Introduction to

Computational Materials
Science
Simulating plasticity at the
mesoscale
Richard LeSar
ICMR Summer School, UCSB
August 2013
Materials Science and Engineering

IOWA STATE UNIVERSITY

OF SCIENCE AND TECHNOLOGY
1

ICMR Summer School on Materials in 3 D: Modeling and Imaging at Multiple Length Scales
Location: ESB 1001

modeling talks at the Summer School

Date
Mon Aug 19th

9:00 am 10:30am
Marc De Graef

Tues Aug 20th

Wed Aug 21st
Thurs Aug 22nd
Fri Aug 23rd

Emmanuelle Marquis

Matt Miller

Stuart Wright

Mike Jackson

Mike Mills

Satoshi Hata

Richard LeSar

Samantha Daly

Dan Gianola

Yunzhi Wang

Marc DeGraef

9:00 am 10:30am

Tues Aug 27th

Simon Philpot

Thurs Aug 29th

Fri Aug 30th

2:00pm 3:30 pm

Peter Voorhees

Date
Mon Aug 26th

Wed Aug 28th

Week 1 Aug 19 23
11:00am 12:30pm

Simon Philpot

Week 2 Aug 26 30
11:00am 12:30pm

Dave Rowenhorst

3:30 pm 5:00pm
Poster Session
(1001 ESB Patio)
McLean Echlin
Poster Session
(1001 ESB Patio)
Joann Kuchera-Morin
Matt Wright

Open Demo Session

2:00pm 3:30 pm

Matt Begley

Michelle Johannes

Mike Uchic

Frederic Gibou

Michelle Johannes

Anton Van der Ven

Baron Peters

Jianwei (John) Miao

Jianwei (John) Miao

Anton Van der Ven

James Rondinelli

James Rondinelli

Free Time
3:30 pm 5:00pm
Poster Session
(1001 ESB Patio)
TBA
TBA
TBA

Barbecue at Goleta Beach

Notes: Group Photoshoot (ESB 2001 stairs) on Monday, August 19th.

Dinner at El Paseo Restaurant on Thursday, August 22nd.

DFT/atomistics

multiscale

mesoscale
2

Definitions: modeling and simulation

A model is an idealization of real behavior, i.e., an
approximate description based on empirical and/or
physical reasoning.
A simulation is a study of the dynamical response of a
modeled system found by subjecting models to inputs
and constraints that simulate real events.
A simulation does not mimic reality, rather it mimics a
model of reality.

modeling and simulation

The accuracy of a simulation depends on many factors,
some involving the simulation method itself (accuracy in
solving sets of equations, for example).
Often, however, the biggest errors in a simulation, as
least with respect to how well it describes a real system,
are the inadequacies of the models upon which the
simulation is based.
Thus, one cannot separate simulations from the
underlying models.

How do we create models?

Useful article by Mike Ashby
(Materials Science and
Technology 8, 102 (1992))
Discusses a systematic
procedure that one can follow to
produce models.
Many models would have been
improved if this process had
been followed.

Think before you compute!

5

abstract, mathematical representations of actual, real-life structures and processes. Constructing

a model starts with a choice of which phenomena should be included (and thus, by implication,

scales of deformation

Table 1 Length scales and timescales used to describe the mechanics of materials, as adapted from
Reference 9a
Unit

Length scale

Timescale

Mechanics

Complex structure

103 m

106 s

Structural mechanics

Simple structure

101 m

103 s

Fracture mechanics

Component

101 m

100 s

Continuum mechanics

Grain microstructure

103 m

103 s

Crystal plasticity

Dislocation microstructure

105 m

106 s

Micromechanics

Single dislocation

107 m

109 s

Dislocation dynamics

Atomic

109 m

1012 s

Molecular dynamics

Electron orbitals

1011 m

1015 s

Quantum mechanics

aIn

the first column, we indicate an important unit structure at each scale; in the second and third columns, the approximate
length scales and timescales; and in the fourth column, the approach used to understand and represent the materials
mechanical behavior at those scales.

we typically have methods and models for individual scales of

behavior - the coupling across scales is referred to as multiscale

Chernatynskiy

Phillpot

LeSar

based on Ashby, Physical modelling of materials problems. Mater. Sci. Tech. 8, 102111 (1992).
6

experiments

strain hardening in single fcc crystals

Mughrabi, Phil. Mag. 23, 869 (1971)

1 m
Szekely, Groma, Lendvai, Mat. Sci.


Engin. A 324, 179 (2002)

1 m

despite 80 years
of dislocations,
we have no good
theories for this
fundamental
structureproperty relation

(slope ~ )
1 m
single crystal under single slip
is the shear modulus

Taylor law:

1/2

\
Mughrabi, Phil. Mag. 23, 869 (1971)
8

strong size effects,

stochastic variation,
intermittent flow, stresses
sufficient to activate most
slip systems.
Dimiduk, Uchic, and coworkers (many papers,
including Science 2004).

