Upload
phamthuy
View
218
Download
2
Embed Size (px)
Citation preview
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Mobility of atoms and diffusion. Einstein relation.
N
1i
2ii
2 ))0(r)t(r(N
1)t(rMSD
• This expression is called Einstein relation since it was first derived by AlbertEinstein in his Ph.D. thesis in 1905 (see note below)
• This expression relates macroscopic transport coefficient D with microscopicinformation on the mean square distance of molecular migration
• The 6 in this formula becomes 4 for a two-dimensional system and 2 for a one-dimensional system (see next page)
• This equation is suitable for calculation of D in MD simulation only forsufficiently high temperatures, when D > 1012 m2/s
• Time t cannot be too large for a finite system (D drops to 0 when MSDapproaches the size of the system)
• For periodic boundaries “true” atomic displacements should be used
• Derivation of this equation is given on pages 78-79 of the textbook by D.Frenkel and B. Smit
In MD simulation we can describe the mobility of atoms through themean square displacement that can be calculated as
The MSD contains information on the diffusion coefficient D,
nsfluctuatioDt6A)t(rMSD 2
Historic note: Before Albert Einstein turned his attention to fundamental questions of relative velocity and
acceleration (the Special and General Theories of Relativity), he published a series of papers on diffusion,
viscosity, and the photoelectric effect. His papers on diffusion came from his Ph. D. thesis. Einstein's
contributions were 1. to propose that Brownian motion of particles was basically the same process as
diffusion; 2. a formula for the average distance moved in a given time during Brownian motion; 3. a
formula for the diffusion coefficient in terms of the radius of the diffusing particles and other known
parameters. Thus, Einstein connected the macroscopic process of diffusion with the microscopic concept of
thermal motion of individual molecules. Not bad for a Ph. D. thesis!
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Derivation of Einstein relation for 1D case
Diffusion equation:
Let’s multiply the diffusion equation by x2 and integrate over space:
2
2
x
t,xCD
t
t,xC
dx
x
t,xCxDdxt,xCx
t 2
222
tx 2
dx
x
t,xCx
xDdx
x
t,xCxD
t
tx2
2
22
2
dxt,xCx
xD2dxt,xxC
xD20
dxx
t,xCxD2
x
t,xCDxdx
x
t,xC
x
xD 2
2
D2dxt,xCD2t,xDxC2
ADt2tx 2 for 1D diffusion
Let’s consider diffusion of particles that are initially concentrated at the
origin of our coordination frame, - Dirac delta function x0,xC
Solution:
4Dt
xexp
πDt2
1tx,C
2
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Derivation of Einstein relation
Diffusion equation:
Let’s multiply the diffusion equation by r2 and integrate over space:
t,rCDt
t,rC 2
rdt,rCrDrdt,rCrt
222
tr 2
rdt,rCrDrdt,rCrD
t
tr222
2
rdt,rCrD2rdt,rCrD20
rdt,rCrD2Sdt,rCrDrdt,rCrD 22
dD2rdt,rCdD2Sdt,rCrD2
AdDt2tr 2 for diffusion in d-dimensional space
= 1
Let’s consider diffusion of particles that are initially concentrated at the
origin of our coordination frame, - Dirac delta function r0,rC
Solution:
Dt2
rexp
Dt2
1t,rC
2
2/d
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Review of calculus used in the previous page
zz
yy
xx
Gradient:
Gradient of a scalar function: zz
yy
xx
)r(
Divergence of a vector function:z
F
y
F
x
F)r(F zyx
)r(F)r()r(F)r()r(F)r(
The divergence theorem: SV
Sd)r(Frd)r(F
2
2
2
2
2
22
zyx
Laplacian:
L’Hopital’s rule: if lim g(x)/f(x) result in the indeterminate form 0/0 orinf/inf, then
dxdf
dxdglim
)x(f
)x(glim
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Mobility of atoms and diffusion (I)
Atomic paths of two atoms in FCC lattice at temperature below the meltingtemperature. Figures are from MD simulations by E. H. Brandt, J. Phys:Condens. Matter 1, 10002-10014 (1989).
Can we use the Einstein relation to calculate the Diffusion coefficient from theseatomic trajectories?
d – equilibrium interatomic distance 50
d
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Atomic paths of two atoms in amorphous and liquid systems. Figures are frommolecular dynamics simulations by E. H. Brandt, J. Phys: Condens. Matter 1,10002-10014 (1989). Two longest atomic paths for each simulation are shown.
Can we use the Einstein relation to calculate the Diffusion coefficient from theseatomic trajectories?
d – equilibrium interatomic distance
Mobility of atoms and diffusion (II)
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Using Einstein relation. Example.
