Anda di halaman 1dari 21

Chapter 3:

Diffusion Equation and its Solutions for Binary Systems

(Ref: Shewmon for Ficks law and Glicksmann for Solutions)


3.1 Ficks Law:

It is interesting to see how concentration gradients give rise to diffusion flux of an element due to discrete atomic jumps. Consider a simple cubic material having concentration gradient along x-direction only. Assume that the concentration of B atoms vary in an A-rich binary solution as shown in Figure 1(a). Consider two adjacent atomic planes, plane-1 and plane -2, separated by a small distance in the material as shown in Figure 1(b).

Let n1 & n2 be the number of B-atoms per m2 on planes 1 & 2 respectively. Since, concentration is decreasing from 1 to 2, n1 > n2. Since, is the planar spacing, the small volume occupied by each element around the plane is .1 = m3. Thus, concentrations along planes 1 & 2 are given by,

C1 =

n1

&

C2 =

n2

Or, n1

= C1 & n2 = C2 (1)

Since, is very small, concentration gradient can be written as,

C C2 C1 = x
Thus,

C2 C1 =

C x

(2)

B be the jump frequency of B-atoms i.e., the number of jumps B atoms can make onto next lattice site per second. Thus, B is the number of vibrations per second that have energy greater than o equal to activation energy for migration of atom ( E Em ) . If we assume
Let, that the jumps are random, there is an equal probability of the atom jumping onto any nearest neighbor site. Thus, the probability of jump onto any neighboring site is 1 , where Z is the Z coordination number. Z for SC, BCC, and FCC are 6, 8, and 12, respectively. Thus, in case of simple cubic structure, the jump frequency onto any neighboring atomic site is

1 B. 6
Now, on Figure 1(b), the number of jumps per second of B atoms from plane-1 to plane-2

is

1 n1 B . 6

Therefore, the flux from plane-1 to plane-2 i.e. number of atoms jumping from plane-1 to plane2 per unit area per unit time is given by:

1 J12 = B n1 6
Similarly, the flux from plane-2 to plane-1 is:

(3)

1 J 21 = B n2 6
Since, n1 > n2, there will be a net flux from plane-1 to plane-2, which is given by,

(4)

1 J = J12 J 21 = B ( n1 n2 ) 6
From eqn. 1,

1 J B = B ( C1 C2 ) 6

(5)

Thus equation (5) indicates that the random atomic jumps will lead to a net diffusion flux down the concentration gradient. Substituting in the equation (5) from equation (2) we get,

C 1 J B = B 2 B ..................(6) 6 x
Or

J B ( x ) = DB

CB ..................(7) x

Equation (7) indicates that the diffusion flux at any position x is proportional to the concentration gradient at the position x & and proportionality constant, DB is called diffusion coefficient, which is given by,

1 DB = B 2 ..................(8) 6
Equation (7) was first proposed by Adolf Fick in 1855. Hence it is also called as Ficks law. It should be noted here that Fick proposed equation (7) based on his experimental observations of diffusion of salt in water. The theoretical proof based on atomic jump model as we used for deriving equation (7) came much later. While deriving equation (6) or (7) we have considered jump frequency (B) which consists of number of vibrations having EEm. However whether the jump will be successful or not will also depend upon the probability of the next lattice site being vacant. A more detailed derivation for DB for interstitial as well as substitutional atoms will be dealt with later. In equation (8) both B & vary with composition. Hence the diffusion coefficient DB is also a function of composition. Further, the diffusion coefficient does not depend upon concentration gradient, although this is not necessary, it has been experimentally proved in most of the systems whether liquid or solid. Thus Ficks law expression is similar to the expression of Ohms law in which the resistance is independent of potential gradient or to Fouriers law of heat transfer, in which thermal conductivity is independent of temperature gradient.