Engineering stress at 1% strain

pure Ni at small scales

1000

100

KD 0.62
10
0.1

Bulk NI

10

100

Specimen Diameter (D) in microns

9

Dirac'on Contrast STEM Tilt Series

0.6 m Thick
Mo fiber

J. Kwon (OSU)
EFRC Center for Defect
Physics (DOE-BES)

courtesy of Mike Mills

10

Plasticity

11

edge dislocations
the Burgers vector b is a
measure of the displacement of
the lattice
distortion of lattice leads to
strain field and, thus, a stress

slip
plane

1934
(Taylor, Polanyi,
Orowan)

movement of an
edge dislocation
note the deformation that
arises from the movement
of the dislocation

annihilation
12

screw dislocations

b ||

1939
(Burgers)
b

the Burgers vector is constant for

a dislocation loop

mixed dislocations
E
F
E

F
B

C
D

F
E

E
F

b ||
13

stress
d

plastic strain

xi

stress

b
macroscopic displacement: D = x
i
d i=1
plastic strain:
N

D
b
N
p = =
xi = b
x

h dh i=1
dh

p = b x
= dislocation density = m/m3 = 1/m2

D
h

p =
D D
= tan
h h
1

Hull and Bacon, Introduction to Dislocations (2001)

14

plasticity

total strain: kl = +
e
kl

stress/strain:

p
kl

ij = c = cijkl kl
e
ijkl kl

app
p
kl

not to scale

for an applied stress of :

p e

elastic strain from: ij = c

e
ijkl kl

plastic strain from: = b

p

dh
i=1

goal of simulations:
calculate plastic strain
15

dislocation generation and motion

dislocations primarily move on slip (glide) planes:

bowing from pinned sites

dislocations grow

Frank-Read source
serves to generate
new dislocations

bowing around obstacles

basis for first dislocation
simulation by Foreman
and Makin (1967)
16

dislocation processes that move edges o

their slip plane
out of plane motion (activated processes):
[ 101]

(111) plane

(1 11) plane
b
S

cross slip:
screw dislocation can move off
slip plane
stress and temperature
activated

climb:
a diffusive process

17

Simulations

18

abstract, mathematical representations of actual, real-life structures and processes. Constructing

a model starts with a choice of which phenomena should be included (and thus, by implication,

scales of deformation

Table 1 Length scales and timescales used to describe the mechanics of materials, as adapted from
Reference 9a
Unit

Length scale

Timescale

Mechanics

Complex structure

103 m

106 s

Structural mechanics

Simple structure

101 m

103 s

Fracture mechanics

Component

101 m

100 s

Continuum mechanics

Grain microstructure

103 m

103 s

Crystal plasticity

Dislocation microstructure

105 m

106 s

Micromechanics

Single dislocation

107 m

109 s

Dislocation dynamics

Atomic

109 m

1012 s

Molecular dynamics

Electron orbitals

1011 m

1015 s

Quantum mechanics

the first column, we indicate an important unit structure at each scale; in the second and third columns, the approximate
What
computational methods we use depends on what our
length scales and timescales; and in the fourth column, the approach used to understand and represent the materials
mechanical
behavior are
at thoseand
scales. the limitations of the methods.
questions

aIn

We start
by identifying
the entities in the model.
Phillpot
LeSar

Chernatynskiy

based on Ashby, Physical modelling of materials problems. Mater. Sci. Tech. 8, 102111 (1992).

19

density functional theory (DFT)

entities
electrons

solve Schrdingers equation: H ( r1 , r2 ,, rN ) = E ( r1 , r2 ,, rN )

too hard to solve directly, so make numerous approximations

approximations and limitations

convert N-electron problem to N 1-electron problems and solve

those for and find the electron density, (the Kohn-Sham method)

solve equations self-consistently in potential field arising from

use approximate functionals (Exc) for potential field (LDA/GGA/...)

limited (generally) to 1000s of atoms

good for dislocation core structures and small numbers of

dislocations

20

molecular dynamics
entities
atoms

force on an atom is: Fi = iU

d ri

solve Newtons equations: Fi = mi ai = mi 2

dt
N 1 N
need description of U
U = ij rij
i=i j=i+1

( )

approximations and limitations

potentials are analytic expressions with parameters fit to experiment

and/or DFT (e.g., LJ, EAM)

reasonably good potentials are available for many systems, but great
potentials are not available for almost anything

limited (generally) to 100s of millions of atoms

time scales: typically nanoseconds

21

limitations of atomistics
Modeling deformation on the scale of dislocation microstructures
cannot be done at an atomistic scale:
1 m3 of copper includes approximately 1011 atoms
time steps in MD: ~10-15 sec
- MD limited to a few hundred million atoms for a nanoseconds
atomistic simulations can describe processes that include only
small numbers of dislocations at fast rates

atomistic simulation of dislocations

showing stacking fault planes between
partials
courtesy of T. German, Los Alamos
22

modeling at the mesoscale

There are numerous methods used at the mesoscale,
some force based and some energy based.
For force-based methods, one must
define the entities
determine the forces
define the dynamics
solve the equations of motion
These methods have similarities to molecular dynamics,
but the entities are collective variables not atoms or
molecules

