that the total amount of solute in the solution and in the sheet remains constant as diffusion proceeds. It is useful from an experimental point of view to have only a limited amount of solution since the rate of uptake of solute by the sheet can be deduced from observations of the uniform concentration in the solution. It is often simpler to do tllis than to observe directly the amount in the sheet, This has been stressed by Carman and Haul [Ll ], \\"'110 have written mathematical solutions in forms most appropriate for the measurement of diffusion coefficients by t.his method.
The general problem can be stated mathematically in terms of a solute diffusing Irorn a wellstirred solution. The modifications necessary for corresponding alternative problems, such as that ofa sheet suspended in a vapour, are obvious.
Suppose that an infinite sheet of uniform material of thickness 2l is placed in a solution and that the solute is allowed to diffuse into the sheet. The sheet occupies bhe space l < x ~ l, while the solution is of limited extent and occupies the spaces la ~ x ~ 1, l ~ x ~ 1 +a. The concentration of the solute in the 80]ut101l is always uniform and is initially Co, while initially the sheet is free from solute.
§ 4.3]
DIFFlTSIO:N I~ A PLAKE SHEET
53
\Ve req uire a solution of the diffusion eq ua tion BO c2C ~D
at ~ 8x2}
(4.34)
with the initial condition
C == 0,
l < x < t.
t == 0,
(4.35)
and with a boundary condition expressing the fact. that the rate a.t which solute leaves the solution is always equal to that at which it enters the sheet over the surfaces x == ± I. This condition is
a cC/of == ~D ac/ox,
x ~ ±l,
t > o.
( 4.36)
"'e assume here that the concentration of solute just within the surface of the sheet is the same as thaf in the solution. This may not be so but there may be a partition factor, K~ which is not unity) such that the concentration just within the sheet is K times that in the solution. This can clear1y be allowed for by using a modified length of solution, alK, in place of a in (4.36) and elsewhere.
A solu tion of this problem by 1v1 arch and "T ea ver [12], based on the use of an integral equation, was cumbersome for numerical evaluation. More convenient forms of solution have been obtained by Carslaw and Jaeger [3~ p. I06]~ ""ilson [1 :3], Berthier (14]~ while Crank [15J has developed solutions particularly suitable for small values of the timet The solution is most readily obtained by the use of the Laplace transform. In a form expressing the total amount of solute, _ft~! in the sheet at time t as a. fraction of .l~l':S,;, the corresponding quantity after infinite time, the sol u Lion is
( 4.37)
where the qn '8 are the nonzero positive roota of
tan qn ::::::;:;; ~a.qn'
( 4.38)
and (l: :::=::::;; all, the ratio of the volumes of solution and sheet, or if there
4
IS a partition factor K then (): == a/Xl. Some roots of (4.38) are given
in Table 4rl for values of a corresponding to several values of final fractional uptake. Roots for other values of (}; are given by Carslaw and Jaeger [3, p. 378] and by Carman and Haul [II]. ·It is sometimes convenient to express a. in terms of the fraction of total solute finally taken IIp by the sheet. Thus in the final equilibrium state, since the
DIFF1JSION IN A PL_4.NE SIIEET
{§ 4.3
total amount. of solute in solution and sheet was originally contained in the solution of concentration 00, we have
w here C07 is the uniform concentration in the sheet. finally. J.11oc ~ of the sheet finally is given by
..:.11 == 2lC = ._ 2aCo ~ == 2aC:iJ.
~ ~ l+a/(lK} l+n:
The fractional uptake of the sheet finaHy is therefore given by
11 1
0C1
The content,
( 4.40)
(4.41 )
~.
~") ..... rr ~t.«._...oo
If, for example, 50 per cent. of the solute initially in the solution is finally In the sheet, ~ == 1. In the particular case of an infinite amount of solute (ex == co) the roots of (4.38) are qn == (n+i)1T~ and we have
(4.42)
which is expression (4.18) for the case of a constant concentration, COt at. the surface of a sheet. The smaller Dt/l2 iSJ the more terms in the series in (4 .. 37) are needed for a given accuracy. When more than three or four terms are needed it is better to use an alternati ve form of solu tion, For most values of a the simplest expression is
Mt == (1 + ~ ){ 1 e TJ 0::1 erf c ( T I 0: 2) t} t ( 4.43 )
Mc;o
where T == Dt/l2. If very small values of 0: are required, corresponding to very high fractional uptakes of solute by the sheet, there may be a range of Dt/12 in which neither (4 .. 37) nor (4.43) is convenient but where the following is useful:
M, _ f _ (X ~ r/1 _ 3cx5 _ }
Moo  (I +a).l 71i1'. + 21T'Ti 41rlTt + ... ·
(4.44)
This is obtained from (4.43) by substituting the asymptotic expansion for exp( 7'tj(l2)erfc( T /ct2)~ when T/a2 is large ..
Fig. 4.6 gives curves showing J.I,IJIoo against (D(l[Z)I for five values of final fractional uptake. Fig. 4.6 shows that the greater the final fractional uptake of the sheet the faster is the solute removed from the solution.. Clearly by comparing the rate of fall of concentration in the solution observed ex perimentally , with the corresponding calculated
1
1
,
Cl
Cl
I
_j
o·
0·8
0·7
0·5
a
•
gl
01
vt
b( tl
Tl
Ul
.
In
§ 4.3J
curve showing .i~11Uoo as a function of Dt/12, the diffusion coetlicient 1) can be deduced. This has recently been suggested by Bcrthier [14] as
'·0
o·
1·1 (Dt/r)1
FIG. 4.6. Uptake by e. plano sheet, from 8 stirred sclution of limited volume. Numbers on curves show the percentage of total solute finally taken up by the sheet.
04
1.0
o·
0·2 0
a method for measuring selfdiffusion using radioactive isotopes. He gives a table of .1llz/.11fCZ) for values of l/a bet.",·ee.n 0 and 1·0 at intervals of 0·1. For precision measurements it is advisable to check Berthiers values as in some instances not enough terms of the series solutions have been retained to obtain the accuracy quoted. The concentration within the sheet is given by the expression
( Z«t 2( 1 +~)eDaAtll3 cos(qu X!"l)}
o == C~ 1+ .
l+a+ct2q! cosqn
n~l
(4.45)
We have considered diffusion into a plane sheet initially free of solute.
There is the complementary problem in which all the solute is initially uniformly distributed through the sheet and subsequently diffuses out into a wellstirred solution. It is easily seen that. the mathematical
DIFFUSIO:N IN .A PLANI~~ SHEET
116
solutions presented above for sorption by the sheet also describe desorption, provided .1lf: is taken to mean the amount of sol u te leaving the sheet up to time t, and .Jlrr, the corresponding amount after infinite time. For the problem of desorption from the sheet we require a solution of (4.34) satisfying (4.36) but with the initial condition (4.35) replaced by
C==Oo, l<x<l, t==04 (4.46)
On writing Ct = coc, (4.47)
(4.46) and the other equations for desorption are identical with (4.34)~ (4~35L (4.36), with 01 written for 04 Hence the equations and solutions for desorption are identical with those for sorption provided AI" _J{ c:o a~e suitably interpreted and C1, (C1)cc replace C, Co:; in expression (444.1). The parameter ex is equal to al Kl as before, but its relation to the final uptake of the sheet expressed by (4.40) and (4.41) no longer holds. Instead we have that the fractional uptake of the solution is given by
st; 1
2lCo = l+lja' (4A8)
~J acger and (~}arkc [) 6 J have presen ted solutions of a n urn ber of other pro hlems in diffusion from a wellstirred solution in terms of certain fundamental functions. Accurately drawn graphs of these functions" from which solutions of limited accuracy arc readilyconstructed, are given in their paper.
4t36. Surface evaporation
In § 3.31 the rate of loss of diffusing substance by evaporation from the surface of a sheet was represented by
D ec lax == o;(COC8)~ (4.49)
where C, is the actual concentration just. within the sheet and Co is the concentration required to maintain equilibrium with the surrounding atmosphere. If the sheet, I < x < l, is initially .at a uniform oonoent.ration C2, and the law of exchange of the type (4~49) holds on both surfaces, the solution is'
~(}C2 . I _ ~ 2L cos(~!I!/I)e~!Dt!_~
CO02 ;S {fi!+L2+L)cos.Bn '
where the f1J~' s are the positive roots of
f3 tan f3 ;;:::; L
( 4.50)
(4.51 ) ( 4~52)
and
a dimensionless parameter 4 Roots of (4 • .5]) are given in 'fable 4.2 for
§ 4.3]
DIFFUSION IN A PLAKE SHEET
57
several values of L. Roots for other values of L are given by Carslaw and Jaeger [3, p~ 37~J. The total amount of diffusing substance, Jl(, entering or leaving the sheet up to time t, depending on whether 00 is great.er or less than C2,! is expressed as a fraction of Llfcc;, the corresponding quantity after infinite timet by
n. ~ 2 L2 e  m. Dill'!
.tlf~ = 1 L ,s!(P!+L2+L)·
n=l
( 4.53)
0·2
,
.

.~
n'n ~
1
~ 
.... bi n'5
~ .
l
.j...
t. J  H~ "1
,
 .
..1

~
'10,
•
I!
I 0·6
0·4
o
1
2
3
4
5
6
FIGr 4~ 7. Sorption or desorption curves for the surface condition (4.49). Numbers on curves are values of L = l~/D.
A solution suitable for small values of time may be obtained in the usual way by expanding the expression for the Laplace transform in a series of negative exponentials (Carslaw and Jaeger [3t p. 252]). The terms in the series expression for concentration very soon become cumbersome for numerical evaluation] however. In practice, it is usually sufficient to use only the leading terms corresponding to the interval during which the sheet is effectively semiinfinite, when the concentration is given by expression (3.35), and (3.37) gives the value of .L~~ for half the sheet.
