Chemical Diffusion

Tapan G. Desai
Material Sciences Department Idaho National Laboratory
May 2007

Approach
Definitions of various diffusion coefficient terms Chemical diffusion - statistical mechanics Chemical diffusion - phenomenological model for ionic material Results on UO2 calculation of self and chemical diffusivities in melt and polycrystalline material hopping analysis radial distribution function comparison

Types of diffusion coefficients

Self diffusion: It describes the migration of the atoms of a pure element in absence of driving force (due to Brownian motion) Intrinsic diffusion: It describes the mobility of a species in a binary solid/liquid Chemical or Mutual or Ambipolar diffusion: It describes the collective property of the mixture

Chemical diffusion in ionic material

In ionic materials to avoid long range charge separation during mass transport the charged species are coupled through a common electric field The chemical diffusion coefficient is an apparent diffusion coefficient that characterizes the overall rate of transport Its value is bounded at high and low extremes by diffusivities of individual species involved

Green-Kubo relations
Self (Intrinsic) diffusion: calculated from velocity autocorrelation
1 Di = v i ( t ) v i (0) dt 30

Chemical diffusion: time correlation function of the relative velocity of

the center of mass of species 1 with respect to the center of mass of species 2 Q D12 = x1 x 2 N [v1 ( t ) - v 2 ( t )] [v1 (0) - v 2 (0)] dt 3 0
where

x1 (ionic fraction) = N1/N, Q - thermodynamic factor v1 (velocity of center of mass of particle of species 1)

= N1 1 v i ( t ) i=1

N1

Green-Kubo relations
Thermodynamic factor Q can be evaluated as follows
1 xx = 1+ 1 2 (GAA + GBB - 2GAB ) Q x1 + x 2

GAB =

[gAB (r) - 1] 4pr 2dr

r=0

gAB is the radial distribution function

The value of Q is unity for ideal solutions, but it is difficult to calculate from molecular dynamics The value of g(r) -1 fluctuates around zero at large r, but the fluctuation will be tremendously magnified when multiplied with r2 Hence most simulations calculate D12 /Q

Chemical diffusion
Q D12 = x1 x 2 N [v1 ( t ) - v 2 ( t )] [v1 (0) - v 2 (0)] dt 3 0

D12 can be resolved into contribution from intrinsic (self) and distinct diffusion coefficient as follows

D12 = x 2 D1 + x1D2 + x1 x 2 Dd
Intrinsic Distinct

Distinct diffusion coefficient is calculated from the velocity cross correlation function cross cross cross

Dd = D11

+ D22
N

- 2D12

N 1 1 cross D11 = v i ( t ) v j (0) dt 3 i=1 j i 0

Chemical diffusion
D12 = x 2 D1 + x1D2 + x1 x 2 Dd
Intrinsic Distinct
If there is no contribution from the cross correlation term, the above equation is reduced to the Darkens equation

D12 = x 2 D1 + x1D2
For equimolar mixtures, D12 is an average of D1 and D2 D12 value is close to the value of D with less content For identical species, Chemical = Self

Chemical diffusion
D12 can also be calculated from the collective MSD of the center of mass of species 1 or species 2

d N1 Q m1 rk ( t 0 ) D12 = c1 + c 2 6x1 x 2 N m2 dt k=1

2 rk (t 0 + t) k=1
N1

The chemical diffusion calculated are subjected to a greater statistical noise than self diffusivity In self diffusion each molecule contribute independent information to which the self diffusivity is averaged Chemical diffusion is a collective property, like center of mass which provides only one piece of information per time step Hence we need atleast N independent time origins over which the chemical diffusivity can be averaged

Phenomenological model
During diffusion, the preservation of electrical neutrality and stoichiometry must come from coupled diffusion of anions and cations (from same source to sink).