23

entities
In DFT and MD, the entities were clear: electrons
and atoms.
At the mesoscale, entities could be defects, such as
dislocations or grain boundaries, or some other
variables that define the physics of interest.
These entities are collective variables, in which the
actions of many smaller-scale entities are treated as
one.
We will often have flexibility in the choice of the entity,
e.g., the many ways to model grain boundaries
Most successful modeling is for cases in which there
is a clear separation into collective variables

24

damped dynamics
Most applications of dynamical simulations at the
mesoscale involve systems with damping, i.e., there
are forces that dissipate the energy.
2
d ri
Standard equation of motion: mi 2 = Fi
dt
diss

Force due to friction is usually velocity dependent: Fi = vi

is the damping coefficient

d ri

Net equation of motion is: mi 2 = Fi vi

dt
2

25

solution in 1D
2

d x
m 2 = F v
dt
For constant F (i.e., no variation with x), the solution is:

F
t /m
v ( t ) = 1 e

tterm 3m /

v* = v / F
t* = t / m

For large damping

(large ), we often
ignore the inertial
effects and assume:
F
v = = MF

this is called the over

damped limit
26

mesoscale simulations of dislocations

model the behavior of only the dislocations by treating

them as the entities tracked in the simulation
three main approaches
- force based (discrete dislocation dynamics)
- energy based (phase field)
- continuum methods (a density-functional theory - see
recent work by Sanfeld, Hochrainer, Gumbsch, Zaiser, ...)

today we discuss discrete dislocation dynamics

on Friday, Professor Yunzhi Wang will discuss a phase-field

approach

27

phase-field and discrete dislocation dynamics

phase-field dislocations
- free-energy-based
- Ginzburg-Landau dynamics
- advantages: links naturally to other phase-field methods,
easy to include energy-based phenomena (e.g.,
partials)

dislocation dynamics
- force-based
- dynamics from equations of motion
- advantages: accurate dynamics (inertial effects), stressdriven processes (cross slip)

the output of both methods are similar

- dislocation substructure evolution in response to a load
28

outline

simple 2D model

basics of 3D simulations

examples:
1. small scale plasticity
2. bulk plasticity
3. strain hardening

what is wrong with the simulations?

connection to experiments

All simulations discussed today are based on isotropic elasticity including anisotropy is not difficult, just very time consuming
29

discrete dislocation simulations in 2D

we start with a simple 2D model that consists of parallel

edge dislocations

first modern dislocation dynamics simulations were

based on this model (Lepinoux and Kubin, 1987; Amodeo
and Ghoniem, 1988; Gulluoglu et al, 1989)

such simulations require (a partial list):

1. representation of dislocations in space
2. description of interactions (forces)
3. boundary conditions
4. description of dynamics

30

Step 1: Simulation of system of straight edge

dislocations: represent as points in 2D

Assume all dislocations have: b = bx and = z

on any xy plane:
slip
planes

at low T, no climb and dislocations can only move on their

slip planes
This model is sometimes referred to as 2.5 D.
31

Step 2a: Interactions between dislocations

Assume: b = bx and = z

Stress from dislocation:

2
2
x
x

y
b j
12 ( j ) =
2
2
2
2 (1 ) x + y

-1

Force from this dislocation on

another dislocation: Peach-Koehler force
0
1

s11 s12 s13

F (i )

s = @s21 s22 s23A

= bi ( j ) i
L
s31 s32 s33
-2

-2

bi b j xij x y
Fx ( i )
= bi 12 ( j ) =
L
2 (1 ) xij2 + y

xij = x j xi

-1

2
ij

2
ij
2 2
ij

NOTE: long ranged

(~1/r)
32

Step 2b. External stress

Let be the applied shear stress: = 12ext
The force from external stress on dislocation i is:

ext
x

(i ) = b
i

Note: Burgers vector has sign:

either +b or -b
stress drives +/- dislocations in opposite directions
Net force (assuming N other dislocations)

bi b j xij x y
Fx ( i )
= bi +
2
L
ji=1 2 (1 )
xij + y
N

2
ij

2
ij
2 2
ij

33

Step 3. Boundary conditions

Put N dislocations at random positions in a 2D periodic
square grid with size D with equal numbers of +b and -b
dislocations
Dislocation density is = N / D
1/2
(all important distances in this system will scale as 1 / )
2

We will look at truncations of dislocation interactions:

1. short-range cutoff
2. no cutoff

34

Step 4. Dynamics
Fx ( t )
Assume overdamped dynamics: v ( t ) = M
L
Assume a simple Euler equation solution:

xi ( t + t ) = xi ( t ) + vi ( t ) t
For no external stress, run the system until converged.
Then apply an external stress, calculate change in
dislocation position, and calculate
N

b
N
p = 2 x i = b 2 x = b x
D i=1
D

35

Results: truncated potential

Not quite converged (=0)

T T
/

T
/..+ .L/
j.j.J.

IT

ITT

At converged solution:
straight lines alternating
+ and - dislocations with
spacing 1/2 the cutoff
distance

J"

I~-'~.

