Documente Academic
Documente Profesional
Documente Cultură
Precipitation Kinetics:
Guinier-Preston Zone Formation in an
Al 2 a/o Cu Alloy
RAMOLS
February 8, 2018
Marcel H.F. Sluiter
M.H.F.Sluiter@TUDelft.NL
TU Delft, Dept. MSE, virtual materials & mechanics
Delft, the Netherlands
1
outline
Some comments on modeling & computational
materials science
2
Computational Materials Science
Conventional: 1900~
model (theory) parameterization
parameters from experimental data
measurement determination of “unknown” parameter
or description/prediction of phenomenon
3
Computational Materials Science
Conventional: 1900~
model (theory) parameterization
parameters from experimental data
measurement determination of “unknown” parameter
or description/prediction of phenomenon
4
Computational Materials Science
Conventional: 1900~
model (theory) parameters (from experiment)
measurement determination of “unknown” parameter
or description/prediction of phenomenon
5
Computational Materials Science
Conventional: 1900~
model (theory) parameters (from experimental data)
measurement determination of “unknown” parameter
or description/prediction of phenomenon
6
Computational Materials Science
Conventional: 1900~
model (theory) parameters (from experimental data)
measurement determination of “unknown” parameter
or description/prediction of phenomenon
7
Computational
Ab initio modeling
Materials Science
Integrated computational materials engineering:
Big Data & the materials genome
What is next?
THIS?
8
Understanding & Insight
9
Atomistic potentials
• ab initio calculations remain too slow for large scales
• Constructing potentials (force fields) is labor intensive
automated tool for generating potentials
10
Lattice gas models
Solid solutions and precipitation: Lattice gas models
Lattice gas:
fixed lattice
at every lattice position
one and only one
atom
multiple atom species
(vacancy is also atom
species)
For a set of structures “s” both H(s) and xa(s) are known
inversion gives Ja
CE: Structure Inversion
How to find the best set of clusters/correlations a?
Al Cu
22
Lattice gas models
Solid solutions and precipitation: Lattice gas models
Thermodynamics (phase diagrams, etc.)
Now realistic kinetics
Vacancy formation energy depends strongly on neighborhood
Barriers for diffusion also depend STRONGLY on neighborhood
23
GP zones in Al 2 a/o Cu at 450 K
AA6061 at room T
with and without Sn
addition
31
Software
Cluster expansion & cluster
variation method:
iCVM
m.h.f.sluiter@tudelft.nl
32
Cluster expansions of diffusion
activation barriers (& attempt freq)
(Non)Constant diffusion
activation barriers
Materials Properties
Most basic properties: thermodynamics!
Derivatives of the energy of a material:
dE
CV =
dT V , Ni ,... d 2E
B = Vo
dV 2 V =V d 2E
o Cij =
d�i d�j
a (T ) � V (T ) : min[G (T , P = Po )]
35
Computational Databases III
Extended by Ceder group at MIT: >154 experimentally known
x-tal structures for >82 binary alloys (ternary alloys are not
public) datamining structure predictor. Fisher et al. Nat.
Mater. 5, 641 (2006)
circles: expt
triangles: other calcs
Blue: magnetic
39
Linear Q(VX)
INSIGHT! 40
fcc
43
44
About 70 bcc
superstructures found!
45
46
Historically:
• 1) composition + (crystal) structure a property
• 2) increasing sophistication to get that one property
• 3) each composition+structure = individual case
• 4) reverse: properties composition + structure
generally NOT possible
We cannot do it yet!
48
Finding Compounds were we know of
none! E.g. Au-Si and Au-Ge
49
E Tasci, M Sluiter, A Pasturel, P Villars, Acta Materialia 58, 449 (2010)
Mimic Nature …
50
Which crystal
structures
match?
51
Si (blue) completely
surrounded by 8 Au (red)
53
54
Quite unlike
Au-Si structure
but
Ge atom again
Au neighbors
only (6)
55
56
Summary (2)
Databases: amazing unforeseen applications
finding superstructures
finding compounds in non-compound forming alloys (Au-
Si, Au-Ge)
59
CE: Why a linear form?
Why “completeness”?
invariance when change of choice of CF