Graphs showing _J~/Jl~ for several values of L are plotted in Fig. 4.7 from numerical values given by Newman [17]. Carslaw and ~J aeger [3, p. 102] give corresponding curves as well as others showing how the concentrations at the surfaces and the centre of the sheet VaT)"" ·with time. Newman [17] also gives a table of values from which .1.1t,/_J!oc can easiJy be deduced for a parabolic initial distribution instead of a uniform
DIFFl~SION IX _.\ PLA~E S}IEET
f§ 4.3
58
initial concentration. All these equations and solutions have a practical application in the drying of porous solids.
Carslaw and Jaeger [3~ p. OS] give solutions of more general problems in w hieh there is an initial ~ nonuniform coneentra tion d istribu tion, or in which the vapour pressure is different. on the two surfaces of the sheet, or in which evapors.tion occurs from one surface only, the other being maintained at a. constant concentration. Jaeger and Clarke [16] have also given in gra.phical form the solutions of a num her of problems with an evaporation t.ype of boundary condition.
The more complicated case in which the rate of transfer on the surface is proportional to some power of the surface concentration has been discussed by Jaeger [18].
4.37. Constani ~fillX Fo at the surfaces
If the sheet, 1 < x < 1 ~ is ini tially at a constant con cen tra tion CO~ and diffusing substance enters at a constant rate, Fo~ over unit area of each surface, i.o.
x==l
,
(4.54)
then
c,co = !P.!{Dt + 3X2.~~_ 2 ~ (l)n elJn2TT'W/~cosn7TX}
D fA 612 '1T2 ~ n2 l ·
n=l
( 4.55)
Obviously the total amount of diffusing substance taken up by unit area of the sheet in time t is 2Fo t. Concentrationdistance curves for various times are shown in Fig. 4.8. An alternative form of solution suitable for small times is given by Macey [19l
4 .. 38~ Impermeable surfaces
An impermeable surface is one at which ·the concentration gradient is zero. This condition holds at the central plane of a sheet. provided the initial and boundary conditions are symmetrical about that plane. It follows therefore that the symmet.rical solutions already given for the plane sheet occupying the region l < x < l apply also to the sheet 0 < x < 1 when the face x = 0 is impermeable. If, on the other hand, both surfaces x == 0 and x = 1 are impermeable and the initial distribution is f(x), the solution is
1 00 l
C = I f f(xl) dx' + 2 ~ eDn~07~iJl~ cos ntrx f f(xr) cos n::~~ da'
l l L l i >:
0 n=l 0 .
( 4.56)
. I:~ ..
§ 4.3J
DIFFUSIOX IN ,..\ PLA~B SHEET
Barrer l2, p. 14] suggests that diffusion from one layer to another as discussed in § 2.24 may be treated by regarding the system as a single
1,4
1·0
.. ··1+ r+ 1. 1 I I
l J 1
[ ,
 I
Ir ~
.
.'
L.<'
..I
V
~
. '"
 ~
 '.,01
·r 0 jl'
. ~
L
~ ·1
~" r
.. L..o ......

..... "'"
..
.I ....
_ ;::;;
I .....
./
l./'
/
r : ,/
"
".
~
'"
"
0·5 L'
f
.. 
 v ~
~ 1..1
I ~ :.J
l,. II
I.' V'
~
.... L,..o
~I ,.... V
l.I
t ~ ~
"1 .... I .....
0·25 iI=':: 1.1
L,.o' V
i..t'"
~ 1/
V
~ ~
~ ... /
.... ""'"
L ,_.. .L
0·1
;...
__ ....
,.
 ~ ... o (CC )
. a
f~1
O·R
0·6
0,4
02
o
0,2
0,4
0·6
o·g
XI"
j L
1,0
Flo. 4.8, Conccntrabion dist.ribu tiona in a. plane sheet for constant flux }"to at tho surface. Numbers on curves arc values of nt/rio.
layer with impermeable boundaries and applying (4.56) where the initial distribution is
fix) == CO~ o < x < h). (4.57)
f{x) == 0, h < x < l
The solution (4.56) becomes
G  C (h + 2 La) I . n7Th Dn'J1Tl.f!P' n7TX 1
 0  _. .. Sln· e COS~L
1 71 n I I J
?l''l
DIFFUSION IN A PLANE SHEET
[§ 4.3
60
This solution is complementary to expression (2.17) which is convenient for small times. Numerical values based on (2.17) and (4.58) arc available _. in the wellknown tables by Stefan [20] and Kawalki [21]~ some of which are reproduced by J ost [4, p. 63 J.
Another special case of ( 4 + 56) has been ev al Hated by Crank and Henry [22] in an Investigation of different methods of conditioning a sheet to a required uniform concentration, They consider the problem in which a sheet, initially at zero co~cen tra tion throughout, has i ts surfaces maintained at a. constant concentration Cfo for a time lOt after which they are rendered im permea ble. The subsequent change in concentra tion is described by (4.56) withf(x) given by
f( ) _ r» [1 4 2:00 1 . (2m+ I )1TX 1J(2m+l)<111T~/ ret]
X  vo  ~ .... SIn e 01 •
11 2m+l l
m=O
( 4.59)
In this case the two integrals in (4: .56 ) reduce to
l [8 00 I
f f(x) dx = C 1 1  ~ ~ eD('2m+l}11'I'~lOl12]
o n2 ~ (im+I)Z '
o m=O
( 4.60)
and
(4.61 )
where 2p is substituted for n since only the terms involving even narc nonzero .
. 4.39. Composite sheet
Various problems of diffusion into a composite sheet comprised of two layers for which the diffusion coefficients are different have been' solved (see, for example, Carslaw and .Jaeger [3~ pp. 265J 302]). The solutions arc similar in form to those presented in this chapter but obviously more complicated. In view of the additional number of parameters involved, no attempt is made to give numerical results here. Jaeger [23] has studied the timelag involved in establishing the steadystate flow through a composite sheet of several layers.
REFERENCES
1. Newns, A. c., J. Textile In.st. 41 {1950} T269.
2. Harrer, R. M.~ Diffusion. in and through Solids (Cambridge, 1941).
3. Carslaw, H. S~, and .Iaeger, J. C.j Conduction. oj Heat in Salida (Oxford, 1947).
4. J ost, W., Diffusion in l)olids, Liquids, GMe8 (Aeadomie ~rcsp., 1952}.
5. Henry, P .. S. H'J Proe. Roy. Soc. 171 A (1939) 215.
DIFFl.TSIO~ I~ A PLANE SHEET
6. McKay~ A. T.~ Proe, PhY8. Soc. 42 (1930) 547.
7. Olson~ F. C. w., and Schulz, O. 'f., Ind. Eng. Chem, 34 (1942) 874.
8. Daynes, n., Proc, Roy. Soc, 97 A {1920} ,286.
9. 'Villiamson~ E. n, and Adams, IJ. H.~ l:Jhys. Rev~ (2) 14 (1919) 99.
10. Gurney t H. P.j and Lurie, J.~ I ltd. Eng. Chem. 15 (1923) 1170.
II. Carman, P. C.~ and Haut R·. A. w., Proc. Roy. Soc. 222 A (1954) 109. 12~ March, II.~ and Weaver, '\1".,. l:Jhys. Rev. 31 (1928) 1081.
13~ \Vilson~ A. n., Phil. Mag. 39 (1948) 48.
14. Berthier, o., J. Chim. Phys. 49 (1952) 527.
15. Crank, J., Phil, J.~I ago 39 {19~8) 362.
16. Jaeger, J. C., and Clarke, ~f., ibid. 3B (1947) 504.
17. Newman, A. B., Trans, A1ner~ [.t18/. ()hem· .. Engrs. 27 (1931) 203.
18. Jaeger, .T. C., Proe. Camb. riu. Soc. 46 (1950) 634.
19. ?ti~ey, .H~ H., Proc, Phys. Soc, 52 (1940) 625.
20. Stefan, J." Sitzunqsber, lViener Akad. lVisBensch. I It 79 {1879) 161.
21. Kawalki, \V., IVild. Ann. 52 {1894) 166.
22. Crank, L, and Henry ~ !\of. E., Proe. PhY8 L Soc. 42 B (1949) 257. 23 .. Iaeger, J .. C~t Quart .. J, Appl. .llath. 8 (1950) 187.
6)
5.1. Introduction
\\l E consider a long circular cylinder in which diffusion is everywhere radial. Concentration is then a function of radius, r, (bud t.inie, t, only, and the <Ii ffusion equation (I. 7) becomes
c .. o 1.(.1 ( [ C·)
 ==   rD .
()t r cr (1
(5.1 )
5.2. Steady state
J f the medium is a. hollow cylinder w hose inner and outer radii are a and b respectively, and if the diffusion coefficient is constant, the equation describing the steadystate condition is
d( dO) _.
dr r dr =::: U)
a < r < b.
The general solution of this is
(} ~ A+Blogr,
•
where A and Bare constants to be determined from the boundary conditions at r =.: a, r === h. If tile surface l' ~:_..:; a is kept at a constant concentration 0., and r == b at (.."12' then
C == C1log(h,/r) +C2 ~,o~(r/a) . (5.4)
log(b/a)
The quantity of diffusing substance, Qt~ which diffuses through unit 1engt 11 (,f the cyi inder ill ti me t J is given by
fJ _ 277 Dt(Co ~ .. ('1)
~{";[  " . (5.5)
, log(h .. /a}
If Qf is measured in a concentrationdependent SYStCTI1, the mean value C2
of the uur usi on cocff cien t 0 btai ned from (5.5) is (~ D de) / (02  (}1) as
Ll
for the plane sheet (see .t.l1). The concentration distribution defined by
( 5.4) is } lot lineat, as it is for the plane sheet. Ty pical distri bu tions are shov ... ·n in Fig, 5.1 for the cases 02 ::. 0, b/a. ;_~.; 2, 5~ 10 .