I~

~1-

T
~

-"

T T l~ ~
T

m'

T..,~-..LL

,~.

l~-

~"

T~.
tl

(b)
h- 3T"

J-,J'~b.,
T

'~L

. L ~ ".:t..8
"rz -L'~I ~:.
T

.1,
. , .~ '.~
T T J ~ ~" I

~
T

iL -~

~'~@T

.i ,,

: r~

ITII
z,.~ht-

J-

"

'r

.~.

.L

h,~
a::~-"-'-

J"

FIG. 1 Simulated dislocation microstructurcs with a dislocation density of 1015 m -2 corresponding to 10 3 edge
dislocations in a 1 gm x 1 grn simulation cell using (a) the infinitely repeated simulation cell (i.e. no cut-off) and
(b) the truncated interaction distance (Pc = 0.5 cell size) methods.

Dislocation distributions in two dimensions, A. N. Gulluoglu, D. J. Srolovitz,

R. LeSar, P. S. Lomdahl, Scripta Metallurgica 23, 1347-1352 (1989).

36

Results: all interactions to infinity

DISLOCATION

DISTRIBUTIONS

Vol.

23, No.

sign-weighted pair correlation function

(a)

shows importance of knowing how to carry out the simulation!

T T
/

IT

/..+ .L/
j.j.J.

~"

Dislocation
distributions in two dimensions, A. N. Gulluoglu, D. J. Srolovitz,
ITT
I~-'~.
I~
~1T T l~ ~
,~.

J"
T
l
~
T
~.
R. LeSar, P. S. Lomdahl, Scripta
Metallurgica 23, 1347-1352 (1989).
Z

m'

-"

tl

T..,~-..LL

37

Your assignment ...

We will email you a zipped folder called LeSarDD.zip.
In this folder is a file (DislocationDynamics.pdf) that
contains a description of how to implement (in
MATLAB) the 2D modeling I just described.
DislocationDynamics.pdf also contains some exercises
to consider.
There is also a folder called code that contains 2 other
folders. The one called DD2D contains MATLAB code
for this problem.

38

2D models show interesting behavior, but ...

many interesting simulations have been done with these
2D (2.5 D) models
Alan Needleman and Erik van der Giessen in particular
have done very nice work, including coupling to
continuum models
However, dislocation plasticity is 3D
Michael Zaiser (Erlangen) once
describe these 2D simulations as
cubist representations of plasticity
La Femme Au Miroir
Fernand Leger, 1920
39

discrete dislocation simulations in 3D

include dislocations moving along all active slip planes to

examine evolution and response

pioneering simulations by Kubin (early 90s). Many groups

doing lovely work: Kubin and offspring (Devincre,
Madec, Fivel, ...), Bulatov, Cai, Weygand, Gumbsch,
Schwarz, Ghoniem, Wang, El-Awady, Zhou, ...

such simulations require (a partial list):

1. representation of dislocations in space
2. description of interactions (forces)
3. boundary conditions
4. description of dynamics
5. approximations, models, etc.
40

fundamental properties of the atoms.

Step 1: approaches to representing dislocations

2.2. Principle of Discrete Dislocation Dynamics

The concept of 3D discrete dislocation simulations was imagined by L. Kubin,

methodologies
into two
Y. Brchet andfall
G. Canova
in theclasses
early 1990s [KUB 92], [DEV 92]. The first code
Micromgas was a simple model for which dislocation lines of a f.c.c. single
crystal are
sub-divided
into sets of edge and screw dislocation segments embedded
discrete
linear
segments
in a continuum medium as pictured in Figure 2.2a.

- pure edge/pure screw (Kubin and company)

- mixed edge and screw (ParaDis, PARANOID)

curvilinear dislocations
2.2. (a) Edge-screw
discretization of dislocation lines;
- parametricFigure
dislocations
of Ghoniem
(b) internal stresses induced by edge and screw segments
- nodal points plus interpolation
- numerical integration along curves

Each dislocation segment generates a long range elastic stress field within the
entire simulated sample. In the case of isotropic elasticity, analytical expressions for
Ghoniem, Tong, and Sun, Phys. Rev. B 61, 913 (2000);
the internal stress generated by a finite segment have been established by J.C.M. Li
Wang, Ghoniem, Swaminarayan, and LeSar, J. Comp. Phys. 219, 608 (2006)
[LI 64] and R. DeWit [DeWIT 67]. Taking into account the anisotropy of the elastic
41

Step 1: parametric dislocations

2
3
2
3
ri = 1 3u + 2u Pi + 3u 2u Pi+1

2
3
2
3
+ u 2u + u Ti + u + u Ti+1

) (
) (

track the motion of

the nodes
remesh as needed

Pi

Pi+1

Pi+2

Ghoniem, Tong, and Sun, Phys. Rev. B 61, 913 (2000);

Wang, Ghoniem, Swaminarayan, and LeSar, J. Comp. Phys. 219, 608 (2006)
42

Step 2: stresses and forces

F = b

force on dislocation
calculated from stress:
Peach-Koehler Force

bA
rA
A

stress comes from many sources

( a)