3.2 Binary Interdiffusion: As we have seen before, for the assumption of constant molar volume, we can write

= 0 , where J is the interdiffusion flux of i i

For binary interdiffusion

+J =0 J 1 2

Based on Ficks law,

= D C1 & J = D C2 J 1 1 2 2 x x D 1 C1 C2 = D 2 x x

However C1+C2== constant

C1 C = 2 x x

=D =D Thus, D 1 2
Thus, for binary interdiffusion, there exists only one interdiffusion coefficient. 3.3 Continuity Equation: Ficks law equation (7) has reduced the problem of diffusion from tracking individual atomic jumps to the one of solving partial differential equations to study variation in the concentrations as diffusion progresses. Equation (7), however, does not give dependence of concentration on time. The dependence of time can be incorporated based on continuity equation, which is derived below

Consider a binary alloy in which there exists a concentration gradient in x direction as shown in the figure 2 (a). The diffusion flux at any position will be obtained based on diffusion coefficient & concentration gradient at that position.

Figure 2: (a) schematic concentration gradient in a binary alloy (b) a slice of small thickness x in the binary material. Now consider a very thin slice of thickness x bound by two planes 1 & 2 as shown in the figure 2 (b). Let A be the cross sectional area of the slice. Thus the volume of the thin slice is A.x . Let J1 & J2 be the diffusion flux at plane 1 & plane 2 respectively. From the concentration profile shown in Figure 1 (a), it is clear that the flux J1 coming into the slice is greater than the flux J2 going out of the slice. If J1 & J2 are represented in no. of atoms per m2 per sec then the total amount of material accumulated in the thin slice in very small interval of time (t) is given by

n = ( J1 J 2 ) A t..................(9)

Change in concentration of the slice in time interval t is


C = n A x

C = ( J1 J 2 )

A t ..................(10) A x

Now, since x is very small, the flux at plane 2 is given by

J 2 = J1 +

J x x

J Or J 2 J1 = x x.................(11)
Substituting from (11) into (10) we get

C J = t x
t 0 ,

! C $ ! J $ # & = # & .................(12) " t %x " x %t

Equation (12) is called continuity equation. It should be noted that equation (12) is obtained based only one simple mass balance & we did not have to involve Ficks law to get the equation. Also the continuity equation is valid for an alloy containing any number of components. For an n components system, we can write

" C % " J % $ i ' = $ i ' # t &x # x &t


3.4 Ficks Second Law:

i = 1to n................(13)

If we substitute Ficks law equation, equation (7) into continuity equation, (12) we get
! C $ ! C $ # & = #D & ................(14) " t %x x " x %t

Equation (14) is called Ficks second law & it gives the variation of concentration as a function of distance (x) and time (t). It is a general diffusion equation which needs to be solved to obtain concentration profiles based on given boundary conditions. The diffusion coefficient D is usually a function of composition & hence in most of cases it is not possible to obtain a general analytical solution to equation (14) without the knowledge of variation of D with composition. The analytical solution of equation (14) becomes much easier if we assume that D is independent of composition, in that case equation (14) becomes:

! 2C $ ! C $ # & = D # 2 & ................(15) " t %x " x %t

Experimentally, it is possible to measure diffusion coefficients as a function of compositions. However, it is also important to understand the behavior of composition profiles with x and t. For this understanding, we will treat D as constant & and try to develop some analytical solutions for different cases. The purpose is to understand the general behavior of composition profiles in different cases. Moreover, the analytical solution developed with constant D can also be used to numerically treat the composition profiles for varying diffusion coefficients in different segments of the profiles. 3.5 Steady State Diffusion: The simplest case to deal with is steady state diffusion, in which mass flow does occur by diffusion but concentration at any point (x) does not vary with time i.e

C =0 t

Thus, Ficks law equation (15) for constant D, reduces for steady state to

d 2C = 0.................(16) dx 2
Note that due to independence of concentration on time, the partial derivatives w.r.t x is now replaced with ordinary derivatives w.r.t. x. Equation (16) is second order ordinary differential equation whose general solution is given by

C ( x ) = x + K ....................(17)
where K & are constants. We require two boundary conditions to get a particular solution i.e to obtain the values of K & . It can be seen that equation (17) is a straight line with constant concentration gradient given by . The actual value of gradient will depend upon the boundary conditions. As example of steady state diffusion, consider a steel plate of thickness w. Suppose that the steel plate is subjected to carburizing atmosphere on one surface & and decarburizing atmosphere on the other surface. Let C s be the carbon concentration on the carburizing side & C0 be the carbon concentration on the decarburizing side. Both Cs & C0 are maintained constant by continuously providing C on the carburizing side & continuously removing C on the other side. Obviously, the steady state will be developed soon & the concentration of carbon along the thickness direction in the steel plate is given by equation (17) Our boundary conditions are: C(0) = Cs & C( w) = C0
7

Substituting these boundary conditions in equation (17) we get,

K = Cs & =

C0 Cs w

Thus the carbon concentration at any point along thickness in the steel plate is given by

C=

C0 Cs x + Cs ..............(18) w

Figure 3: Concentration profile developed at steady state in the steel plate subjected to carburizing atmosphere at one surface and decarburizing on the opposite surface.

