Materials Design and Discovery, and the Challenges of the Materials Genome Nicola Marzari, EPFL
The good news: Europe at the forefront in codes for materials simula@ons To name just a few: Quantum- ESPRESSO (Trieste/ Cineca/EPFL), CP2K (UZH), CPMD (IBM Zurich), Abinit (Louvain), VASP (Wien), Castep and Onetep (Cambridge), SIESTA (Madrid), Wien2K (Wien), FLEUR (Julich), ExciUng/ELK (Berlin/Halle) E.g. 700 papers published in 2012 using Q- E
Accelera@ng materials design
Digging for gold
5 or shopping for diamonds?
Some background on first- principles (i.e. quantum mechanical) materials simula@ons
AROSA, CANTON DES GRISONS, 27 DECEMBER 1925 at the moment i am struggling with a new atomic theory. i am very optimistic about this thing and expect that if I can only solve it, it will be very beautiful. e. schrödinger
A lab that doubles every 14 months Total computing power in TOP500 10 16 Computing power (FLOPs) 10 15 10 14 10 13 10 12 6/1993 6/1994 6/1995 6/1996 6/1997 6/1998 6/1999 6/2000 6/2001 6/2002 6/2003 6/2004 6/2005 6/2006 6/2007 6/2008 6/2009 6/2010 11/1993 11/1994 11/1995 11/1996 11/1997 11/1998 11/1999 11/2000 11/2001 11/2002 11/2003 11/2004 11/2005 11/2006 11/2007 11/2008 11/2009 Jack Dongarra, CCP 2011
! Most cited papers in the history of APS! Journal! #!cites! Title! Author(s)! 1! PRB!(1988)! 39190! Development!of!the!ColleBSalvetti!CorrelationBEnergy!! Lee,!Yang,!Parr! 2! PRL!(1996)! 25452! Generalized!Gradient!Approximation!Made!Simple! Perdew,!Burke,!Ernzerhof! 3! PRA!(1988)! 22904! DensityBFunctional!ExchangeBEnergy!Approximation!...! Becke! 4! PR!(1965)! 20142! SelfBConsistent!Equations!Including!Exchange!and!Correlation!! Kohn!and!Sham! 5! PRB!(1996)!!!13731! Efficient!Iterative!Schemes!for!Ab!Initio!TotalBEnergy!! Kresse!and!Furthmuller! 6! PRB!(1976)! 13160!!!!Special!Points!for!BrillouinBZone!Integrations! Monkhorst!and!Pack! 7! PRB!(1992)! 10876! Accurate!and!Simple!Analytic!Representation!of!the!Electron!! Perdew!and!Wang! 8! PRB!(1999)! 10007! From!Ultrasoft!Pseudopotentials!to!the!Projector!Augmented!! Kresse!and!Joubert! 9! PRB!(1990)! 9840! Soft!SelfBConsistent!Pseudopotentials!in!a!Generalized!! Vanderbilt! 10! PR!(1964)! 9789! Inhomogeneous!Electron!Gas! Hohenberg!and!Kohn! 11! PRB!(1981)! 9787! SelfBInteraction!Correction!to!DensityBFunctional!Approx.!! Perdew!and!Zunger! 12! PRB!(1992)! 9786! Atoms,!Molecules,!Solids,!and!Surfaces!B!Applications!of!the!! Perdew,!Chevary,!! 13! PRB!(1986)! 9313! DensityBFunctional!Approx.!for!the!CorrelationBEnergy!! Perdew! 14! PR!(1934)! 9271! Note!on!an!Approximation!Treatment!for!ManyBElectron!Systems! Moller!and!Plesset! 15! PRB!(1994)! 9100! Projector!AugmentedBWave!Method! Blochl! 16! PRL!(1980)! 7751! GroundBState!of!the!ElectronBGas!by!a!Stochastic!Method! Ceperley!and!Alder! 17! PRL!(1987)!!!!!!7663! Inhibited!Spontaneous!Emission!in!Solid>State!Physics!! Yablonovitch! 18! PRL!(1986)! 7589! Atomic!Force!Microscope! Binnig,!Quate,!Gerber! 19! PRB!(1991)! 7425! Efficient!Pseudopotentials!for!PlaneBWave!Calculations! Troullier!and!Martins! 20! PRB!(1993)! 6925! Ab!initio!Molecular!Dynamics!for!Liquid!Metals! Kresse!and!Hafner! 21! PR!(1961)! 6467! Effects!of!Configuration!Interaction!on!Intensities!and!Phase!Shifts! Fano! 22! PR!(1957)! 6260! Theory!of!Superconductivity! Bardeen,!Cooper,!Schrieffer!!