.. Another steadystate problem leading to an interesting result is that of the hollow cylinder whose surface r == a is kept at a constant concentration 01 ~ and at r ;;;:::;: b there is evaporation into an atmosphere for
§ 5.1]
DIFF U 8IO N I N ~.:\ C,\YL I N D E R
63
which the equilibrium concentration jnst within the surface is ()2' The boundary condition, with the constant of proportionality denoted by h, i~
and we find
de +h(C()2) == 0, dr
(7 == Cl{_l_t~j~_ ]og(blr)}+~hbC~}og(~.~iia.~. 1 +hb Jog(r,/a)
(G.7)
(5.G)
l'O~~~~~~~~~~~~
c.c,
0'4
3
4
5
8
6
7
FIG. 5.1 + Steady state coneentraticn distributions through cylinder wu.ll, K umbers on curves are values uf b/a.
The ou tw ard rate of cliff usion per unit length of the cy linder is Qt~ '"'~ here hb
Qt == 21TD(C1C2) hb 1 .
1 + og(b/a)
( 5.8)
By differentiating this expression with respect to b, it is easily seen that if aJt > 1 the rate of diffusion decreases steadily as b increases from a, but if aft < 1 the fate first increases and. later decreases, passing through a maximum when b == l/h~ This is due to the two opposing changes associated with an increase in b. On the one hand) the rate of evaporation is increased because of the increase in area of the surface, r == b, as b Increases, but on the other hand, the gradient of concentration through the cylin der decreases as b is increased, In certain circumstances, therefore, the rate of diffusion through the wall of a pipe nla.y be increased by making the wall thicker [I]. This is illustrated in Fig. 5.2 for ah  .!.
If the surface condi tions are
(5.fJ a)
DIFFUSION IN A CYLINDER
[§ ,5.2
64
c == ahl C}(I+bhzlogb/r)+bh2Cz(1 +ah1log ria),
ah, + bh2 + abh ; hzlog b [a .
Q ~ 27TDtabht h2(C2(}I)
·t ah1+bhs+abh1h'}.logbJa
we find
(5.7 a)
and
(5.8 a)
o·so.._ ._ ....... ....&... ......... .....
10
I· 5
b/a
35
2·5
FlO_ 5.2. Effect of thickness of cylinder wall on steadystate rate of flow.
Other problems on diffusion in regions bounded by surfaces of the cylindrical coordinate system and in which the flow is not necessarily radial are treated by Carsla "'~ and Jaeger [2, P4 1 ~ 7].
5.3. Nonsteady state. Solid cylinder
Following essentially the meth ad of separa t.ing the variables described in § 2.3t we sec that
(/ ;:;:::: 1. te  P x ~J
(5.9)
is a sol uti 011 of (5. I) for D constan t provided u is a function of r only,
satisfying d2u I du 2;
dr2 + r d:r + ll: 1t =:; O, (5.10)
which is Be8sel~s equation of order zero. Solutions of (5.10) may be obtained in terms of Bessel functions, suitably chosen so that the initial
DIFFeSIO~ I)l A CYLIXDE.H.
tiil
and boundary conditions are satisfied. Thus if the initial concentration distribution is j(r) and the surface r == a is maintained at zero concentration~ a solution of (5~ 1) is wan ted sa t.isfying
o == 0, r ~ at t ? 0, (5.11)
C == 1(r),
o < r < a,
t == o.
(5,) 2)
The boundary condition (~) .11) is satisfied by,·
00 ;
C == '" A J. (a r)elJLr.: I
£., nO n ~
n=l
(5.13)
provided the ~n's are roots of
Jo (aa.,J == 0,
(5.14)
where Jo(x) is the Bessel function of the first kind of order zero. Roots of (5.14) are tabulated in tables of Bessel functions (see for example reference [3], p. 171). For this function 0 is finite at r == O. The initial condition (5. 12) becomes
00
f(r) == LAn Ju(rLl!·n)J
ft=l
(5.15)
it being assumed that f(r) can be expanded in a series of Bessel functions of order zero. The A,...'s are determined by multiplying both sides of (5.15) by rJo(ctn r) and integrating from 0 to a using the results,
a
f r~(ctr}Jo(fJr) dt == 0,
..
e
when Ci and f3 arc different roots of (5.14), and
(5.16 )
a
J r{Jo(CtTl}2 dr = !aIlJ!(a.:xn),
o
where Jt(x) is the Bessel function of the first order and a is a root of (5.14). The derivation of the relationships (5.16) and (5.17)j and of corresponding expressions which hold when a is a root not of (5 .14 ) but of alternative equations which commonly arise in diffusion problems, is given by Carslaw and Jaeger [2j p. 172]. Finally the solution sat.isfying ( 5 .11 ) and (5.12) is
(5.17)
(5. 18)
Alt.ernati vely, solutions for both large and SOl all t i Illes can be 0 bt ained by use of the Laplace transform,
ls 5.3
66
5 ~3 t ~ ~,,1 urface concentration constant. Initial distribution f( r) I f in ths eyl inder of radius a the conditions are
C = G~, r == a ~ t ;;: 0,
(5.19) (5~2(})
C === !(r), 0 < r < a, t == 0,
t he solution 18
where the a.ft's are the positive roots of (~).14).
If t.he concerrtrabion is initially uniform throughout the cylinder, f(r) = 01 and (li.tI) reduces to
001 == 1 ~ ~ efu,VJo(ra,,). (5.22)
CUC1 a L (XnJl(aan)
n,:::!:!:::l
If J.l~ denotes the quantity of diffusing substance which has entered or 1eft the cylinder in time t and J.ll~. the corresponding quantity after infini te time, then
(5.23)
The eorres pondi ng s01 ution useful for small times is
(!~Cl ai (Lr (ar)(Dta)l. ar
 ..  == erfc ~ "+~, .ierfo ~ +
(lO· . (}1 'rt i\/{ IJl) 4ar1 2~(Dt) .
{9a2.7r22ar)Dt'2 ar
+ 32a1rt 1 erfc 'l.,l(jjt) + ... , (5.24)
which 110]U8 provided ria is not. small, The case of rla small is discussed by Carsten and ]\·1cKerrow (4]. They give a series solution involving modified Bes~el functions of order n±l. "fhe necessary functions are tabulated in t hell' paper and numerical calculation is straightforward.
.i\]SO for small times we have
J~ == i(Dt)! _ P! __ I _(Dt)~ + ....
Jla:; 17' a2 a2 37Tt a2
Clea rly these sol n t ions are not as val ua ble as the corresponding o~es for the plane sheet. In practice the range in t over which they are convenient fur evuluation is less than in the plane case.
(5.25)
* 5.31
DIFFUSION IN A CYLI:;JDER
67
The solutions for the cylinder can be writ ten i 11 terms of the t\VO dimensionless parameters Dt,/a2 and ria. Curves showing (CC1 )!(Co~ C1)
_~~c] ..
C C'
(} 1
0·4 r/a
FIG~ 5.3. Concentration distributions at various t.imos with initial concentration C1 and surface eonecnt.ration ego )l' um hers 011 curves are values of nt/a'}..
a
O~6
0·8
0·2.
as a function of ria for different values of Dt/a2 drawn by Carslaw and Jaeger [2, p, 175] are reproduced in Fig. 5'~4 The curve of Fig. 5.7 for zero fractiona1 u ptake shows how .J11/ illXJ depends on Dt/a2 when the concent,Tution at the surface of the cylinder remains constant.
5.32. Variable sur face concentration
If the ini bia] conecn tration in tho cy finder is zero and that at the surface is q;(t)J the solution is
. ~ l
C' = 2D_ .2 cD"'~t (_':!' .,~)(~~,,) J en('(~ 'Aq_,()..) o, (5.2 fj)
a tJ1 (a·~'lll )
n=l 0
where the O::n's arc the roots of (5.14).
(i) As for the plane sheet" a case of practical interest is when
~ (t) == C~ ( I  e  fit), ( 5 . 27 )
0·8
0·0
0·4
0·2
68
DIFFUSION IN A CltLINDER
[§ 5.3
representing a surface concentration which approaches a steady value, (}o, but not instantaneously. The solution (5.26) then becomes
o
_,; ,,,,
t )
.~
~.
I.
t ..
~. II
· ... fJ
I
(5.28)
~.
~ +
~
4
( Dt/tJ)!
FlO. [).4r Calculated sorption curves for surface conecnfrabion given by Oo( 1  e.fit)r Numbers on curves are: values uf (3afl./LJr
:0·1
..
j_
••
nmml ••
~~j : +:ttlj±tltt:1.i::tl
+
.... ,~
............ .j,...j,..i.. ~
I~
2
3
and the sorpt.iontimc curve is given by
( .5.29)
5
l~~ig. ;").4 shows uptake curves for different values of the parameter f3a~/J). (ii) If the surface concentration varies linearly wit h time, i.e .
.p(t) ~ kt,
( 5430)
.....
70
DIFF1:S10N 1).1" A CYLI:';VEI{
[§ 5.3
t.he solution is
C k( a.2~ r2) 2k L~ D~~ I ~lfl(riXn)
=== t  +   e ~ a~· ·
4D alJ O::n rfl (aex.,~ )
n=l
In Fig. 5.51 curves showing DC/(ka2) are drawn against ria for different
values of Dtja". Numerical values are given by Williamson and Adams [ 5]. The corresponding expression for J.lf, is
. ~ TTka4 4rrk ~ e.net!'
illt == Trka""t  D + ~ D~  L ~.  ,
H ~n
n···l
(5.31 )
(5.3:!)
!.