= ext + defect + self +

( ab)

bB

dlB rB

ba

stress from an individual dislocation

bn

2
ij =
kmn R,ijm ij R,ppm d k
R,mpp jmnd i + imnd j +

8
1

3
R
1
R,ijk =
2
xi x j xk R

evaluated numerically:
f (r ) d

Nsegment N
int

w f (r
=1

q=1

basic method: Wang, Ghoniem, Swaminarayan, and LeSar, J. Comp. Phys. 219, 608 (2006)
43

Step 2: stresses and forces

[1]

[3]

[2]
2

[4]

[5]

B
i
A

see Hirth and Lothe, Theory of Dislocations

Devincre, Solid State Comm 93, 875 (1995)

[6]

if use discrete linear

segments

stress at point i from a segment

from A to B is an analytic
expression made up of more
tensors
from stress can find the force.
below we will discuss how to
calculate stress on nodes from
stress on segments
44

Step 2: self stresses

Two terms:

core energy increases with length of dislocation so it

opposes dislocation growth:
E core
L

= T = b2

interaction of one part of a dislocation with itself

45

Step 4: equations of motion and dynamics

full equation of motion (required at least at large ):

ma = F v

over-damped limit (ignore inertial effects): v = F /
for force velocity: determine forces on the nodes and
solve the equations of motion

r (t + t) = r (t) + v(t) t

time step limited by largest force

Note: as we shall discuss below, the forces and velocities
of the nodes requires some attention.
Dislocation motion in high-strain-rate deformation, Wang, Beyerlein, and LeSar,
Phil. Mag. 87, 2263 (2007).
46

Step 5: some approximations

dislocation reactions
treated with models
ignore partials (more on this below)
coupling of dislocation motion to geometry
we ignore the effects of lattice rotations caused by the
dislocations
many issues with boundary conditions, with the
modeling of bulk plasticity being much more
challenging than modeling of small-scale plasticity

47

Step 5: models
A Monte Carlo method is used to
determine whether cross-slip is activated.

cross slip:
[ 101]

(111) plane

(1 11) plane
b
S

L t
P=
e
Lo to
P =1

V
o CS

k BT

for CS < o
for CS > o

Kubin LP et al., Scripta Mater. (1992)

Frank-Read sources, annihilation, and junction forming

what else should we include:
e.g., climb, ...?
annihilation

junction

48
48

Summary of the approximations

all atomistic-based processes described with models
most codes ignore partial dislocations
most materials are anisotropic not isotropic
we ignore the effects of lattice rotations caused by the

dislocations (except as a post-process)

boundary conditions are a challenge, with the modeling
of bulk plasticity being more challenging than modeling
of small-scale plasticity
we typically approximate long-range interactions
...

49

steps in a simulation
choose initial conditions and stress
- place dislocations randomly on possible slip planes
calculate total stresses on each node by integrating

over all dislocations and find forces

calculate if cross slip occurs
solve equations of motion (nodes move)
check for junctions, annihilations, etc.
repeat

from movement of dislocations, calculate plastic strain

analyze dislocation structures, densities, etc.
50

An example:
a simple Frank-Read Source

51

example: a line-tension based Frank-Read

source
DislocationDynamics.pdf also contains an explanation of
this model. A folder called DDFR in the folder code
contains MATLAB code this model.
The goal is to simulate:
Nota bene: for a more complete discussion,
see
Computer Simulations of Dislocations, V. V.
Bulatov and W. Cai, (Oxford University Press,
New York, 2006
and for more complete MATLAB codes, see
http://micro.stanford.edu/~caiwei/Forum/
2005-12-05-DDLab/

52

the model

We start with a dislocation

pinned at both ends.

y
x

This segment has

b = b ( 0,1,0 ) and = (1,0,0 )

and is thus what kind of dislocation?
The glide plane is in the xy plane. The forces in the glide
plane are:

= b
L

F
= b yz y , x = b y , x
L

53

representation of the dislocation and nodal

forces
Add nodes to dislocation
and connect with
straight segments.

[1]
1

[3]

[2]
2

[4]
4

[5]
5

[6]
7

Force on segment i is constant along the segment:

s
Fi = b y , x i
Force on node i:
weighted average of force on segments

s s
for constant force on segments Fi = Fi + Fi+1 / 2

Line tension force on i: F = E +

i

i1
i

E is the line energy

54

velocity
[1]
1

[3]

[2]
2

[4]
4

[5]
5

[6]
7

velocity of a node is dependent on velocity of the segment

each point on the segment has a different force, and
thus different velocity (v = M F)
as Bulatov and Cai describe, the velocity is a weighted
average
they give an approximate
expression, which we use
n
Fi
n
vi
B ( i1 + i ) / 2
use Euler equation (or something better) for EOM
55

limitations of model code

no interactions between segments
no annihilation
...
However, it will show you how the basic method works.
And you can make movies.

56

a better calculation of an FR source

operation of a source in a thin film

57

Applications

58

1. Some successes:
small-scale
plasticity
59

dislocation dynamics simulations have proven

very successful for small-scale systems
4 examples showing how changes in interfaces change
plasticity:

small scales with free surfaces (micropillars)1,2

small scales with coated surfaces (micropillars)3

polycrystalline thin films with free surfaces4

polycrystalline thin films with coated surfaces5

1 Zhou,

Biner and LeSar, Acta Mater. 58, 1565 (2010).