ZU JU = ZO JO

The total driving force is electrochemical potential, i = chemical potential + electrical potential

i = mi + Z i Ff
Ji = -

(F = Faraday constant)

c i Bi h i N a x m f m f ZU cU BU U + ZU F = ZO cO BO O + ZO F x x x x

( Z + Z )D D dc U O U O U JU = - Z D + Z D dx U O O U

Mutual diffusion
Dmutual ( Z + Z ) D D U O U O = Z D +Z D U O O U

The value of mutual diffusion coefficient is bounded by diffusivities of the two ions and the slower species is rate limiting - If DO >> DU, then Dmutual ~ DU -When the two ions are not so dissimilar in diffusivity, Dmutual is an intermediate value - When ions have more than one diffusion paths then
Dmutual =
l gb l gb ( ZU + ZO )[DU + pdDU /d ][ DO + pdDO /d ] l gb l gb ZO [ DU + pdDU /d ] + ZU [ DO + pdDO /d ]

Diffusion of the slowest species along the fastest path is rate limiting

Calculation of diffusivities in UO2 melt

1

0.8

0.6

Oxygen Uranium Mutual

2.5

T = 4500 K

Oxygen Uranium Mutual Ion model Darken

0.4

1.5

0.2

1
0

Diffusion coefficient (angs/ps

0.5

Normalized velocity autocorrelation function

-0.2 0 0.2 0.4 0.6 0.8 1

0 0 0.5

time (ps)

time (ps)

1.5

Self diffusion coefficient from velocity autocorrelation function matches with the Einsteins equation within 5% error Mutual diffusion coefficient agrees with Darkens equation closely for UO2 at T=4500 K. It is averaged over 20 different samples

Calculation of chemical diffusivity in polycrystalline UO2

4 107

mutual-pc
3.5 10
7

3 107 2.5 107

7

T=2700 T=2800 T=2900 T=3000

MSD

2 10

1.5 107 1 107 5 106

0 0 10 20 30 40 50 60

time (ps)

MSD of the center of mass of oxygen ions

Hopping analysis
How is the motion of the oxygen ions in the superionic state ? Is it liquid-like or it hops between vacancies and interstitials ? This can be done by examining the spherically averaged van Hove selfcorrelation function Gs(r,t), where the quantity 4 r2Gs(r,t) dr is the probability of finding a particle at distance between r and r+dr from its position a time t=0. The Gaussian approximation is given by
-3 2 2 - 3r 2 2 Gs ( r,t ) = p r( t ) exp 2 r t 2 3 ()

The validity of this equation can be tested by calculating the moments

r( t )
n

drr n Gs (r,t )
n ( t ) = r 2n by
n Cn r 2 - 1

Thus a departure of Gs(r,t) from a Gaussian form can be expressed in terms of (t), defined
Cn =1 3 5 7...(2n + 1) 3n

Hopping analysis
2 ( t ) = 3 r 4
2 5 r 2 - 1

For a non-Gaussian Gs(r,t) the 2 will not vanish, else 2=0

1

0.8

(=1800 ) (=3000 )

0.3

0.25

=4500

Uranium Oxygen

0.2
0.6 Uranium (T-3000) Oxygen (T-3000) Uranium (T-1800) Oxygen (T-1800)

() 2
0.4

() 2

0.15

0.1

0.2

0.05

0
0 0 5 10 15 20 25 30

()

10

15

20

25

30

()

Radial distribution function

3.5
4

20

T - 4500 K

T - 3000 K
3.5 U-U O-O U-O

T - 100 K
15 U-U O-O U-O

2.5 U-U O-O U-O

2.5

g(r)
1.5

g(r) 2
1.5

10

g(r)

0.5

0.5

0 0 2 4 6 8 10 12 14

0 0 2 4 6 8 10 12 14

0 0 2 4 6 8 10 12 14

r ()

r ()

r ()

Weakly correlated anion-anion interaction Is it true at temperatures just below the sub-lattice melting (T< 2000K)

Importance of cross correlation function

The cross correlation function is a measurement of displacement of particle i due to the displacement of particle j, where the particles are distinct (the species can be same of different) It is also a good indication of coupling between the velocities of 1-1, 2-2 and 1-2 species The main factors that influence are: Electrostatic interactions between cations and anions Mass and size differences

Definition
Einsteins Formula

: Jump distance : Jump Frequency

1 2 D= l G 6

e-E*/kT barrier

E* = A0 exp kT = fraction of vibrations which have sufficient energy to overcome the