I' ++++++.111+++++++4
and j_lJtD/(7Tka<1) is plotted against Dtlo? in Fig. 5.0. These solutions can be extended by superposition as in § 3~!3 (v}, to cover modified surface condi ti ons such as 1>( f) :::;;;; Co + let and a non zero ini ti al concentration.
5.33. Diffu.sion frorn (1. stirred solution oj limited. volume
The problem differs only in detail from the corresponding problem considered for the l)lanr sheet in § 4.3G and the results can be written down \\or j thout l~ X plana tion.
Su ppoHe t hat the cylinder occupies the space r < a wh i]e the crosssection of the bath of solution in which it is immersed is 11 (excluding t.ho space occupied by the cylinder). The concentration of solute in the solution is al "rays uniform and is initially Co. The cylinder is initially free from solute. The total amount of solute, J.lli, in the cylinder after t ime t is expressed as a fraction of the corresponding amount JISt:; after infinite time by the relation 1.6]
tr::!
J/i ~ 40:( 1 + n:) _ J)tJ~ t;a.1.
I.~ ) <l'2f .. ,
_;_l/~ ~ 4+4~+ ~"qn n.=l
( 5.33)
§ 5.3J
1)IFFl)"SlON IX ~\ CYLI~DElt
71
where the qn's are the positive, nonzero roots of aqn JO(qN) + 2J1 (qn) ~ 0,
(5.a4)
and ct == A/1Ta2, the ratio of the volumes of solution and cylinder. If there is a partition factor, K, bet ween so I tl te ill eq ui li hri 11 In in the cylinder and hI the solution, ~ == A/( 1Ta2 K)~ The parameter ex is expressed in terms of the final fractional uptake of solute by the cylinder by the
•
expression
.JI 'f..
1
'I'he roots of (5.3..J:) are given in Table 5.1 for several values of C( in order to assist. the evaluation of U5+33). The convergence of the series in (5.3~)) becomes inconveniently slow for numerical evaluation when Dtia» is small. An alterna tive solu tion suitable for small Dt,..'u'1. when rY is moderate is [7]
Mf_ = ~_±_~ [ 1  c4( 1 .} ~);l VI/( II ~ il ~ , o rf c { :! ( I + I,."\:) ( 1) t .a 2. ) ~ ,f ~ } J ~ ( i) . ~) H )
,II 1 IcY
.... ~ I 4
Carman and Hall I [H J ha ve derived an al terna t i ve eq nat.ion \\" hich is leR.~ easy to use but which is accurnte up to considerably higher values of .JJr,/.JI~. Their equation is
'''!~ = __ .Y3. _ .. t4ylDli((t'il'J erfcfYi!(D~)! l, f
ill JJ Y::J+Y4 ~ (t a.2 J
_) ... _~~_._ cty~ Dli(q2lJ.~' crfc{ _ ~y ~ (D:) ~.) ) ( ~. 37)
Y31Y4 {l'; a J
Y3 :.~: ~{( It a)! + ILl'" == Y3]. (5.3~)
For ,_"X very small it. is con venien t. to use the asy III ptot i c ex pansion for erfc in (5.~~ti) and hence to write
J/, _ I t':t f ct(Dt,/a2)1 ~ (\{I( Dt .... a'2) ~ 3,~,!)(J)t..:·a2)~ I
~)f;  I + 1 tl: t I  2171 (1 + !(~) I i t);&"( f=tlcr; )3  I ~ ~7Tf( I +1 Cl;)5 : ... •
(G.30)
If ct is very large, the following expression is more convenient,
_~l/r : (1 + (r)f !(Dt) ~ _ (_~ + ~)P} ._,'~ ~ 4 (1 ~ ~ ~ .~ __ )(l)~)~ + . ..1,
JI~ C'( t rrl a2. :! (i: n.2 31T~ S 0.: 0;2 a2 J
'5.40) which fur the special case of (l == C/), becomes
ill, == 2J 2_(1)t)~ ~~ _~ ~?! __ I_(_D~)J .J. (5.41)
JJI~ t 1T~ (12 2 a2 U7T~ a2 + j
rrhe deri vat.ion of these sol ut ions f01' small times is given by Crank [_7l
Fig. 5.7 shows curves of _1l~/_lll~ against (Dt/a'l)l for five values of the final fractional uptake. Berthier [9] gives a table of .ftt,/J/oo for values of I let between 0 and 1 at, intervals of 0·1. As in the plane case his values are not always reliable to three decimal places.
]·0 i i
tf'  _Io_
W U I+++i': ++++i! ++ ~ ~> ,....;1' r] i,.. ;..: __.j.rrJ 11" ..4 
, 1 f .... ~ ~ I......}"' ~ ....
.. I' DO ~ v p Ie f  ~ ~ ~n .V L  +++++1+1+1111
1++++_ .  r 9,2 '" ~ l,..r... J ,..r 1 ~r ~ ~ ~ r~ 4+++++41
O 9 Lr 1 V V V v _J.. i
. V 'I V l/ ~ _p.V 1 __ +++++tr'++++++++~+IIIII
1/ 1 I ~ V.   V t+   ,. rft.++;.......+~+++++++_+_+__+_I
11++++1,.'1 _ _ ~ i); V / " ~ i'I
0+ 8 ~..j.....! 1++i1.41/1 ~I+ l'h7 V,I+'+I'+' 4'[j4l/~:,.It.t_~~;'!~~~~~~~"tt.ti1.1~ ~~~;~~~~~ttt_tt~j_~101
I i r/. , .' I J r
!T i ~  .     5 ozttzrttztttttrrttrtttt: ~ ~  j.    r
If v . i1  V ~ j'
: / I j 3 ('II J f ~ V~ ~  ~
1 : 1/ I
O'7~~~~~~/~~/~/~/~ __ ~h~~~'~~~~r~!~r~r, ~~_+~~~+_ ++~~Hf~+++4~:~~~"~
~ [ ! _ 0/: I t !" ~    ~  r' 
""·:1 . 10 V i ++1+ J..:L.......f.I
__'___:'_/I+ ~  It       II  ff : ++:f_l___ll++++++++I+I++l++i...___......__,' ~
O~6 I; , J 1 1 1
~:/:; :1+: I V vll_r~1~ +~, I
1Ij.j.....~_llIf f r~;III  i' ., F'"   1  t i J r 11.  ; +t+l+I+++~+++++~.IoI
, 1  ~ II I I: I I ;
t+++++~7f___+_7++h++++++I__i_j___i_+_+++++~~l+_J____.I___+_i  I r +++IoloI.1
~ ~,  ~  II t J.: j II . ' ~ ~ F~ +10+4+1
O ~ [I i 1
... , _J V;_! 1I r '.
re [I 7 '_1. J ~  ~u •• ,~~~
  r+ :I I :!!   t, , ~ +.f+t+I ++++r+~~.__+__l+++!++I_I___I,;i+L++++I
1+i4+t+~I,'I___+_,I/I4++_+_++_+_I~I___l_++IIIl +14___"+ r    J  +++++T+oI+I+II ++~.loL~++++_+__+_+IIII___l_+___i_+_+__+_+++ t+  +  . .I++Iolot.lll.lII
7:J
DlltFITSION I~ A CYLINDER
[§ 5.3
As for the plane sheet, § 4.35, the a hove cq ua tions also describe the course of desorption into a wellstirred solution, initially free from solute, from a cylinder in which the concerrtra'tion is initially uniform and equal to Co' The only modifications are that. (5~35) is to be replaced bv
..
The concentration of solute C within the cylinder is given by
c === C~oo{ I ~ ~_(C(+ 1 )eDq!Ual ~(qn_~/'!)}.
+ L (4+4a+ll:2q;) Jo(qn)
n=l
.711;£
__ .... "'T"I
7Ta2(;o
1
1 I ~ 1 + ... Ci.
( 5.42)
(5.43)
§ 5.31
DIF~"USION I~ A CYLINl)ER
73
and Cl~ (Cl)~ replace C, c: in (5.42) where
(5.44)
5.34., Surfare euaporation.
If the cylinder is initially at a uniform coneen tration C2, an d there is a surface condition
D Be/or == a(C~Co)~
where Os is the actual concentration just within the cylinder and (:'0 is the concentration required to maintain equilibrium with the surrounding atmosphere, the required solution is
0C2 _ I _ ~ 2LJo(rf3n/a.) _ {J;'IJljal
C{lC2  ;S (f3!+L2)JoC8,,) e ·
The f3n ~H are the roots of
(5.4n)
and
(5.47) (5~48)
a dimensionless parameter. Roots of (5.4 7) are given in Ta ble 5.2 for several values of L4 The total amount of diffusing BU bstanoe, Jft~ entering or leaving the cylinder l depending on whether Co is greater or less than O2, is expressed as a fraction of jlloc, the corresponding quantity after infinite time, by
::, = 1  ~ ;t~!:~:; ·
The solutions suite ble for small values of time, provided r /a is not small ,
( 5.49)
are
_C_~z = 2h(Dta)I ierfe _~=:. _ + 4ha1!!!{ .~_ + ~ h} i2erfc a:r __ f .~.,
Co~Ca rl 2, .. /(Dt) r1_ 8r 8a 2.J(Dt)
(5.50)
v/here It. ~== 0; l D and
i· ,
.1lft == 2DtL ~_ SL2(Dt)i _ L2(Dt)2 (1 ~~ L)
M '2 312 2 2 ••••
a:: a .7r a a
(5.51)
 ~ Tabulated values of _l~/jrfx are given by Newman (10] from which the graphs of Fig. 548 are drawn. Newman [10] also gives values of a second function from which ,tlt,jJIOC) can easily be deduced for a parabolic initial di stri bution.