2 Zhou, Beyerlein and LeSar, Acta Mater. 59, 7673 (2011).
3 Zhou, Biner and LeSar, Scripta Mater. 63, 1096 (2010).
4 Zhou and LeSar, Int. J. Plasticity 30-31, 185 (2012).
5 Zhou and LeSar, Comput. Mater. Sci. 54, 350 (2012).

small scale plasticity is the perfect problem for DDD - small

numbers of dislocations and straightforward boundary conditions
60

free surfaces: boundary element method

Total displacement and stress fields:

and

are the image fields that enforce the boundary conditions.

and

are the displacement and stress fields in an infinite medium from all dislocations.
El-Awady, Biner, and Ghoniem (2008)

61
61

sample preparation (no ions required)

Before relaxation
Density = 2.71013 m-2

D = 1.0, 0.75 and 0.5m

D:H=1:2

After relaxation
Density =1.81013 m-2

(Dotted lines are BEM meshes)

initial conditions shown to have large effect of calculated response in

small systems: Motz, Weygand, Senger, Gumbsch, Acta Mater 57,
1744 (2009).
62
62

stress-strain behavior of Ni
!

1 micron samples

calculated
onset of flow:
~ 0.003

1 micron samples

experimental
onset of flow:
~ 0.008

differences
arise from:
strain rate,
equation of
motion, ...
63

loading and flow

loading
loading regime:
intermittent flow, but
are they avalanches?

loading

flow

see Scale free intermittent flow in crystal plasticity, Dimiduk,

Woodward, LeSar, and Uchic, Science 312, 1188 (2006).

64

a simple model for a free-standing polycrystalline

thin film
simulation cell

free surfaces on all sides

9 grains, all with same
orientation
tension under constant strain
rate
all stress-strain-density
behavior calculated from
internal grain only, averaged
over 10 realizations

65

Bulk measurements based on X-ray diffraction and convergent beam electron diffraction support this view (e.g. [47
49]). The direct observation of frozen-in radii of bowedout dislocations in Al5 wt.% Zn right at LAGBs indicates

DD model of dislocation-grain boundary

interactions
(a) direct annihilation of incident dislocation

(a) twist annihilation of incident dislocation

sin-1 (b/s) / degree

0.1

str

on
g

rea
cti

/ GPa

on

we

ak
r

0.1 1
10
Y
X

Burgers vector [1 1 1] [1 1 1] [1 1 1] [1 1 1] 1 0 0

eac
t

102

Y
X

ion

this LAGB

10

We investigated t
tion processes associ
as a hexagonal netwo
vectors using DDD s
The LAGB poses
tration with a wide
depend on the slip sy
location as well as on
ments in the network
to the strength of ju
network dislocation
sequence of increasin
the weakest, to terna
gest interaction. The
boundary impenetrab

103

s/b [1 1 1] [1 1 1] [1 1 1] [1 1 1] 1 0 0
Burgers vector

.
Fig.
8.
Estimated dependence
of disintegration
LAGB penetration
resistance
(b) final network
precedes penetration.
(b) after successful penetration from situation
(a) #1
r / s=b lns=b on network mesh size s based on penetration strength
7. Network penetration sequence of negative 1=2 1 1 1"1 !
1 0 incident
Network penetration sequence of positive 1=2 1levels
1 1"1 !
1 0
incident at Fig.
observed
s=b
200 in this study.
dislocation (on semitransparent glide plane). Vertical gray bar indicates

low angle GB (0.3) modeled with

stress (full height = 1.5 GPa).

interpenetrating mesh of dislocations
modeled transmission of dislocations
through the LAGB

Fig. 6.
dislocation (on semitransparent glide plane). Vertical gray bar indicates
tress (full height = 1.5 GPa).

5. Concluding remark

Defining steady state

approximate constancy of
meant.

atomistic simulations are also

being done, but each
orientation and dislocation are
different

Simulation of dislocation penetration through a general low-angle grain boundary

B. Liu, P. Eisenlohr, F. Roters, D. Raabe, Acta Mater 60, 5380 (2012)
66

transmission model at grain boundaries

grain boundary/dislocation interactions (de Koning et al.
2002)
Outgoing dislocation, b2

Grain 2
GB
Grain 1

Residual dislocation, b

Incoming dislocation, b1

transmission stress adjusted

to match experimental
stress-strain relation

67

setting transmission stress: GB

data by Xiang et al (2006)

GB = 5 FR

68

comparison with experiment

700

D = 250 nm
D = 500 nm
D = 1000 nm
D = 1500 nm
Cu freestanding (Gruber et.al.)
Cu freestanding (Xiang et.al.)

Yield stress, y (MPa)

600

Excellent agreement
with experimental
stress-strain behavior,
but ...

500

400

is this a definitive test

of the method?

