Loading in 2 Seconds...
Loading in 2 Seconds...
The Third International Workshop on DFT Applied to Metals and Alloys. Integrating Informatics with First Principle Calculations: Building a Materials Genome project. Krishna Rajan NSF International Materials Institute Combinatorial Sciences and Materials Informatics Collaboratory CoSMIC-IMI
Integrating Informatics with First Principle Calculations: Building a Materials Genome project
May 3rd 2007
Data + Correlations + Theory =Knowledge Discovery
Information is multivariate, diverse , very large and access / expertise is globally distributed
Ideker and Lauffenburger:(2003)
From a set of N correlated descriptors, we can derive a set of N uncorrelated descriptors (the principal components). Each principal component (PC) is a suitable linear combination of all the original descriptors. PCA reduces the information dimensionality that is often needed from the vast arrays of data in a way so that there is minimal loss of information
( from Nature Reviews Drug Discovery1, 882-894 (2002) : INTEGRATION OF VIRTUAL AND HIGH THROUGHPUT SCREENING Jürgen Bajorath ; and Materials Today; MATERIALS INFORMATICS , K. Rajan , October 2005
Functionality 2 =F ( x1 , x2 , x3 , x4 , x5 , x6 , x7 , x8 ……)
X1 = f ( x2)
X2 = g( x3)
PC 1= A1 X1 + A2 X2 + A3 X3 + A4 X4 …….
PC 2 = B1 X1 + B2 X2 + B3 X3 +B4 X4 …….
PC 3 = C1 X1 + C2 X2 + C3 X3 + C4 X4…….
Mooser-Pearson map (’59)
Villars map (’83)
Makino map (’94)
Modeling & calculation
For AB2, AB3, A2B3, A3B5
= 7832 compounds
For AB compounds
= 3916 compounds
Possible structure type
Establishing association rules for crystal
∆X = 0.85978
ΣVE = -0.30361
∆nav = -1.61573
∆nws1/3 = -0.93887
2x∆X = -0.14441
∆Φ* = -1.68529
Structure decision route
Possible structure types of AuBe2
OUTPUT: Structure type
1. MgCu2 (-3.65757 eV)
2. PbCl2(-3.60992 eV)
3. OsGe2 (-3.58157 eV)
4. CaF2 (-3.46498 eV)
5. AlB2 (-3.46430 eV)
First principles calculations
Chinget.al J. Amer. Ceram.Soc. 85 75-80 (2002)
1. Assess influence of latent
variables ( i.e. electronic structure parameters) on properties of known data
2. Establish heuristic relationships on database of all input variables instead of phenomenological relationships in bivariate manner
3. Use statistical learning to
predict new materials behavior
on new multivariate input data
4. Inverse problem approach to formulate quantitative structure-property relationships
along body diagonal of unit cell, Wycoff notation) Lattice sites along the cubic unit cell body diagonal in the spinel, AB2X4
As distance from origin of PCA plot increases, the intensity of DOS peak is increased
Can quickly determine:1) Co-Al and Ni-Al alloys have DOS mostly below EF and Ti-Al alloys and Al have DOS mostly above EF.
2) DOS at EF highest for Co3Al and Ti-Al alloys.
All of these conclusions are correct
Demonstrate the usefulness of PCA in screening for these above points --- more useful as number of systems examined is increased
Blue: E < EF
Red: E > EF
Normalize the DOS curves by shifting the curves so that EF of each alloy is at the same energy
Reason for doing this:
Two common things looked at in literature on DOS curves.
DOS relative to EF and DOS value at EF
We can quickly screen if an alloy’s DOS are primarily above or below EF.
Among a series of alloys, we can say which alloys have a high value of EF