F
il
D IFF IT S I 0 ~ I N ~\ V Y L I ).l" I) L~~ It
I § l}.3
f I ~ ~~~ t1' • ~1··· ~J...~·l~ ~ .i. +
~ 11"" 1. i I _.
t. rAL :/ •• I..,.o!" . .
"" r~. .. ~ f . l/n:'" I f~  i  _ ,:~ll+ ~ t r J · ~  
t;f~H ~ 't . 10 j
. ~. Ie f ~2: .H _p;5i • ·t + ~ ~.  ~ I fI
~ '__ IT 1 if I ·v .~ .;. ~ l.K+ .... ,._  + J I ~ t
1.U I p.ll ~ Q:2_.: I + t. · .
11 1.  r. ~ .. ~r~ r : ..... , 1.
j : ~~. [; ~ _j +
H ~·l 't  ~~·~t t I·
rl rrn .~ '"t ~ ,. ·r·)!' r:r hi : ,
f ·t· ~ r
u . ....
.  ~ , H + ' ~?
r •• : i j :r l+  : ...
I ; i  j  f
h' .~ _[l I v: '" + i j .1 ~ + · T~
i .... , +
Ij : tIl ./ 'r~ Iot p ! .
. ~. . i .r I j I ~~. ~ r
+ _A. .~ ·t·t· 
y ~ .....
.. I+ 1[. f 1 r+ .~ rt" ,
I p ,. I . ! ...... ..... + 4
:1 .~ ... ·t· ~l .+
~i T ++;!. r r
l1·iJ . .r+ ~ r . ~
.. J .... I~""t . t· ._ .. 14 .
+ t II ~~ !iilr ~ , tJ "T""'ttr+: L ? ......... . ]
.I. 01. 1' .~ 1 . ~ . I ·f
.. P" ~ r+ 
~j  1' ~ + ..  • .~. 
+ ~J ~f 17~ r I
·~i . .;,. , :
 j ~ , ..... H
I. ,I ~. r ~ l+ 1 ~.
11, r  . !'t'   ...
11 .j ~....j._1"'" r .r • .: .~ ·t
 ' ..... r
.jloo  + r _ .. r
:J I l ,
II· ._( r  .. .. "[ H· l.  ;
10 ~ , .J. 
r1 ""t" r .p ·T
/.. , ... ' , ~ft:1 f ........ ., ,
[I ! f. ft r r :
:1 It ... t i : l t !
Ii J .;. ,_. . + +  +. ..
rJ· •• ~ ..j. "J I~ r Hj:'t , +.
. t"h
.~ 4 .;..
i ...  1+ i
v ~ u, ·Io·t& _, t . 1 L
~ ..i..... ~~ .. ,.
:J I i
II ,  L .; l ; H  r r i 
,
·1 L , ..... .....
J Y . In • ~ 1·  ...  [
..... f~ I' ~ . r r""t
~ ·r ~, ~ , I .j.. ,
... . ,. .......' .~. . 1  . ..
: 0·8
0·6
0·4
o
6 7
(Dr/a:) I
FIG. J+8. Snrpt iori and clrsorpt ion curves [01' l ho BUl'fa('p roudit.icn (5.~Hl}. N urn bers OIl curves are values of L ~  ({..:J.! D.
3
4
5
.,
...
1
5.35 .. Constant jll~X .F~ at the surface
If the cylinder is init.ially at. a uniform conccntrabiou G~, and there is a constant rate of transfer of diffusing substance }~ per unit area of the surface, i ,C4
D ~C ,'::1 L'
 (_.: _.: l c.r == _II o~
r == (1,
( 5.52)
where the n: n ' s are the posi ti vo roots of J1 ( ();) == o.
Roots of (.5~54) are tabulated in reference [~), p. 171 j, and the first fi ve roots are to be found ill 'Tahle 5.2 when L =:: O. ()bviolUdy the amount of diffusing substance lost by unit length of the cylinder in time t iH 27TaFO t, This is a problem which has been discussed in connexion wit.h the drying of clay by Macey [11], I aegor [12J, and others. A solution useful for s111aO values of t.imc is
(!~Co == _~(2(pat)~jerfC ~,~~+ I)t~a~:~~~}i2erfe_.a.,~~ ··f· ... }~
D t r 1.\" (Dt) 2a~r~ 2,;, (Dt) 
(5.55 )
Concentrat.iondistancc curves, plotted frorn (I3.5:_Q are shown in :Fjg~ 5.n.
DIFFUSIO~ I~ A CYLINDEH
l§ 5.3
76
5t3b. Lmpermeable surface
If the surface of the cylinder is impermeable and there is an initial concentration distri bu tion f( r) then
{fl. OC! a }
C  2 J r:l("') dr' + ~ eJ)Ct~t .fJo(r~~l I rj(r') 1. (rl:: rJ) dr'
 "2 L..t ~ J2() .. (J '. II., ~
a () a(Y.'l
o ~l ·1 0
(5+5fl)
where the cl:n ~8 are roots of
{5.57)
which arc tabulated in reference [3, p~ 171].
5.37. Composite cylinder
V arious problems of diffusion into a composite cylinder comprised of t wo coaxial cylinders having d i fferen t diffusion coefflcien ts have been solved. {See, for exam ple, Carslaw and ,T aeger [2, p. 2 S+ J.) The extra parameters involved make any attempt at general numerical evaluation too formidable to be attempted here. Some solutions relating to diffusion accompanied by nonlinear absorption are discussed in Chapter XII.
5.4. The hollow cylinder
Carslaw and Jaeger [2, p. 180J give the general solution t.o the problem of the hollow cylinder with the surface r ::...:..:: a maintained at a constant
....
concentration Cit and r == b at OZ! when the initial distribution is f(rL in
the region a ~ r ~ b. In the special case of a constant initial concentratinn, f(r) == Co, and when 01 = O2, the solution is
(fJ.58)
where
(5.59)
and the Ct:n ~s are the positive roots of
(5.60)
Roots of (5.60) are given in Table 5.3 for different, values of b]a. This table is reproduced from Carslaw and Jaeger's book [2~ P4 379]. In (5458) and (5.59) Jo and Yo are Bessel functions of the first and second kind respectively, of order zero. They are both tabulated in reference [3].
§ 5.4]
DIFFUSION IN A CYLINDER
17
• !++++I
0·1
0·3
0·7
0·4
Flo. 5+ 10. U pta.ke curves for hollow ey liu( lor,
The expression for the amoun t. of diffusing s 11 bst ance en tering 0 r lea vi ng the region a ~ r ~=.; b in ti me t is given by
(5Ji I)
In Fig4 5.10, curves of JI,/ .lllcr) are plotted against {Dt,l (b  a )2}1 for di fferen t valnes of bla
I •
78
DIF}'USION IN 1\ CYLINDER
I"§ 5.4
5.41 ~ Flow through cylinder uiall
If the surface r =::; a is maintained at C1) r == b at C2~ and the region a ~ r ~ b is initially at Co, the concentration approaches the steadystate distribution discussed in § 5 ~2 according to the expression
where the an's are roots of (5.60). The amount of diffusing substance entering or leaving the cylinder wall in time t is given by J.l/u where
Normally a quantity of greater practical interest is the amount, Q~, escaping from unit lengtll of the outer surface, r == b, This is readily deduced from (5.62) and is given by obtaining 21TD(r8c/ar)r~b and integrating with respect to timet t, In the most. commonly occurring case Co = C2 ~ 0, and we then find
~ _ 2(!?t~L) _ 4 ~ Jo(aLl:n}~(ban)eDctlt nC1  log(b/a) :S ~~{.J~{actn) ~J3(ban)} .
For a given btu the graph of Qt/(rrb2C1) against t approaches, at large t, a straight line which makes an intercept L on the t axis given by
( 5.64)
a2b2+(a2 + b2)log(bja)
•
4 Iog(b/a)
As Barrer [J 3, p. 37] suggests, this intercept provides a means of measuring the diffusion coefficient, D ~ for a material in the form of a hollow cylinder. Fig. 5.11 shows graphs of Qd(7rb2()]), the amount evaporated from unit area of the outer surface r = b, as a function of Dtj(ba)2 for different values of b!a.
!
T
...i....!..
T 100'
't' '7 ~ ~l)
1. lEo
• "J I. V'
7
I ......
Lf
, ~t,.  . 'J'V + .. ...,j r
 ·f
1,,(
~ 'i.o'"
. ~ H+~ rH+~+~~~~~4~H+~·~
~ !"' t' ...
++Io+H++I~~+:IJo9fh L.a.. ~ • ..~ + lo ; 1. . ~...; .... ; • I ~. . ~
:~:::, !±1~_..jj:~ T"'r+ e, · ,_++ __ ,~~~ =+~t:.. 1"'=: ~+i"++~ ++1j~ 1+. +f+Hi _H:1+r++_ .... rt ~~..:..4. ++l+~ fl~ _ ·Tt rt. ~ ~. f . :t +n t .~ .~. t ~
o ~r ...jC!I: ~ • ~ r j . t·  T 1= .1  ~.~. H H · t : l ~
0·1 0,2 0·3 04 0,5 0·6 0·7 0+8
Dt/(ba)l FIG. !).11. Approach to steadystate flow through wall of cylinder. Numbers on curves are values of bin.
o· I ~_+.~ _ , + _
7H
,
:."
1
i
......
.... ~  .t t
f 5t42. A general boundary condition
I aeger [14] has given the solution to the problem of di ffusion in to a hollow cylinder in which the concentration is initially zero and the boundary conditions on til e two su rfaees are
.....