300

H/D = 1.0

200

100

5
-1

Reciprocal of film thickness, 1/H (m )

69

dislocation structures

D = 500 nm

H = 250 nm

H = 500 nm

H = 2 m
70

dependence
on D and H

dependence of yield
stress on D as a
function of film
thickness

approaches Hall-Petch
relation for thick films
(H = 1.5 m)

71

outlook
for small-scale samples, DD simulations have led the

way in developing understanding of the deformation

mechanisms
by controlling physics, we can examine the role of

different physical processes (e.g., cross slip)

limited by the assumptions mentioned earlier
with better connection to experiments (esp. structure),

we should be able to examine the correctness of the

predictions, which can guide the development of better
models

72

A challenge: bulk
plasticity
73

2. Can we model with DD the development of

dislocation substructures?
Single crystal Cu in single slip: Stage III

we will use as an example

a study of high strain rate
response done in
collaboration with Los
Alamos

1 m
Szekely, Groma, Lendvai, Mat. Sci. Engin. A 324, 179 (2002)
74

computational challenges
the dislocation density increases with stress

1/2

calculations slow with increasing stress, with the

computational time per time step increasing as

nnode L

tcomp n

2
node

time steps decrease with increasing dislocation density

typically use dynamic time stepping because of very

large dislocation interaction stresses (based on a

maximum distance a dislocation can move)
time steps are small (can be as small as 10-10 sec)
as density increases, have more pileups and larger

stresses and thus smaller time steps: t 1/

75

boundary conditions
must have transport of dislocations into and out of the

simulation cell
periodic boundary conditions can be a problem.
- Madec, Devincre and Kubin, On the use of periodic boundary
conditions in dislocation dynamics simulations, Solid Mechanics
and Its Applications 115, 35-44 (2004)

can avoid many of these issues with large cells (slow

calculations), but still have long-ranged interactions

may be best to use large cylinders whose properties

seem to be bulk-like

76

periodic boundary conditions

1.0

0.8

0.6

0.4

0.2

0.0
0.0

1
Frank-Read
source
0.2

0.4

0.6

0.8

1.0

77

a typical initial structure

!
assume an initial set of Frank-Read
sources randomly placed on
a
available slip systems

under stress, they bow out

and interact with other
dislocations
as the length of dislocations
increases, so does the
complexity of the microstructure

5 m cube
78

studies at high strain rate on Cu single crystals

effects of loading
on dislocation
microstructures
and deformation
Schmid factors:

[111] < [211] < [100]

interest in high
rates: e.g.,
impact damage
Plastic anisotropy in fcc single crystals in high rate deformation,
Wang, Beyerlein, and LeSar, IJP 25, 26-48 (2009)

5 m cubes
79

slip bands in [111] loading

viewed along [100]

viewed along [110]

5 m cube

Is this an artifact of the boundary conditions?

Plastic anisotropy in fcc single crystals in high rate deformation,
Wang, Beyerlein, and LeSar, IJP 25, 26-48 (2009)
80

dislocation density on 1 11 slip planes

4 1

= 10 s

5 1

= 10 s

p = 0.15%
B
slip bands are
approximately 0.2 m
wide
band spacing varies
between 0.1-1 m
spacing between fine
bands within bands is
0.01 m

10 s

10 s

without cross slip

81
81

hardening

= e + P

high-rate behavior differs from that at low rates, which is

controlled by obstacles and thermal activation
essentially all dislocations are glissile
slip bands act as deformation highways: processes in slipband formation are the dominant deformation mechanisms
What is the eect of the boundary conditions?
82
82

asurements is indicated in parentheses in the top row. The

s Khkl can be compared to the ones calculated from Eq. 2
solved shear stress of 10 MPa in copper, these values are
ee paths expressed in microns.

mesoscopic to macroscopic dimensions. For

purpose, use is made of a crystal plasticity c
(4, 15), which is a specific type of finite elem
code that takes into account the crystallogra
examine
the
based
nature of rate
dislocation
glide,on
deformation conditi
k (9) Kubin et
K112al.(3)used DD
K111to(3)
K001
(3)hardening
and lattice rotations during plastic flow. We
calculating
of the
1 0.015
10.42 the
0.4 mean
7.29free
1.6path4.57
0.3dislocations along various
integrated the
set
of dislocation-based equat
slip planes
(replacing
dependence
on
experiment
in
Kocks
(11.87)
(7.38)
(6.21)
on a meshed tensile specimen (fig. S1).

.6

Application
0.6 B to various metals at
14
s
room temperature:
they obtain qualitative, but not
0.4
12
quantitative pagreement with
experiment.
0.2
10
in many ways a 2D model
s
(mean free path on different slip
0
8
systems)
0.8
0
0.2
0.4
0.6
0.8
t

10016
14

80

(MPa)

log[ (m-2)]

[123]

model), assuming uniform densities per slip system.

2]

3. stage II hardening by DD simulations

Ni

12

60 10

/0

40
20

Cu

Ag

Al

4
2

00

5
0.1

10

0.215

20
0.3
/0

25

0.4 30

35
0.5

0.6

40

Kubin, Devincre, and Hoc, Intl. J.

