Synthesis and Metrology of Combinatorial Materials Libraries
in discussion with combinatorial materials science community, computational materials science community
Ichiro TakeuchiUniversity of Maryland
Supported by DOE, DOD (ARO, AFOSR, ONR), NSF
Evolution of combinatorial approachesOvercoming challenges
Throughput trade-off
High-throughput synthesis
High-quality synthesis
Rapid characterization
Throughput trade-off
On-the-fly real-timeanalysis
Integration with theory, databases,standards, etc.
Accurate quantitativecharacterization
20 years ago (1990s):
Superconductor library (Xiang/Schultz)
Dielectric composition spread (van Dover)
Combinatorial electrochemistry (Mallouk)
the renaissance of the high‐throughput approach
Library synthesis under epitaxial growth conditions
F. Tsui (UNC)H. Koinuma, M. Lippmaa, T. Chikyow (COMET)
Materials are of the same quality as single composition depositions
J.‐C. Zhao (OSU)
Diffusion multiples
J. Vlassak(Harvard)
Combinatorialmicrocalorimetry:
Glass transition,
latent heat, et.
Combinatorial alloy development
Combinatorial arc melting hearth (2014)
J. Cui (PNNL)
Spin-offs from combinatorial approachesCharacterization tool: microwave microscope
Xiang, Wei (LBNL, 96)(US Patent 5821410)
Dielectric mapping (LBNL, 98)
1”
Atomic resolution microwave microscope
(UMD & Intematix, 2010)
(Agilent, 2008)N9416S Scanning Microwave Microscopy
10 years ago: we all had ideas about combinatorial workflow
Where are they now? Do we (still) use them? If not, why not?
PI: A. Zaban (Bar Ilan)
Major combinatorial projects around the world
DOE Hubs:
Combinatorial approaches are integral part of major programs
JCESR
3.84 M Euro for 3 years
prepared by RIST
E. Amis, M. Fasolka, K. Beers (NIST)
Advanced microfluidics were developed
NIST Combinatorial Methods Center(1998 – 2007)
MGI and integration of high‐throughput computations and experiments
• Combinatorial experiments are the natural counterpart to computational efforts
• Accelerated experimental validations of computed results
• But how do we couple the efforts in a meaningful way?
• How do we close the gap between theory and experiments?
“ Ichiro, there is no communication gap between theorists and experimentalists. We just need to go drinking more often” ‐ famous MGI theorist
Integration of theory and experiments in high-throughput materials science:
w/ S. Curtarolo, Duke
Step 1 Step 2 Step 3
Integrated materials discovery engine
Synchrotron diffraction set up at SSRL The entire 3” wafer (300 spots) can now be measured in 2 hrs
Rapid structural mapping of combinatorial wafers at synchrotron:moving from demonstration experiments to routine measurements
XRF carried out simultaneously
Each wafer produces: 300 MB to 2 GB of image data
Reflection set up with in-situ heaterTransmission set up
w/ J. Gregoire A. MehtaM.J. Kramer, Ames
In-situ reaction experiments
Non‐negative matrix factorization
Other techniques
• Multidimensional scaling with k-means
• Spectral clustering
• New graphical model and independent component analysis techniques
Fast and computationally inexpensive
Ensemble approach
FCC Fe22%
BCC Fe41%
FCC FePd31%
Machine learning techniques for automated XRD analysis (C. J. Long, A. G. Kusne)
Clustering: Mean shift theory
Each point on the ternary phase diagram is one X‐ray spectrum (expt or simulated)
Points on binary lines are simulated spectra from ICSD
They are rapidly mined/analyzed together
Integrating databse (ICSD) with combiXRD data
Fe
Fe0.6Pd0.4
Fe0.6Ga0.4
A. G. Kusne (NIST)
Algorithm used: mean shift theory
XMCD (X-ray magnetic circular dichroism)
E. Arenholz, ALSNext: high-throughput ARPES (angle-resolved photoelectron spectroscopy)?
Element sensitive magnetic characterization
Combinatorial investigation of hard/soft magnetic coupling
Data mining (predictions) for novel materials:Need experimental databases
Problem: no appropriate databases exist
Solution: create our own databasefrom literature
Multi volume compilation of raw published data in forms of figures (graphs and tables) from articles
Excel sheet (compiled by 3 undergrads over 2 years)
Experimental Materials Databases- Exist in many fields, but often separately and in disparate
forms
- What you really need does not exist in databases or is not accessible
- Data-mining of experimental databases is valuable
- Some exceptions are crystallographic databases (ICSD 150,000 entries) and the NIMS databases
http://www.cmdnetwork.org/content/cmdnetwork/about.jsp organized by ASM
Data mining for novel multifunctional materials
Problem: no appropriate databases exist
Solution: create our own databasefrom literature
Multi volume compilation of raw published data in forms of figures (graphs and tables) from articles
Excel sheetSuperconductor database
Decades of tedious manual labor
Machine learning/reading techniques can be used to automatically go through PDF files
Automated knowledge discovery from multilingual science PDFs
w/ Synthesis Partners, LLC
Successfully demonstrated on 200 selected journal articles: Table of key magnetic parameters
So what are the challenges and opportunities now?
Combining combinatorial experiments with (high‐throughput) computational approaches can greatly help facilitate accelerated materials discovery
Handling large amount of data: need infrastructure: software development; need to do boring things like database curation
More advanced characterization tools are out there: XMCD, ARPES
Problem:lack of funding for concerted efforts
Can we establish a (virtual) combinatorial/high-throughput materials synthesis center?
Many combinatorial synthesis and characterization techniques exist in a distributed way throughout the world (some duplications are needed!)
A Center would allow/force meaningful development of (centralized) data repository, management system(s) – solid foundation for theory/experiment integration also with databases
A. Mehta
(similar proposal being developed in China – Yalin Lu (USAFA)
Dec 2013
Combi2014
8th International Workshop on Combinatorial Materials Science and Technology
6 – 8th October 2014 CAIRNS, AUSTRALIAhttp://www.csiro.au/Combi14