What can first- principles do for me? Fairly straightorward, but *fundamental*: equilibrium structures, thermodynamic stability, thermomechanical properues, electronic structure, energeucs and reacuons Harder: vibrauonal and magneuc spectroscopies (IR, Raman, NMR, EPR), XPS/XANES, BCS superconducuvity, opucal absorpuon, phase diagrams Jedi master: thermal and electrical conducuviues, photoemission spectroscopies Predic@ve accuracy is a key challenge
Electrical and thermal transport from first- principles Nanoscale devices for microelectronics (either silicon or carbon based) Bulk or nanostructured thermoelectrics (harvesung waste energy, cooling) Source: Nature Materials
Microscopic origin of transport Density- func@onal perturba@on theory hxp://www.quantum- espresso.org
Valida@on through spectroscopies P.H. Tan et al., Nature Materials (2012)
Thermal conduc@vity: Si 1- x Ge x J. Garg, N. Bonini, B. Kozinsky and N. Marzari, Phys. Rev. LeX. (2011)
Mean free paths: Si 1- x Ge x J. Garg, N. Bonini, B. Kozinsky and N. Marzari, Phys. Rev. LeX. (2011)
Engineering high- performance thermoelectrics
Making sense of experiments: graphene Simula@ons: A. Cepellob, N. Bonini, and N. Marzari (2013) expts from A. Balandin (Nature Materials 2011)
ρ e ph (Ω) 25 0 100 10 1 Theory Experiment (Efetov & Kim) Upper bounds on intrinsic mobility 100 200 300 400 T (K) n (10 13 cm 2 ) 1.36 2.86 4.65 6.85 10.8 (c) 100 200 300 400 T (K) n (10 13 cm 2 ) 1.36 2.86 4.65 6.85 10.8 (d) 0.1 10 100 T (K) Theory 10 100 T (K) Experiment (Efetov & Kim) C. H. Park, N. Bonini, G. Samzonides, B. Kozinsky, and N. Marzari (2013)
How many materials do we know? ICSD (InternaUonal Crystallographic Structure Database): 143,000 entries Very lixle is known for the vast majority of these Basic properues for all of them can be calculated in a few hours on petaflop- class machines
Materials design: Once a material property can be calculated accurately from quantum simula@ons, it becomes straightorward to explore rapidly and systema@cally thousands of compounds for improved performance.
Proper@es/performance è descriptors è high- throughput searches Courtesy of S. Curtarolo / Nature Materials (2013)
Novel ferroelectric perovskites G. Pizzi, A. Cepellob, S. Halilov and in collabora@on with M. Fornari, G. Ceder, R. Armiento, B. Kozinsky
Ferroelectric phase transi@ons Cubic Orthorhombic Tetragonal Rhombohedral Preliminary results F(T) of rhombohedral phase taken as reference Crossing points where phase transitions occur T (K) DFT Exp T-C 670 403 O-T 309 278 R O T C R-O 102 183
Run computer laboratories with a materials informa@cs platorm: prepare, organize, run, store, datamine, share, repeat Developing AIDA ( Automated Infrastructure and Database for Ab- ini@o calcula@ons - EPFL+Robert Bosch RTC) a librarian: job tracking and organizauon of all present and future acuviues of a group a library: databases of jobs and job results a network of connected libraries: possibility to share databases with of other groups
User interface Select: Structure Codes Desired proper@es AIDA server Backed up data storage and databases Codes database Structures Database Calcula@on Database: History Permanent results storage Calcula@on Data analysis CSCS Input file genera@on Output parser Work submission Execu@on on clusters User interface
AIDA network Clusters Users Group 1 AIDA server Databases Private data Public/shared data Permissions for access to each entry (calcula@on, structure,...) UUIDs associated to each entry Scripts to download/upload entries from/to remote DBs or sync two AIDA DBs Some data shared AIDA server AIDA server Group 2 Some data shared Group 3
Rapid and long- term reproducibility of calcula@ons: workflows and dependencies
Conclusions First- principles calcula@ons have reached predic@ve accuracy for many (but not all!) materials and materials proper@es Intelligent high- throughput searches are the next, natural step, and lead to materials design and discovery Computa@onal laboratories can and should be organized with an informa@cs infrastructure to organize, run, store, datamine, share, and repeat calcula@ons