3.6

Non-steady State Diffusion:

In non-steady state diffusion, the concentration at a position (x) keeps changing with time i.e., Ci = Ci (x, t). In such a case, Ficks 2nd law equation i.e.

C 2C =D 2 t x

(19)

Has to be solved for x & t. Since equation (19) is a second order partial differential equation in x & t, we need two boundary conditions and one initial condition to solve it. 3.6.1 Laplace Transform:

In this chapter, we shall solve a few boundary value problems, which are of practical importance. We will make use of Laplace transform to solve these problems. Hence, it is important to refresh the concept of Laplace transform. If F(t) is an object function, the Laplace transform of F(t) is written as,

L { F (t )} = F ( p ) =

e
0

pt

F (t ) dt

(20)

Thus, Laplace transform of a time-dependent function converts that function into a timeindependent function. Here, F is called the Image function and e pt is called Laplace Kernel.
Inverting a transform converts an image function back to its object function. A few important object functions and their Laplace transforms i.e. image functions are given in Table-I.

Table I: Table of a Few Laplace Transforms


Sr. No. Object Function F(t) Image Function Condition L{F(t)}

1.

cos t

p p +2
2

2.

sin t

p +2
2

3.

exp(at )

1 pa 1 p
1

p>a

4. 5. 6.

1 d(t)

tn

n! p n+1 G (n + 1) ( p a )n+1
exp a p

n0

7.

exp(at ) t n
" a 2 % a exp $ ' 2 t 3/2 # 4t & " a 2 % 1 exp $ ' t # 4t &

p > a, n > 1

8.

a>0

9.

1 exp a p p

a0

10.

! a $ erfc # & "2 t %

1 exp a p p

)
)

a0

11.

( 2t )

3/2 n

! a $ erfc # & "2 t %

(1+n /2)

exp a p

a 0, n 1

10

3.6.2

Thin Film Solution/ Instantaneous Planar Source:

Figure 4: Instantaneous planar source (Thin Film Geometry) Consider an instantaneous planar source of an element i onto the surface of an alloy. The thickness of the deposited layer is infinitesimally small and the total amount deposited is M. Now, if the system is annealed at a constant temperature, the element i will diffuse into the material in a direction perpendicular to the thin layer. Initially, before diffusion, the concentration i in the material is zero. Also, as diffusion progresses, the concentration at a very long distance (x = +) is going to be zero. We assume constant diffusion coefficient (D). Thus, we have to solve the diffusion equation

C 2C =D 2 t x
with the initial condition, & boundary condition,

(21) (22) (23)

C ( x, 0) = 0

at x > 0

C (, t ) = 0 at t 0

Also, at any time, total amount of i in the material = M.

Therefore,

Cdx = M
0

(24)

Now, multiplying both sides of equation (21) by


pt

e pt and then, integrating for all times,

2 C pt C e . dt = D e . 2 dt t x 0 0

(25)

Lets first consider the integral on RHS of equation (25). Since,


e pt is not a function of x,
11

therefore,

e
0
0

pt

2C dt = 2 x

2 pt e C dt 2 x

Interchanging the order of integration and differentiation, we get

e pt

) 2C 2 & pt dt = e C dt ( + x 2 x 2 ' 0 *

However,

e
0

pt

C dt = L {C ( x, t )} = C ( x )
pt

(26)

Hence,

e
0

2C d 2C dt = 2 x 2 dx

(27)

{Ordinary differential used because C = C ( x) }.