BO B(}
k1" +k2" + k3 C ~ k4t
(.:1 or
k' Be k,8C J. r C k~
.  + "'2  + fl.') , == '4
1 r.lt e . .~ ,
(.. or
These conditions include as special cases:
(i) Constant con een tra tions, C1 on r = a and k} ::::::: lc  k'  1/  O· k 'k:  C k' l t: ~ C
2  1 ~ k 2  ) 4./ 3  1 ~ 41 h. 3  2 •
(ii) I~v aporation conditions on the surfaces
DcC/er == Yl(Cl~~~CL
D ''.("1:::1.  (C'" C1)
 (. / c r  Y2 !;!'
r ::::: (1,
r ;:;:; b.
(!J.(7)
C2 on r == b, when
r == u.,
(5.68) (!3.{)B)
r == b,
80
DIFF1}SION IN A CYLINDER
[§ 5.4
when kl == k~ ~ Ot k2 == k~ == D~ ka = Ylt k. = Yl G1! k~:::::::: 'Ylt k~ == yz Cz• This includes the obvious modification for an impermeable
surface.
(iii) Diffusion proceeding from a wellstirred solution occurring in the region 0 ~ r ~ a, the concentration at r = a being always the same as that throughout the sol ution; the surface r == b impermea ble, At. r ~ a the condition is
r:;;:::;;;:;:;:; a,
that is
and we have also
bGjor == 0, r ==: b,
(5.71)
so that kl == k; == I, k2 = 2DJa, k; == k; = k~ = ka == k. == O. Other cases, such as that of diffusion from a wellstirred solution in the region o ~ r ~ a with the surface r ==== b maintained at a constant eoneentration, or the surface r == a maintained at a constant concentration ","bile there is loss by evaporation from the surface r == b, and other combinations of these boundary conditions are all deducible from the general solution. The deri vation of the solution by the use of Laplace transforms is given by Jaeger [14]. The final result is
C = ak4J k;~bk; log(rjh rcbk~{k2ak3Iog(1:/~U_ aka k;  bk2 k;  abk'J k; log (a/b)
tX;I
'1T ~ eDCt~[ F(an)G'o{r; LYn){ k4[A~ Jo(bCJ.n)k; an J1(bCl.n)] ~ ?1::::r:l
 k~[ A 'I Jo(aa:n) k2 an J1 (arxn )]},
A'  k' tn: '2}
n~ 3 Ian
B' = k;+2Dkilb '
(5.72)
where
An = k3Dk1 a:!; B == k2+2Dk1/a;
Co(r; Cfn) = Jo(ra,.}[An Yo(aan)k2 ~n :t;(ao:n)]
Yo(TCtn)[An Jo(at1:·n)k2 0:11 J1(aan)],
F(et. ) == A~ Jo(b~n) k; Cln J1 (bCXnL~= __
" [A~ Jo(bO:n)  k; Cln J1 (blln)]2(A; +k2 Bet!) _ 1
 [An Jo(a~.n)k2 C(n J1 (aO:n)](A~+ k; B' a!)
and where the O!:.n '8 are the positive roots of
[(ka kl Dn:2)Jo(aa)k2 Ml(a.a)][{k; ~ k~ Da.2)Yo(bo)  k; O:Yl(ba.)][(k; k~ DLX2Jo(ba) k~ a.~ (ba}][(k3 kl DQ:2)Yo(aa) ~k2 aJ; (aa)] == O. (5.76)
(5.74)
(5.75)
f
. , . ..., ""r .
1.
!
J
I
I
:1
•
,.  •
.,
,.
II
I
~
IL
1:1
j II
i Ij
I ,
Lf
.1 I
, f
.j 4
rI .1
I~ r
LI
1
, ,
1".
II rI
17
J
II'"
 '( J
.J .I
J
J ,
v IJ
f7 J
) J J
I ....
IF
v ~
rr ~
~ l.oIII!
J I
J ,
v FI'
, ~
~ """"" ,
V ~
, "
~
I0I'l
~ ~ ~
~ ~ ... ....
i.' ,; .... "'"
L..o ""'" _,~
i.I' ~ L..,.'"
~ L.oo ioo'
l..o ~ wo ...
W" ~
 0 CO !IoO 'II!i' N
W~\J 0 0 0 0
1 ,
vG'" c
.
~
1....
•
oM
_t:

....:.
~
~
0 0
v'1 00
Qi
i
~
ffi
;,...
~
e
c:;.
.::
0
~
¢ ~
~ ~
Z
e
t\
~
=
0
......
~
~
A
..,;.
0 c:::
...;..' '4
w
Q)
~
~
~
~
ID
c:;.
=
~
+ ...
fI:1
I
=
0
'4
..;;0
~
..,;.;.
0 ==
N £I
Q
::;
8
•
eq

L"'='
+
Co
....
~ o

i§ 5.5
5.5. The region bounded internally by the cylinder r ~ a
(i) If the initial con cen tration throughout th e region r > a is G~ J and the surface T == a is maintained at C1, then
~
G"l~_0) = 1 + 2 f e_J)u1~JO{ur~}~(ua)Yo(ur)Jo('~) dJu.
t\ (,~"IT J 5(tta) + y ~(~t(l) U
4)
A solu tion useful for small times is
._~~(~!! == (a)i erfc L:": + (r:::_a)(Dt)I ierfc_:__a +
C\  (}o r ;L .. ,/(Dt) 4a.,i 2"l(Dt)
+ Dt(9a22ar7r2) +2 f. ra + (h.78)
,,  I er C 40.. 'OJ
32 a,ilr i 2, . .,/ (Dt)
(5.77)
Fig. 5.12 shows 110W concentration depends on radius at successive ti mes. The expression for the amount of diffusing su bstance, F, crossing unit. area of the surface, r = a, in unit time is
0:)
F ~ D(~) ~ 4(CtCo)D J l)u'J/ du
 or '~n ~ 172.a e u[Jg(a·u) + Y ~(au)J·
o
(5.79)
Numerical values of the integral in (5.79) have been tabulated by Jaeger and Clarke [15] and are shown graphically by Carslaw and Jaeger [2, p.283l For small times we have
and for large times
2D [I Y ]
F = ti(C1GO) log(4T)2y  {log(4T)2y}Z" ,
(5.81)
where T = Dtja2 and y == 0·07722 is Euler's constant,
(ii) If the region r > a is initially at a uniform concentration Co, and there is transfer of diffusing substance across r == a according to
r == a,
(5.R2)
then we have 0°1
0001
co
==. _ 2h J e_DUIIJO(ur}[ u¥;_ (1Ul}+hYo(ua)] Yo(ur)[ uJ1 ('00) +}tJo(ua)] ~u ,
'1T [ 1tJ1 ( ua) + hJo( ua )]2 + l uYi (ua ) + hYo( ua)]2 U
o
(5.83)
§ 5.G]
DIFFUSION II'; .A. C)_LINDElt
where h ~ ex/D. Graphs showing how (C1_·C1)/(COC1) at the surface r =:.:= a varies with time have been drawn by Carsla w and Jaeger [2, p~ 2R3], for several values of ah, Clearly because of (5J~~) these curves also show the rate of transfer across unit area of the surface r = a. Carslaw and Jaeger [2, p. 283] also give the solution of the corresponding problem when there is a constant rate of flow of diffusing substance across the surface r == a.
RE~"E R EN CE 8
I. Porte]', A. 'V., and :\Iart.iut T I R., Phil. 11tJ aq. 20 (1910) 5l1.
2. Carslaw, H. S~, and .Jaegor, J ~ C., Conduction of Heat in Solids (Oxford, 1947}.
3. British Association. Muthematical Tables, 6 (Cambridge, 1(37).
4, Carsten, It R. F., and JIcKerro\v, N. 'V., Phil. Jlag+ (7) 35 (1944) R12. 5. ,\rilliamsont E, D~, and Adams, L, H·~. Phys, Her, 14 (1919) 99.
6~ \Vil son t A. IT., Ph il. ill Gq. (7) 39 (1948) 48.
7. Crank. .r., ibid. 362,
8. Carman, P. C.J and Haul, l{. A. "r.t Proc. Roy. Soc. 222 A (1954} 109.
9. Berthicr, G., J. Chim, Phys. 49 (1952) 527.
10. NC\VJUan, A, E., Trans. AnI. Inst. ou: Enqrs. 27 (1931) 203. II. )1"acey, H. H., Proc. Phys. Soc. 52 (1940) 625; 54 (1 tl42) 128.
12. Jaeger, J .. C.~ ibid. 56 {1944) 197.
13. Barrer, R. l\r.~ Diffusion in and through Solids (Cambridge, 1941). 14. Jaeger, J. C" Proc, Roy. Soc. ~lIS,lV. 74 (1940) 342.
15. and Clarke, !'tl,.,11 Proe. Roy. Soc. Edinburgh, 61 ... 4. (1942) 229.
•
VI
DIFFUSION IN A SPHERE
6.1. Introduction
IF we restrict ourselves to cases in which the diffusion is radial, the diffusion equation for a constant diffusion coefficient takes the form
o.f! = D(020 + 2 ac). (6~1)
ot or2 r or
On putting (6.1) becomes
u == Or,
au (J2U
~ ..  = D.
at or2
(6.2) (6.3)
Since this is the equation for linear flow in one dimension, the solutions of many problems in radial ftow in a sphere can be deduced immediately from those of the corresponding linear problems.
6.2. Steady state
In this case the equation is
~(r2dC) = 0, dr dr
of which the general solution is
C = B+A/t,
(6.5)
where A and B are constants to be determined from the boundary conditions. If in the hollow sphere t a ==;; r ~ b) the surface r = a is kept at a constant concentration Cl~ and T == b at C2, then
o = aC1(br}+bC2(ra) . (646)
r(ba)
The quantity of diffusing substance, Q" which passes through the spherical wall in time t, is given by
ab
Qt =::; 47rDt ba (02~Cl)· (6.7)
If Qt is measured in a concentrationdependent system, the mean value of the diffusion coefficient obtained from (6.7) is
(f'DdO)/(C1C2)
('~
as for the plane sheet (see 4.11).
