Slide 1

Slide 1 text

Computer–Accelerated Materials Design Prof. Aron Walsh Thomas Young Centre Department of Materials EFL Lunchtime Webinar (May 2020)

Slide 2

Slide 2 text

Theory & Simulation of Crystals

Slide 3

Slide 3 text

Decrease Technology Timeline J.P. Correa-Baena et al., Joule 2, 1410 (2018) Use simulations & statistics to reduce barriers for synthesis, device assembly, and diagnostics

Slide 4

Slide 4 text

Materials Design INPUT OUTPUT Property or Metric Composition and Structure

Slide 5

Slide 5 text

What is Computational Chemistry? “Using mathematical methods for the calculation of chemical properties or for the simulation of molecular/materials behaviour. It also includes, e.g. synthesis planning, database searching, combinatorial library manipulation” IUPAC (2014) IBM 5MB hard drive (1956) IBM quantum computer (2020)

Slide 6

Slide 6 text

Webinar Outline A. Past Interacting ions and electrons B. Future Power of data and statistics

Slide 7

Slide 7 text

Models of Interacting Elements erg was a popular unit of energy (10-7 J) J. E. Jones, Proc. Roy. Soc. A 106, 463 (1923) = !" − # Repulsive Attractive Molecular mechanics (MM): application of classical mechanics to model chemical systems. Often analytic functions (bond stretching, bending, etc.) John Lennard-Jones (1894–1954)

Slide 8

Slide 8 text

Models of Interacting Elements Some models implemented in GULP https://gulp.curtin.edu.au Molecular mechanics (MM): application of classical mechanics to model chemical systems. Often analytic functions (bond stretching, bending, etc.)

Slide 9

Slide 9 text

Models of Interacting Elements Molecular mechanics: simulations of >1 billion atoms now possible (e.g. protein folding, radiation damage, crystallisation, correlated diffusion) Y. Shibuta et al, Nat. Comm. 8, 10 (2017) Solidification of Fe with Finnis - Sinclair potential using molecular dynamics

Slide 10

Slide 10 text

Models of Interacting Electrons Quantum mechanics (QM): numerical solutions of the Schrödinger, or relativistic Dirac, equation to describe electron distributions (chemical bonds) The master equations appear simple, but for >1 electron these partial differential equations cannot be solved exactly and require approximations Many-body wavefunction

Slide 11

Slide 11 text

Models of Interacting Electrons Density functional theory (DFT) replaces the many-body N-dimensional electronic wavefunction by the 3-dimensional electron density Quantum mechanics (QM): numerical solutions of the Schrödinger, or relativistic Dirac, equation to describe electron distributions (chemical bonds) Walter Kohn (1923–2016)

Slide 12

Slide 12 text

Computational Bottleneck of DFT O(N3) scaling due to matrix diagonalisation, e.g. 10⨉ atoms cost 1000⨉ more processing time Density functional theory scaling data in FHI-AIMS courtesy of Volker Blum https://aimsclub.fhi-berlin.mpg.de Using ELPA massively parallel eigensolvers https://gitlab.mpcdf.mpg.de/elpa

Slide 13

Slide 13 text

Recovering Chemical Concepts Covalent bond, ionic bond, oxidation state… Simple concepts can be difficult to extract from QM Five (of the many) ways to assign electron density to atomic centres in crystals “Oxidation states and ionicity” A. Walsh et al, Nature Mater. 17, 958 (2018)

Slide 14

Slide 14 text

Open Source Materials Modelling Selection of DFT packages for calculating electronic structure and properties of crystals Code Features Link ABINIT Plane wave basis set – many property calculators https://www.abinit.org CP2K Mixed basis set – fast for large and complex systems https://www.cp2k.org GPAW Flexible basis set – modern python interface https://wiki.fysik.dtu.dk/gpaw Quantum Espresso Plane wave basis set – many property calculators http://www.quantum- espresso.org Starting point: Atomic Simulation Environment https://wiki.fysik.dtu.dk/ase/