Now, to solve integral on the left side of equation (25), we use the integration formula
b

udv = [uv] vdu


a a a

Substituting, Therefore,
pt

C u = e pt & dv = dt T
du = pe pt dt & v = C

Thus,

C e . dt = udv = [uv ]0 vdu t 0 0 0


pt pt % =# $e C &0 C pe dt 0

0 % + p C e pt dt =# e C e C ( t = ) ( t = 0) $ & 0

Using equation (26),


12

e
0

pt

C dt = p C C ( x,0) (28) t

Substituting equation (27) & (28) into (25) we get,

p d 2 C C ( x, 0) .C 2 = D dx D
Now, C ( x,0) = 0 , we can write (from equation (23))

(29)

d 2C p .C = 0 dx 2 D

(30)

Equation (30) is a second order ordinary differential equation, which can be solved easily. Thus, Laplace transform has helped us to convert the partial differential equation (21) in x & t to an ordinary differential equation (30) in only x. Let C = y Let

from Equation (30), y "


be the solution of equation (31). Thus,

p y = 0 (31) D

y = ex

p x .e = 0 D p or , 2 = D p = D

2 ex

Therefore, the general solution of equation (30) can be written as

p p C = A exp x + B exp + D D x
However, as we know, concentration field should behave or remain bounded at all x, including x = . This is possible only if B = 0. Hence,

p C = A exp D x

(32)

13

Now, we need to find the value of A. For this, we make use of equation (24). Taking Laplace transform over all times in equation (24),

$ ' + pt . pt & ) e C dx dt = e M dt 0 & ) , / % ( 0 0 0


(e
0 0

pt

M pt . M . Cdt dx = + e / = + ,0 1/ , 0 p p

Cdx =
0

M p

(33)

Now, from equation (32),

p M A exp x dx = D p 0

p M x = exp D p 0 M A [0 1] = pD D A. p or , A = M pD

Substituting for A into equation (32) we get,

C=
Now, from Table-I, we have,

M p exp D x pD

(34)

# ' * a 2 % 1 % 1 L $ exp a p ( = exp , / 4 t % % + . t & p )


1

for a > 0

Using a =

x in equation (34), we get D

14

{}

x 2 M 1 C = . exp D t 4 Dt

x 2 M exp (35) Or, C ( x, t ) = Dt 4 Dt


Equation (35) is very useful and has been extensively used in order to determine the diffusivity (D) for a dilute concentration of solute. To achieve this, a very thin layer of solute is deposited onto the surface of the matrix-alloy. The solute is then allowed to diffuse at a constant temperature; and after the annealing, concentration is measured as a function of depth (x). Since the original film is very thin and so the concentrations are dilute, the diffusivity is assumed constant and hence, the concentration profile follows equation (35). Since it is very difficult to measure very dilute concentration accurately, usually, a radioactive isotope of the solute is used for these experiments. The intensity of radio-active radiations is plotted against the depth and it is taken as a measure of concentration. Since an isotope is almost same as the solute atom, the diffusivity of the radioactive isotope (called tracer) is taken as impurity diffusion coefficient (D*). From equation (35), we can write

" M % " 1 % ' ln C ( x, t ) = $ * ' x 2 + ln $ $ * ' # 4D t & # D t &


Thus, ln C is plotted against x 2 and the slope of the plot gives the tracer diffusivity D*.

(36)

The thin film solution is also useful in semiconductor industry. For doping silicon with an n or p-type impurity, the dopant is first deposited onto Si as a thin layer. The Si with the deposition is then annealed at a higher temperature for sometime so that the dopant diffuses into Si. The depth of diffusion and control of concentration is very important for appropriate functioning of device. Equation (35) is very useful in determining the appropriate time of annealing and developing right concentration profile of the dopant.

3.6.3 Sandwich Structure:


Consider thin film geometry with an infinitesimally thin layer of solute deposited onto a solvent material. Now, place another block on top of the film such that the structure now forms a thin layer of solute sandwiched between two alloy blocks. This is illustrated in Figure 5(a).
15

Thus, the initial condition is

C ( x, 0) = 0
Boundary conditions are:

at any x {except at x = 0}

C (, t ) = C (+, t ) = 0

Figure 5: (a) Sandwiched thin layer geometry, and (b) Concentration profiles at various times in sandwiched thin layer geometry

With the same procedure as employed for thin film solution, we can show that the general solution for C in this case is,

p p C = A exp x + B exp + D D x

(37)

Since concentration field should be well behaved at x = + as well as x = , we can say that