§ 6.2]
DIFFUSION IN A SPHERE
85
If the surface r = a is maintained at a concentration (:1 and at r = b there is evaporation according to the condition
dO
dr +h(O~02) == 0, r == b, (6.8)
we find
c =::;; aCt {hb2 + r( 1 hb)}+ hb'!.f)2( r  a) •
. r{hb2+a( 1 ~hb)}
(6.9)
The amount" Q~, passing through the spherical wall in time t is now given by
Q _ 47rDthab2(C1CZ) , hb2+a(1hb} +
(6. J 0)
If ak > 2, the rate of diffusion decreases steadily as b increases, but if an < 2 the rate first increases and later decreases, passing through a maximum when b = 2/h. As in the case of the cylinder, this maximum is due to the combination of a decreasing gradient and an increasing surface area as b is increased.
If the surface conditions are
oCIBr+h1(C1C) === 0, r = a;
8Cjer+h2(OC2) == O~ r === b, (6.11 )
the solutions are
C == Ct a2hl{b2h2r(bh21)}+C2b2h2{r(ahl + 1)a!h1}, (6.12)
r{b2h2(ah1 + 1) a2h1(bh21)}
Q _ 41fa2b2hl h2 JJt{ C1  02)
~~  b2h:li(ah1 + 1)~a2hl(bh21)·
6.3. Nonsteady state
and
(6. J 3)
6~31. Sur face concentration constant. I nitial distribution f (r)
If we make the substitution u == Or, suggested above, the equations for u are
(6414) (6415) (6.16) (6~ 17)
u = O~ r = 0, t > 0,.
u == aCo~ r = a, t > 0,
u == r Iv). t = 0, 0 < r < a·,
where C~ is the constant concentration at the surface of the sphere. These are the equations of diffusion in a plane sheet of thickness a, with its ends, r == 0 and r = a, kept at zero and aCo respectively, and with the initial distribution r j(r}4 ThifO; problem has been considered in § 4431, and the solution follows immediately by making the appropriate substitutions in equation (4.16)~ If the sphere is initially at a
J ) IFF L"": S 1 U N I ~ _A S P II ~~ R ~:
[§ 6.3
uni form concent.rat.ion, (/1' and the surface concentration is maintained constant. at. (/0' the solution becomes
0,6
04
(}(:1
___ .0"
r» 11
uOVl
(6.18)
1,0
1 " . ~ "'''':;1
u , "
_! .. _., rr ~r _~ .. ~. ~t:;;.~"'.,,, ~
~~ F~ ,.~ f "",:;:;;'" "~~.J /,""
_ J ._, ~  ,_ ~, r  ~.
... ~ /.1 tV'
J 01. ·1 ~~ ~., ·n ... 1 . ..oj "'"
... .... ~
I .1 .. r~ '," ~~ 1' L..ofo'" i"'" l..I 1/ JI,
, 1·_· r~ r ~~ .... ., LI f J , I,ll
f ~, I I' t' .... 11 l.,oo J f7 ,J II
F 1,00""'" "'" ..t I_) L "
~~.~ _, ~ i.o' IJ. .J I I I J I
L..<o""" ~ Lt 'f r1
U· 1 ....  tI" ~.~ t/:... ,·V ~ t
L"" J '1 I
,., ~ 1.1' r ~ II
~~ .... ~ ,j I " /, ~
If'o ~ ~" /: r J I
I~ _ ~i"'" I.' v J I I
:.. .... ~ v. ./ II
~ ./ II' J
n ~iI[ .... , ~ II L II
~ IIW _ ..... ,_ v J I IJ
'"" .J I ~
~ ".. I , II
17 ~ .J J 1
~ .. 'f 'I .. I
tl". 10 v , J 'I
~ ~f.' ". ~ f7 v
~ , I) 1 I J
tI", r: IJ , I
~ ". ,, II
~ Ii"" r ~ J I
I .... L..oo " v I
I\. ••• .... ... IJ , , II II 1
.... fiI'" , ~ J J r f
~ ~ ~ J J
~ t.I V I~ , II
., ~ , .I. J
~ 1,111 ,I L..P ., ~ r 1 1
~~ ~ ... Iii ... ~ I~ J ,
.... .... JI ~
II"'" L v I II
~ ~ l.J' ".
1 P S loP' ~ .J  ~ ~ _ ~'~ r
r "'! .... .... ~ ~ ~, ;..
~ , ..... ~ ~
.... ....  ~, I ' ·1
 ~ t.,.P ,I ~ f( ~ ~. 'I . I ~: ~ ( ff
• H """ ,.,~ l ~ ~~, i..o' '''~
U [l. •• '", • ~ ~r t~ ~ ..... 1!
~ "'" r· ::;~ 1 J r
.... o
1()
0,8
FIG~ 6. L Cnucontrat.ion dist.ribut.iuns a.t various t.imes in a sphere 'with initial eorieent.ration C1 and aurfuco cnnccntrut.ion Co~ Numbors on cur v es are values of Dtfa.2,
The concentration at the centre is given by the limit as r ~ ()~ that is by
~~1 = 1+2 :f (_I)neDn.;"Jtfa~. (6.19)
CO~Ll n ~ 1
The total amount of diffusing substance entering or leaving the sphere is given by
(6.20)
The corresponding solutions for small times are
a ~oo { f (2nt] )ar f. (2n+ I )a+,}
~ er c ~ I  er c , '
} 21O.,/(Dt) 2~1 (Dt)
nJO
(6.21 )
§ 6,31
~ ...
. j
~"f: (Di)! { _1 • ~ L1J"..I'. n.a} t»
 === ()  7T l r 2 ierfc : .. ~..~  3  .
.i.lfoo a2 n .~" I '\;i( Dt} a:!
These solutions can be writ.ten in torrns of the two dimensionless paramt.~tcrs Dl!a,2 and r]a, Curves showing (C(\)!(({,(/l) as a function of ria for different values of l)t/a2, dra.\lln by Carslaw and Jaeger [1, P4 20 I J~ are rcprod need in Fig. 6.1. 'I'ho cu rye for zero (factional ex ha ustion on Fig. H.4 shows _J~/JfCfj as a function of Duo",
and
(6.22)
6.32. Vari(Jblc surface concentration.
If the init ial coneen tration in the sphere i~ zero an d tha t at the surface is ¢( t) the sol ution is
,
C 21) L:~ J) 2 :!/::! • n1T1· J D 1: 'H' Z.rI... \ d \
.... ;;:.:; ~  (1 )nB  . u or iN nsr SIn _ .... _ e n 1r 1\,« ..,...(1\) Il.
ra a
n=l
o
(6.23)
(i) lVhcn
(6.24)
(6~23) becomes
_? == I ~ a e .. ~fit sin{({3a2/D)fr/a}_
(,~ r sin {fla2 j Dp
2f3a3 L~ n e 1Jn.I7r~tifl'l. ~ n7Tr
 (.l)  81n~
71Dr n(n21T2f3a2f D) a t
n=l
and the sorptiontime curve is given by
6Pa'l. 5~ e D" '''''/Ia'
+ 7T2D n2.(n21TzfJa2/ Dr (6.26)
n=l
Fig. 6.2 shows uptake curves for different values of the parameter pa2/D. (ii) If the surface concentration varies linearly with timet i.e,
(6.27)
_Ll~ ( a 2) 6a 2 Lm I ~ ~ II:
". == t  + _ eDn 1T ~ja
~1Ta3k 15D 7T4D . n" "
'h=1
~ D/(~1Ta5k} is plotted against ]Jt/a2 in Fig. 6.3.
and
(6.29)
88
3M, 4TTaiCf;I
0·8
5 . .
0·2
o
DI:FFUSION IN A SPHERB
l§ 6.3
1.0
0+.
o
}()
FIG. 6,2+ Calculated sorption curves for surface concentration given by ao( 1 e.fit).
Numbers on curves are values of patlD~
6.334 Diffusion from a wellstirred solution of limited volume
The problem and method of solution are very similar to those of the plane sheet and the results can be given without explanation. Suppose that the sphere occupies the space r < a, while the volume of the bath of solution (excluding the space occupied by the sphere) is V. The concentration of solute in the solution is always uniform and is initially G~. The sphere is initially free from solute. The total amount of solute, .Mi, in the sphere after time t is expressed as a fraction of the corre .. sponding quantity after infinite time by the relation
M, = I _ ~ 6a(a+I)eDq~lfa'. (6.30)
~ Lt 9+ 9~+q! tt!
n=l
where the qn's are the nonzero roots of
tan qn = 3 3qn'l I +cr.qn
(6.31 )
§ 6.3]
DIFFUSION IN A SPIIERE
8ll
and Ct = 3 V I( 41Taa)~ the ratio of the volumes of solution and sphere, or if there is a partition factor, K ~ between solute in equilibrium in the sphere and the solution, ex == 3V/{47Ta3K). The parameter a:: is expressed in terms of the final fractional uptake of solute by the sphere by the relation
1
(6.32)
0·5

~
..
 L"
fI
~
I~
I...J
LI'
1..
1.1
l.I
[I'
v
? 
.
1..1
;.I'
L,oo
!o':
l.."oo
1.0'"
l...o
~ 0·4
0·2
o
o· f
0·2
0·3
0·4
Dt/a2 FIG. 6.3. Sorption curve for sphere with surface concentrat.ion kt:
The roots of (6.31) are given in Table 6.1 for several values of ct. An alternative solution suitable for small times given by Carman and Haul [2] is
;1_ = (l+al[l_ Yl eerfc{3Yl(n;)1}~~.eerfC{ 3Y2(P;)i}]
co Yl+Y2 a a Yl+Y2 ~ a
+higher terms, (6.33)
'I
where in their notation
)/1 = ,{(I +~a)t + l}~ Y2 = Yl ~ l~ (6434)
and e erfo z ~ exp Z2 erfc z. (6.35)
In Fig. 6.4, M,/Mf.Q against (Dt/a2)t is plotted for five final fractional
Q
I§ 6.3
Of}
,
uptakes. Berthier [3] gives a table of .1~! .1.1100 for values of 1/ a between o and 1 at intervals of 0·1. His values are not alwa.ys reliable to three decimal places.