Changes in atomic mobility during solidification from the melt.
nsfluctuatioDtAtr 6)( 2
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Green-Kubo formula for diffusion coefficient
An alternative way to define D in MD simulation is through VelocityAutocorrelation Function (Green-Kubo expression):
• For reliable calculation of D trajectories should be computed for as long as thevelocities remain correlated
• Green-Kubo and Einstein expressions for D are equivalent
• We will discuss the meaning of the Velocity Autocorrelation Function later
• Derivation of this equation is given on page 80 of the textbook by D. Frenkeland B. Smit
• To get most from your MD trajectories you can use averaging over startingtimes.
dtvtvN
dtvtvDN
iii
0 10
))0()((3
1)0()(
3
1
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Temperature dependence of diffusion
Assuming Arrhenius behavior for the jump-frequency one can extract a vacancy
migration energy or an average activation energy for atomic migration in a
disordered system,
Tk
EDTDB
aexp)( 0
Note the deviation of the diffusivities from the Arrhenius behavior at hightemperatures, when kBT becomes comparable or larger than the activationenergies. This indicates a change in the mechanism of atomic mobility (changein Ea, collective motions..?).
Arrhenius plot for an amorphous alloy, by E. H. Brandt, J. Phys.:Condens. Matter 1, 10002-10014 (1989).
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Temperature dependence of diffusion
Diffusion of a cluster of 10 atoms on a substrate.
Shugaev et al., PRB 91, 235450, 2015].
Tk
EDTDB
aexp)( 0 Ea = 0.421ε
Lennard-Jones potential with parameters σ and used for the substrate A cutoff function defined in [Phys. Rev. A 8, 1504, 1973] is applied at 3σ
LJ parameters for atoms in the cluster are σcc = 0.60σ and cc = 3.72ε
LJ parameters for cluster-substrate interactions are σcs = σ and cs = 0.5ε
The mass of an atom in a cluster is mc = 1.74m, where m is the mass of an atom in the substrate.
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Temperature dependence of diffusion: Fast diffusion paths
Diffusion coefficient along a defect (e.g. grain boundary) can also be describedby an Arrhenius equation,
with the activation energy for grain boundary diffusion significantly lower thanthe one for the bulk. However, the effective cross-sectional area of theboundaries is only a small fraction of the total area of the bulk (an effectivethickness of a grain boundary is ~0.5 nm). The grain boundary diffusion is lesssensitive to the temperature change – becomes important at low T.
Tk
εDD
B
mG.B.G.B.G.B. exp0
Self-diffusion coefficients for Ag. Thediffusivity if greater through lessrestrictive structural regions – grainboundaries, dislocation cores, externalsurfaces.
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Partial diffusivities. Diffusion in heterogeneous systems.
Partial diffusivities in multicomponent systems can be calculated by averagingover atoms of one type only.
Heterogeneous diffusion. In calculation of the mean square displacement onecan average not over all the atoms, but over a certain subclass. For example, theplots below show the difference between atomic mobility in the bulk crystal andin the grain boundary region.
The plots are from the computer simulation by T. Kwok, P. S. Ho, and S. Yip. Initial atomicpositions are shown by the circles, trajectories of atoms are shown by lines. We can see thedifference between atomic mobility in the bulk crystal and in the grain boundary region.
Mean-square displacement of all atoms in the
system (B), atoms in the grain boundary
region (C), and bulk region of the system (A).
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Estimation of the diffusion coefficient from the jump-frequency information
If diffusion proceeds through vacancy migration, then one can obtain a diffusion
coefficient from the jump-frequency information. Assuming Arrhenius behavior
for the jump-frequency (R) one can extract a vacancy migration energy (Em).
T. Kwok, P. S. Ho, and S. Yip, MD studies of grain-
boundary diffusion, Phys. Rev. B 29, 5354 (1984).
Tk
ERRB
mv
v exp0
vvv RNaD 2~
Rv - the effective vacancy jump frequency
R0 - pre-exponential factor ~ effective coordination number × attempt frequency
Nv - equilibrium vacancy concentration that is ~ exp(-Evf/kT) where Ev
f is the effectivevacancy formation energy
a2 - effective squared jump distance
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Diffusion in nanocrystalline materials: Examples
Arrhenius plots for 59Fe diffusivities in nanocrystalline Fe and other alloys compared to the crystalline Fe (ferrite).[Wurschum et al. Adv. Eng. Mat. 5, 365, 2003]
image by Zhibin Lin et al.J. Phys. Chem. C 114, 5686, 2010
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
Spatially heterogeneous mobility of atoms. Example.
Atomic mobility is much more active at the front of crystallization. D is notreally a diffusion coefficient in statistical thermodynamics sense, but rather aquantity that reflects an average mobility in this material undergoing phasetransformation.
Changes in atomic mobility during crystallization of amorphous metal.