16

A = 0 for x < 0 & B = 0 for x >0


Thus, the general solution for the thin-film sandwich structure is given by:

C = A exp = B exp +

p x D p x D

for 0 < x < +


(38)

for < x < 0

Further, since total amount (M) of the solute should be conserved,


+

Therefore,

C ( x, t )dx = M
+ 0

By symmetry (for same material on both sides of x = 0 ) we can write


0

C ( x, t ) dx = C ( x, t ) dx =

M 2

(39)

If we take Laplace transform on both sides of Equation (39), we get

$ ' + pt M . pt & ) e C dx dt = & ) - e 2 dt 0 ,0 / %0 ( 0


(e
0 0

pt

Cdt dx =

M + pt . M e / = + 0 1. / , 0 2p 2p,
M

Thus,

Cdx = 2 p
0

If we substitute from (38) for x = 0 to x = we get,

17

p M D p A exp x dx = = A . exp D D x 2 p p 0 0 A. D M . ( 0 1) = p 2p A= M 2 Dp

Similarly, we can show,

B=

M 2 Dp

C=

M exp 2 Dp M = exp + 2 Dp

p x D p x D

for 0 < x < + for < x < 0


By looking at Table-I we can write:

Substitute &

x D x a = D a =

for 0 < x < + for < x < 0


x 2 M 1 C= . exp 2 D t 4 Dt x 2 M 1 = . exp 2 D t 4 Dt

for 0 < x < + for < x < 0


Thus, the solution for thin film sandwich structure is given by

x 2 M C= exp 2 Dt 4 Dt

(40)

18

The concentration profiles at different times for a sandwich structure are presented in Figure 5(b). The concentration profile is a typical Gaussian-distribution profile. The effect of diffusion is to broaden the width of the peak and lessen its height.

At x = 0, C =

M = C ( 0, t ) 2 Dt
1 t
.

thus, the concentration at the plane at x = 0 decreases as

Also, if xe is the distance from x = 0, at which the concentration reduces to 1 times the e maximum concentration then, we can see

xe2 = 4 Dt i.e. xe = 2 Dt
Thus, the distance xe increases as decreases in height with time. To further understand how the concentration profiles would evolve, consider Figure 6 (a),

t , i.e., the concentration profile broadens in width &

C 2C & 2 respectively with x at a given time (t). (b) and (c) showing variation of C , x x

19

Figure 6: For a thin film sandwich geometry, (a) C vs. x, (b)

2C C vs. x & (c) vs. x x x 2

As we can see from Figure 6 (b), the concentration gradient is zero at x = 0 & is positive for x < 0 & negative for x > 0. Since diffusion flux will only be scaled by D, the flux profiles C will have same behaviour as that of except for the reversal of signs. Thus, the diffusion x fluxes will be negative for x < 0 & positive for x > 0 and zero at x = 0. This means that the solute will diffuse from right to left for x < 0 and from left to right at x > 0. At x = 0, the diffusion flux is zero. Next, inspection of Figure 2(c) shows that the curvature of the concentration profile

C is positive upto point b and beyond point c. Between b & c, the curvature is negative. x 2 This means that the concentration will decrease with time between b & c. However, concentrations at x < b & x > c will increase with time. This is because,
i.e.

(C t )

2 =D C

x 2

. 20

Another important point to be noted here is to understand the infinite boundary condition. In reality, a plate cannot have an infinite thickness. However, as long as the solute does not reach the extreme surface of the plate in x-direction, we can say that the boundary condition is infinite. Typically the boundary condition is treated as infinite as long as the thickness in xdirection is greater than 4.6 Dt .

Dt term is very important in diffusion. It is also called as characteristic diffusion depth as it has a dimension of length. Many a times, distances in diffusion are referred to in terms of Dt .
Thus whether a given boundary condition is infinite or not will depend not only upon diffusion coefficient but also on time. Typically, diffusion coefficients can be of the order of 1014 m2 s . Even if we consider diffusion time of 100 hrs, we have

Dt = 1014 x 3600 x 100 = 60 x 10 x 107 m = 60 x 106 m = 60 m 4.6 Dt = 276 m


Typically we will always have section thicknesses of the order of a few mm. Thus, infinite boundary conditions are typically observed.

21

Anda mungkin juga menyukai