The concentration of solute C within the sphere is given by
O  Q {I 200 6(1 +LX)eDq!tIu;~ a sin(qn. rIa)}
~ ~ +  ~~ .
9+9a:+q!ct2 r sinqn n=l
(6.36)
The above equations also describe the course of desorption into a wellstirred solution, initia.lly free from solute, from a sphere in which
§ 6.3J
DIFF{T"S IOK IN A SPHJt~ltE
91
the concentration is uniform and equal to Co. The only modifications are that (6.32) is to be replaced by
3.2f.frc I
~ = ,
4rra3Co 1+1/0:
and 01' (C1)oo replace OJ c: in (6436)" where
01 === Co 0, (Ot)co == GoC(X).
(6.37)
(6.38)
6134. Su.rface evaporation
If the sphere is initially at a uniform concentration CH and there is a surface condition
where G~ is the actual concentration just within the sphere, and Go is the concentration required to maintain equilibrium with the surrounding atmosphere, tho required solution is
aco 2La L:oo eDf3!'ta~ sin {1nrfa
(6.40)
CICO  r n=l {,s! +L(L~ I}} sin ~n •
The fin's are the roots of
PncotPN,+L1 === 0,
and L :::::=: arx/ D4
(6.41)
(6.42)
Some roots of (6~41) are given in Table 6.2. The expression for the total amount of diffusing substance entering or leaving the sphere is
j~ ~  2:aj 6L2e~~~D'fa'J
M  I 2{,82 r
co fin n+L(L~ I)
n=l
(6.43)
Fig. 645 shows curves of ... l~/.V~ plotted as functions of (Dt/a2)f for several values of L for which N ewman [4] gives tabula ted solutions.
6t35 .. Constant jlux Fo at the surface
If the sphere is initially at a uniform concentration CO) and there is a const:ant rate of transfer l~ per unit area of surface, i.e,
~D BG/er == Fo, r == a, {6444}
then ,\ve have
Co a ;;::= Fo a {3])~ + 1 !:_~ __ ~~ __ 2 a ~ sin((}:" r) eD..;,t}. (6.45)
D a2 2 a2 10 r Lt a!a2sin(ct a) ,
n=l n
where the acc:n's are the positive roots of
aCln cot acrn === I. (6.46)
The amount of diffusing substance lost by the sphere in time t is 41Ta2 Fo t,
DIli'FUSION IN A SPHERE
[§ 6.3
92
Some roots of (6.46) are given in Table 6.2 when L = O. Fig, 6.6 shows curves of D(GoO)/(Foa) plotted against. ria for different values of Dtjaz. ]0
..
~ """ .".,.".
~ . ~
I.
~ 1
u S~ h'l2 11 ~
:~ 1,1 Ii':
0·5
II.
Ij_ II ·2
IJ
.Ill
L1 v
I) IJ
1 J
II
.l
~
J
.;_
"1
II
"1 i'.
J
tl ~ .i
r.
]
It
.. . 0.6
0·4
o
• 1
2
3
F1G4 6~5. Sorption or desorption curves for the surface condition (6.39)~ Numbers on curves 8.00 values of L = uajl),
6.36. I mpermeable surface
If the surface of the sphere is impermeable and there is an initial concentration /(r) then
a ro a
o = ~ f r2f(r} dr +! , e ~D~~I s!n Q;·n r J'" j{r')sm rx r' dr',
as ar ~ Bln2an a n
o ft 0 (6.47)
where the Cin'S are the positive roots of (6.46)4
6~37.. Composite sphere
Problems of diffusion into a composite spherecomprised of an inner core and an outer shell for which the diffusion coefficients are different ha' ... e been considered by Carslaw and Jaeger [I, p~ 288], Bromwich [5], Carslaw [6], Carslaw and Jaeger [7], Bell [8], and others. The extra parameters involved make any attempt at general numerical evaluation too formidable to be attempted here.
6.4. HoUow sphere
Carslaw and Jaeger [1 J p. 208] give the general solution to the problem of the hollow sphere with the surface r = a maintained at a constant
+
r+
I
•
 •
0
9
~~ . ~
I
..,... J
~ 0 ..
4
00
I
•
.p.
...,...t
•
0
'_
'~
~
 QQ l \0 V') V ~ M '1""'""1 0
+ . •
c O' 0 0 0 .0 0 0 0
I
~.
~
~~ o
· 
~
~ ....
...
C,.l
d
0
j
a
='
z
It;J
~
::0.
~
rn
oD
...e
.......
~
tI!
~
~
~
q:::
":'oN +
..I ~
....
ce
.,..;.. ......
I;QC!
=t.tt
o 0
c;,:.
~ .:tI
~
$~
2f ~
..8 ~
~IIo
ro =
d
_.
...
. 
rn
I.'::::
0
.
.....
]
.t
~
+l
rn
.
'"'t:I
d
0
~t
+l
:tI
:c.
~
:::
Q.')
~
....
....
0
C;
~
~
~
+
C
11
~ Qoo.
•
o
OQ
•
o
\0
+
o
M o
·
o
00
94
DIJrFT.:SION IN A SPHERB
coneen tration C1, and r = b at C2, when the initial distri bution is f( r) in the region a ~ r ~ b. Some special cases have been considered by
,~, .
ioo .+++++++1
....
.;..
j
...
07 0·8
{ Dt/(ba)l}l
FIG. 6+7. Uptake curves for hollow sphere.
Barrer [9] who also suggests a number of practical systems to which his solutions might be applied. In the special case of a constant initial concentration, f(r) =::;: Co, and when (}l == (}2~ the solution is
?SG = I + ~ ~ (b cos n7Ta) sin 1trr(ra) e1Jt,!!n!itl,:(lJll)!I. (6.48)
t.1Lo 1Tr L n ba
n=l
The total amount of diffusing substance entering or leaving the hollow
§ 6.41
DIFFUSION IN A SPJJ:ERE
95
sphere a ~ r c; b in t.ime t is given by
_~~~. == 1 ~~~_~6 ~. __ ~ (b ~osn.7Ta)2e_Dnfl":jlla}_(r}~. (fi.49)
_.iJ,f~ n2{a2+ab+b2) c: n
n=l
Gra.phs of .1lft/ j_1/x against {Df I (b' (t) 2 p. are shown in Fig. tl. 7 for diffefPn t values of b]a.
6~41. Fluw lhrouqh. sphrrical ioall
If the surface r == a is maintained at C1, and r = b at 02:' and the region a ~ r ~ b is initially at (}o, the concentration approaches the steadystate distribution discussed in § 6~2~ according to the expression
o == ~~:~+ (bC2aC!])(ra) +~ ~ b(C2G~)cosn7Ta(G)1G~) X
r r(ba) r7T L n
1i=1
x sin n7T(ra) eDn'rr1t/(ba):'. (6J>O) ba
The total amount which accumulates in the spherical wall in time t is ..t~~ where
M; = ~3~ (ba>[ (a2+ a;)C1 +(b2+ a:)C2 (a2+ab+b2)G'o
 :2 .i (a2Cl(a2+b2)Co+b2C22ab cosn'IT(q~02Co)} X
u=l
e Dnt:rr't,i(bU)'2]
X .
n2.
(6.51 )
The quantity which is usually of greater practical importance is Q~, the amount escaping say from the outer surface r == b. In the simplest case, when Co === C2 ~ 0, ,ve find
Qt == ]Jt ~~_! ~ (l)n e~Dn~1T~f(ba)'. (6~52)
47TlLb(b~a)(}1 (ba)2 6 1T2 Lt n2.
7i. ::1
As t + co, the graph of (6.52) when plotted against 1)t/(b~a)2 approaches a straight line which has an intercept on the time axis given by
Dt I (ba)2  64
(6.53)
As in the case of the plane sheet and the cylinder, this intercept can be u~d to determine the diffusion coefficient D experimentally [10, p. 31J. Flg~ 648 show~s the way in which the graph of (6)i2) approaches the straight line,
96
DIFFUSIO~ IN A SPHERE
[§ 6.4
0·5
Q, 4rran (b~ll') c
.
 .
} ~
, . ,~ ." 

j... .~.
~
,~ ~
.
r.~  c  
",II
L

'"
:.r
r.,.
 r 
;r
~. 
I! , 
~ .
._ ,~ ~, _I
.~ 0·4
0,3
0·2
0,1
0·5 00
Dt/(baj;
Ftc. 6.8. Approach to steadystate flow through the wall of a. sphere.
0·1
02
0,]
6.421 ISurface eoaporaiion.
Carsla w and Jaeger [I, p. 209 ] give a sol u tion to the pro blern in w hich evaporation into an atmosphere of zero vapour pressure occurs at r == a and r ::::::;; b according to the expressions
r =:: a,
(6.54)
If = b.
(6.55)
If the initial concentration distribution in the spherical wall is J(r), the solution is
.co b
C = ~ L: eDc<~IR.(r) J r'R,,(r')f(r') dr',
n=l
a
(6.56)
where
(6.57)
(H2 + b2k~ ~~)t { G sin( r a )O:n, +ak1 l"tn cos ('r ~ a )ll:'n}
{(b~aJ(a2k~ 0:; + G2)(b2k~ Cl:; + H2) + (Haklt Gbkz)( GH +abk1 k'}. rl:;)}" (6.58)
This action might not be possible to undo. Are you sure you want to continue?