Diffusive Molecular Dynamics

22
Diffusive Molecular Dynamics Ju Li, William T. Cox, Thomas J. Lenosky, Ning Ma, Yunzhi Wang

Transcript of Diffusive Molecular Dynamics

Diffusive Molecular Dynamics

Ju Li, William T. Cox, Thomas J. Lenosky, Ning Ma, Yunzhi Wang

2

Traditional Molecular Dynamics

• Numerically integrate Newton’s equation of motion with 3N degrees of freedom, the atomic positions:

• Difficult to reach diffusive time scales due to timestep (~ ps/100) required to resolve atomic vibrations.

{ }, 1..i i N=x

3

Diffusive MD: Basic Idea

Ferris wheel seen with long camera exposure time

Variational Gaussian Method

Lesar, Najafabadi, Srolovitz, Phys. Rev. Lett. 63 (1989) 624.

{ }, , 1..i i i Nα =x

DMD

ci: occupation probability(vacancy, solutes)

Define µi for each atomic site,to drive diffusion

{ }, , , 1..i i i i Nα =x c

Phase-Field Crystal: Elder, Grant, et al.Phys. Rev. Lett. 88 (2002) 245701Phys. Rev. E 70 (2004) 051605 Phys. Rev. B 75 (2007) 064107

change of basis: planewave → Gaussian

4

( ) ( ) ( )

0 0 0

3 23 2 2 2

1

Gibbs-Bogoliubov Free Energy Bound:

1 exp exp | |2

(| |, , )

Nji

i i i j j j i j i ji i j

i j i j

F F U U

u d d

w

αα α απ π

α α

∞ ∞

−∞ −∞= ≠

≤ + −

′ ′ ′ ′ ′ ′= − − − − −

∑∑ ∫ ∫ x x x x x x x x

x x

2

1

3 2ln thermal wavelength 2

Ni T

B Ti B

k Te mk T

α ππ=

Λ+ Λ =

Variational Gaussian Method

{xi,αi}true free energy

VG free energy

5

Comparison with Exact Solution

Lesar, Najafabadi, Srolovitz, Phys. Rev. Lett. 63 (1989) 624.

6

7

DMD thermodynamics

( ) ( )2

1 1

1 3(| |, , ) ln ln 1 ln 12 2

N Ni

i j i j i j B i i i i ii i j i

F c c w k T c c c c ce

αα απ= ≠ =

Λ≤ − + + + − −

∑∑ ∑x x

{ }Add occupation order parameters to sites: , , , 1..i i i i Nα =x c

VG view DMD view

01

=

c

10

=

c

8

2

1 1

The chemical potential for each atomic site is easily derived:

1 3(| |, , ) ln ln2 2 1

N Ni i

i j i j i j Bi i j ii i

A cc w k Tc e c

αµ α απ= ≠ =

∂ Λ = = − + + ∂ − ∑∑ ∑x x

DMD kinetics

nearest-neighbor network

( )1

1 , if and are nearest neighbors2

0 otherwise

Ni

ij j ij

i j

ij

c kt

c ck i j

k

µ µ=

∂= −

+ − =

2B 0

calibrate against experimental diffusivity:Dk

k T a Z=

9

log(D)

Atomic Environment-Dependent Diffusivity

Atomic coordination

number

12(perfect crystal)

9(surface)

10,11(dislocation core)

experimental or first-principles

diffusivities

10

Particleon surface

(largeparticle)

11

Particleon surface

(smallparticle)

12

Sinteringby hot

isostaticpressing

(porosityreduction in nanoparticlessuperlattice)

13

Sinteringby Hot

IsostaticPressing

(randompowders)

14

Nanoindentation

(only atomswith coordination

number ≠ 12are shown)

15

Small Contact Radius, High Temperature

16

Indenter accommodation by purely diffusional creep

17

coordination number coloring, showing edge dislocation

Dislocation Climb

vacancy occupation > 0.1

18

• DMD is atomistic realization of regular solution model, with gradient thermo, long-range elastic interaction, and short-range coordination interactions all included.

• DMD kinetics is “solving Cahn-Hilliard equation on a moving atom grid”, with atomic spatial resolution, but at diffusive timescales.

• The “quasi-continuum” version of DMD can be coupled to well-established diffusion - microelasticity equation solvers such as finite element method.

• No need to pre-build event catalog. Could be competitive against kinetic Monte Carlo.

19

Quasicontinuum - DMD?

image taken from Knap and Ortiz, Phys. Rev. Lett. 90 (2003) 226102.

DMDregion?

continuum diffusion

equation solver region,

with adaptive meshing?

20

Stress-Induced Bain Transformation

FCC

BCC

21

22