Slide 15

Slide 15 text

Open Source Materials Modelling Selection of first-principles properties that are used in materials science Class Property Measurement Structural Ground-state crystal structure X-ray/neutron/electron diffraction Vibrational Phonon frequencies and lifetimes X-ray/neutron scattering IR/Raman spectroscopy Thermodynamic Internal and free energy changes Stability and chemical reactivity Point Defects Concentrations, barriers, diffusion rates Impedance spectroscopy; isotope tracers, etc. Extended Defects Changes at surfaces and grain boundaries Tomography; microscopy; spatially resolved probes

Slide 16

Slide 16 text

Thermal Transport in NMC Batteries Mixtures of LiNiO2 , LiMnO2 , and LiCoO2 (NMC) are used for high-performance Li-battery cathodes Hui Yang et al, ChemRxiv Preprint Server (2020)

Slide 17

Slide 17 text

Webinar Outline A. Past Interacting ions and electrons B. Future Power of data and statistics

Slide 18

Slide 18 text

Data-Driven Science Agrawal and Choudhary, APL Materials 4, 053208 (2016) Entering a new phase of scientific discovery, harnessing advances and tools from data science

Slide 19

Slide 19 text

Machine Learning (ML) Chemistry Statistical algorithms that learn from training data and build a model to make predictions Quick-start machine learning guide: K. T. Butler et al, Nature 559, 547 (2018) Applications include: • Faster (or better) computational tools • New knowledge from existing data • “Self-driving” research workflows

Slide 20

Slide 20 text

Machine Learning (ML) Chemistry Images from: https://vas3k.ru/blog/machine_learning/ Predict a category, e.g. decision trees to predict reaction outcome Predict a value, e.g. regression to extract a reaction rate Group by similarity, e.g. high-throughput crystallography Maximize reward, e.g. reaction conditions to optimize yield

Slide 21

Slide 21 text

Machine Learning (ML) Chemistry https://www.coursera.org/specializations/mathematics-machine-learning

Slide 22

Slide 22 text

Machine Learning (ML) Chemistry https://xkcd.com/1838/

Slide 23

Slide 23 text

Open Tools Example of Python tools that can benefit all research (whether “theory” or “experiment”) Transform Model Visualise

Slide 24

Slide 24 text

No content

Slide 25

Slide 25 text

Open Data • Reproducibility – direct comparison with published literature beyond static tables and figures, e.g. raw spectra • Reuse – facilitate meta-studies comparing results from multiple samples, e.g. variation in UV-vis features / band gap assignments • Statistical Models – power of machine learning depends on the quantity, quality, and diversity of training data For data-management guidance: https://www.go-fair.org/fair-principles/

Slide 26

Slide 26 text

Crystallography in the Lead https://www.ccdc.cam.ac.uk / https://checkcif.iucr.org Cambridge Structural Database (from 1960) …. 1 million 2019 Human and Machine Readable Community Databases of Known Materials Standard Format

Slide 27

Slide 27 text

Thermoelectric Devices

Slide 28

Slide 28 text

Energy Storage Devices

Slide 29

Slide 29 text

Solar Cells https://doi.org/10.1088/2515-7655/ab2338

Slide 30

Slide 30 text

Unexplored Crystal Space D. W. Davies et al, Screening all inorganic materials, Chem 1, 617 (2016) Large number of candidate systems • Composition (AB, A2 B3 …) • Structure (polymorphs) • Stoichiometry (defects) • Morphology (facets)

Slide 31

Slide 31 text

Time to Curate your Data https://cen.acs.org/content/cen/articles/98/i20/time- curate-data-Use-shutdown.html Use lab downtime to launch a new data regime

Slide 32

Slide 32 text

Conclusions Materials Design Roadmap https://doi.org/10.1088/1361-6463/aad926 Slides: https://speakerdeck.com/aronwalsh @lonepair There are many exciting developments in materials modelling and design. The transition from analytical to numerical methods is now being augmented by statistical approaches – limited by available data! Thanks to research group, collaboration network, and the EFL team!