Mater. Res.
(2009).
crossing
of the [001]
and [111]
curves
fcc single
crystals
at room tem- active slip systems.
ModelThenot
essentially
different
from
the was ex
Devincre,
Hoc,conventions
Kubin, Science
imentally observed.
This crossing occurs
becauseof
of aKoslowski
competition between
n using
traditional
for 320,
2D
phase-field
modeling
et
1745
(2008)
ress/strain curves were resolved on orientation dependencies of stages II and III. (B) Copper, [123] orienta

ientation effect and the occurrence

orientation. The strain hardening

al., resolved
PRL 93,
265503
(2004)
Densities (r) and
strains
(g) on the
primary (p) and secondary (s)
systems as a function of the total resolved strain, showing the transition betw
83

stage II: their conclusions

1. they have not solved stage II, though they made
progress
2. they believe that it is possible
3. they cannot predict the transition between stages

There is presently no generally accepted theory explaining

how and why organized dislocation microstructures
emerge during plastic flow. Kubin, Devincre, and Hoc, Intl.
J. Mater. Res. (2009).

84

IV. some more examples

strain hardening: e.g., Kubin and coworkers mapped out
aspects of Stage II strain hardening in fcc crystals, Devincre, Hoc,
and Kubin, Science 320, 1745 (2008); Kubin, Devincre, and
Thierry, Intern. J. Mater. Res. 100, 1411 (2009).
fatigue: Fivel and coworkers applied 3D dislocation simulations
to the early stages of fatigue, with applications to fatigue in: steel
[Dprs, Robertson, and Fivel, Phil. Mag. 84, 2257 (2004); ibid.
86, 79 (2006)], fcc materials [Dprs, Fivel, and Tabourot, Scripta
Mater. 58, 1086 (2008)], and precipitation hardened materials
[Shin, Robertson, and Fivel, Phil. Mag. 87, 3657 (2007)]
silicon: see work from K. W. Schwarz, for example, with
PARANOID code
Others of note: Weygand/Gumbsch/... : many calculations,
including small-scale plasticity
85

Outlook and
needs
86

Summary of some of the approximations

all atomistic-based processes described with models
most codes ignore partial dislocations
interactions are long-ranged
most materials are anisotropic not isotropic
we ignore the effects of lattice rotations caused by the

dislocations (except as a post-process)

boundary conditions are a challenge, with the modeling
of bulk plasticity being much more challenging than
modeling of small-scale plasticity
...
we need more sensitive tests to tell us what is and is not

important. Structures matter so microscopy is key.

87

Outlook for calculating bulk mechanical

properties
Simulations are limited by: size, boundary conditions, and

all the approximations mentioned above.

Limited success for DD in modeling dislocation

substructure development.
Thus, prediction of hardening has not been possible.
The present dislocation-based models for strain

hardening still have difficulties integrating elementary

dislocation properties into a continuum description of bulk
crystals or polycrystals. As a consequence, current
approaches cannot avoid making use of extensive
parameter fitting. - Devincre, Hoc, Kubin, Science 2008.

88

Needs
Better connection to experiment: The feedback loop

between theory, simulation and experiment for plasticity

and property modeling, T. M. Pollock and R. LeSar,
Current Opinion in Solid State and Material Sciences 17,
10-18 (2013).
As noted above, we need to link to experimentally

measured dislocation substructures

We will do this in collaboration with Marc DeGraef, using

methods he discussed Monday

With better connection to experiments (esp. structure), we

should be able to examine the correctness of the

predictions, which can guide the development of better
models.
89

(S)TEM Defect Simulations

Darwin-Howie-Whelan equations:
X0 e ig
dSg (z)
= 2isg Sg (z) + i
dz
qg
0
g

0
i(g
g )

1
1
e

+i
qg
g
g0
Extinction distance

g0 (r)

g0

g (r) 2g

Sg0 (z)

Nd
X
i=1

Displacement field

Ri (r) = 2g Rt (r)

Absorption length
Scattering matrix

Reformulate as a matrix problem:

dS(z)
= iA(r)S(z)
dz

S() = e

iA

S(0) = S()S(0)

Propagation = scattering matrix product

S(z) = S(zm )S(zm

From:
Marc De Graef

1 ) S(z2 )S(z1 )S(0)

90

Example
parallel illumination
From:
Marc De Graef

fcc-Al, 200 keV

100 nm foil, slightly bent
6 perfect dislocations, 3 stacking faults
50 spherical voids, 50 spherical inclusions
(020) systematic row, 9 beams

g020

bright field

dark field
91

Calculated (S)TEM results

calculation of the displacements

bi 1
ui R =
+
4 8
2
R
R,ij =
X i X j

( )

1
ikl bl R,pp + 1 kmn bn R,mi d k

Nsegment N
int

f (r ) d w f (r
=1

q=1

Complication: is discontinuous in dislocation plane.

From displacements, we will calculate the expected STEM

signal, enabling us to better compare with data.
92

acknowledgements
who did the work:
Caizhi Zhou, MUST
Zhiqiang Wang, U. North Texas
Irene Beyerlein, Los Alamos
Sriram Swaminarayan, Los Alamos
Marisol Koslowski, Purdue University
collaborators:
Nasr Ghoniem, UCLA
Dennis Dimiduk and coworkers, Air Force Research
Laboratory

93

Questions?

94