for the
Materials Scientist
Dane Morgan
University of Wisconsin, Madison, WI
computing power:
exponential scaling with
time
O(N3)
Run-time: polynomial
scaling with number of
atoms
Burkett, T. et. al. Phys. Rev. Lett. 93 (2004) Norskov, J. et. al. MRS Bulletin 31 (2006)
Calculating properties of
realistic nanostructures
ab initio
Machine learning in
Computational Materials Design
wrap-up
NGDM, October 10, 2007
computational materials design strategies
getStablePhases(i);
...
...
Depends on which
calculateProperty(i); phases are stable and
i = nextChemistry(); their structure
}
getStablePhases(i);
...
... Machine Learning
calculateProperty(i); needed on
Depends here !! phases are
which
i = nextChemistry(); stable and their structure
}
Material
DFT Code Property
Predictions
Material
DFT Code Property
Predictions
Database
Machine Learning of
Framework Computed and
Experimental
results
'Commodity'
computational resources
Open source
electronic structure
software
Computing budget
First Entry: 1913 ~50k compounds/year
# of entries: 100,243
# usable compounds: 29,962
NGDM, October 10, 2007
The structure search problem
complex energy
landscape
Proposed Solutions
Calculate energy of a finite set
of structure prototypes
Proposed Solutions
Proposed Solutions
Heuristic Rules
Use previous experiments to suggest
what to calculate
How ?
Identify a set of simple parameters
based on alloy constituents
Heuristic Rules
Plot stable structures in space of
parameters 1986: Pettifor
1983: Villars
r A ,B nate
Heuristic Rules
Plot stable structures in space of
parameters 1986: Pettifor
1983: Villars
r A ,B nate
Heuristic rules efficiently code
historical knowledge
provide transfer of knowledge
Knowledge Base
Experimental Data
Pauling File binaries edition (Villars, P. et. al. J. of Alloys and Compounds, (2004))
3975 non-unique
compounds
4263 compounds total
x= x A , x0 , , x 1 , , xB
2 Low temperature state of alloy
x= x A , x0 , , x 1 , , xB
2 Low temperature state of alloy
Material
DFT Code Property
Predictions
Set of likely
structure candidates
Machine Learning Database
Framework of
Computed and
p x ∣e Experimental
results
Pair Cumulant
p xi , x j
gij x i , x j =
p x i p x j
1 uncorrelated
0
NGDM, October 10, 2007
how probabilities represent physics of mixing
p xi , x j
gij x i , x j =
p x i p x j
8.48 Fe3C
MgCu2
1 1
0 4 3
1 2 3 1
3 4
2
cB
Data from Pauling File, Binaries Edition
p xi , x j
gij x i , x j =
p x i p x j
8.48 Fe3C
MgCu2 ~0
1 1
0 1 2 3 1
4 3 3 4
2
A
B
A
A A
B B
A
A
B
Mutual Information
Ii , j= ∑
xi , x j
p x i , x j log
p xi , x j
p xi p x j
Ii , j= 〈 log [ g ij xi , x j ] 〉
Each element of
matrix is correlation
between Xi and Xj
degree of correlation
Ii , j= ∑
xi , x j
p x i , x j log
p x i , x j
p xi p x j
e.g.,
Xi=”AB prototype”
and
Xj=”A2B prototype”
??
LiRh MgCu2 CuPt7 a.k.a. MgPt7
AlB2
Suggested
Known phases
phases
Set of likely
structure candidates
Machine Learning Database
Framework of
Computed and
p x ∣e Experimental
results
~28
candidates
req'd for freq.
10 candidates
--> 95%
chance of
seeing GS !!
Including structure
Independent Variables correlation
Nature Materials, 6, 641-646, 2006
NGDM, October 10, 2007
Some open questions
...
...
Now have efficient
calculateProperty(i);
i = nextChemistry(); tool for this
}
Material
DFT Code Property
Predictions
Set of features
● Charge Density
● Total energy
● Bulk moduli
● Coordination
DFT Code
● Bond strength
● Bond character
● Magnetic moments
● Polarization
● ...
Material Properties
DFT Code -catalytic activity
-conductivity
-plasticity
-voltage/energy density
Machine Learning
Database
of Framework
Computed (functional mapping)
results
1335 alloys
3975 non-unique
compounds