Real and virtual screening for materials discovery through first principles calculations

Size: px

Start display at page:

Download "Real and virtual screening for materials discovery through first principles calculations"

Gwenda Morris
5 years ago
Views:

Materials Science and Engineering, Kyoto University, JAPAN 2

University, JAPAN 3 Center for Materials Research by Information

1 Real and virtual screening for materials discovery through first principles calculations Isao Tanaka 1,2,3,4 1 Department of Materials Science and Engineering, Kyoto University, JAPAN 2 Elements Strategy Initiative for Structural Materials, Kyoto University, JAPAN 3 Center for Materials Research by Information Integration, NIMS, JAPAN 4 Nanostructure Research Laboratory, Japan Fine Ceramics Center, JAPAN

2 Enormous number of chemical combinations Periodic Table Number of chemical elements Number of chemical combinations (only for simple composition ratio) 1 ~100 2 ~100,000 3 ~10,000,000 4 ~1,000,000,000 2

3 Inorganic Crystal Structure Database (ICSD) World largest database for known inorganic crystals. 177,000 crystal structures 40,000 structures excluding duplicates, incompletes, etc. Many systems are yet-unexplored! 3

4 First Principles Calculations Useful to fill the gap between enormous number of chemical combinations and experimental database. Electronic structure calculations entirely based on quantum theory without using empirical parameters. Total energy Force/Stress Electronic/Magnetic structures Physical properties But this is not a magic box!! 4

5 Vast chemistry space to explore Simple chemical combinations AaBbyCcDd (a,b,c,d <10) ~1B experimental database aflowlib for crystal structure OQMD ~1.3M hypothetical ~400k hypothetical ICSD compounds compounds ~40k for given structures for given structures Materials Project ~60k first principles database IMPORTANT! First principles calculation requires an initial guess of structure, otherwise it becomes too expensive to explore the vast space.

6 Vast chemistry space to explore Simple chemical combinations AaBbyCcDd (a,b,c,d <10) ~1B thermodynamically (meta)stable compounds experimental database for crystal structure ICSD ~40k OQMD ~4.5k candidates for thermodynamically stable ternary compounds (Chris Wolverton s talk) thermodynamically unstable compounds

7 Vast chemistry space to explore thermodynamically (meta)stable compounds experimental database for physical properties >>1k experimental database for crystal structure ICSD ~40k Systematic method to add as-yet-unknown & stable compounds to the inorganics database is desired. It is still worthy to explore the space of known compounds.

8 High throughput screening 1 Real Screening DFT database Construction of DFT database High throughput screening with descriptors Experts knowledge Physical Rule

9 High throughput screening 2 Virtual Screening 9

10 Contents Materials discovery through first principles calculations 10

11 11 Accelerated discovery of a novel Sn(II)-based oxide for daylight-driven photocatalyst Hiroyuki Hayashi, Shota Katayama, Takahiro Komura, Yoyo Hinuma, Tomoyasu Yokoyama, Kou Mibu, Fumiyasu Oba and IT Advanced Science 9, (2016) Hiroyuki Hayashi

12 12 Sustainable energy: Sunlight Solar cells & daylight driven photocatalysts Intensity (arb. units) Target band gap range Solar cell absorber: ev Photocatalyst: ev IR AM1.5 spectrum *) UV Photon energy (ev) Band gap (ev) Ga 2 O 3 SnO 2 ZnO In 2O 3 TiO 2 Oxides Cu 2 O BN AlN GaN InN Nitrides Popular oxide/nitride semiconductors UV IR *)

13 Lone Main pair group oxides: element oxides Sn(II), Pb(II) & Bi(III) Energy Characteristic electronic structures Lone pair Sn O Narrower band gap Wider valence band Sn 5p cation sp Sn 5s O 2p

13 13 Lone Main pair group oxides: element oxides Sn(II), Pb(II) & Bi(III) Energy Characteristic electronic structures Lone pair Sn O Narrower band gap Wider valence band Sn 5p cation sp Sn 5s O 2p Conduction band Valence band Main group Sn(II) element oxides oxide Band gap (ev) Ga 2 O 3 Band gap (ev) SnO 2 ZnO In 2O 3 TiO 2 Oxides Cu 2 O BN AlN Ga 2 O 3 SnO 2 ZnO GaN In 2O 3 TiO 2 InN Cu 2 O BN AlN PbOGaN Bi 2 O 3 Sn 3 O 4 SnS SnOInN Lone pair Nitrides Oxides materials Nitrides Popular oxide/nitride Popular oxide/nitride semiconductors semiconductors UV IR

14 14 Known oxides: Sn(II), Pb(II) & Bi(III) Sn(II) M O Pb(II) M O Compounds System registered in ICSD Sn(II) M O 47 Pb(II) M O 210 Bi(III) M O 213 Bi(III) M O Sn(II) oxides are yet unexplored!

15 15 Chemical composition of interest Sn(II) M O SnO MO q/2 4A 6A transition metal compounds Commonly used for photocatalysts ex. TiO 2, WO 3, NaTaO 3, TaON, Wide band gaps SnO MO q/2 pseudo binary systems Sn(II) oxides Relatively narrow band gaps q M Known Compounds 4 Ti, Zr, Hf SnTiO 3, Sn 2 TiO 4 5 V,Nb, Ta SnNb 2 O 6, Sn 2 Nb 2 O 7, SnTa 2 O 6, Sn 2 Ta 2 O 7, SnTa 4 O 11 6 Cr, Mo, W SnWO 4, Sn 2 WO 5, Sn 3 WO 6 Only 10 oxides are known. Reported as good photocatalyst.

16 Inorganic Crystal Structure Database (ICSD) World largest database for known inorganic crystals. 177,000 crystal structures 40,000 structures excluding duplicates, incompletes, etc. 9,100 structure prototypes (e.g. rock-salt, perovskite,...) Number of chemical elements Number of structure prototypes in ICSD , , ,300 16

17 Crystal structures Prototypes of A(II) M(q+) oxides q Number of prototype structures in ICSD Examples 4 586 Pb 3 O 4, AlFeO 3, NdYbS 3, 5 207 K 2 ZrF 6, Bi 2 WO 6, 6

17 17 Crystal structures Prototypes of A(II) M(q+) oxides q Number of prototype structures in ICSD Examples Pb 3 O 4, AlFeO 3, NdYbS 3, K 2 ZrF 6, Bi 2 WO 6, CrUO 4, β SnWO 4, Total : 1,161 crystal structures SnO MO q/2 pseudo binary systems NdYbS 3 type NdYbS 3 NdYbS 3 type SnTiO 3 Total: 1,163x3 = 3,483 candidate oxides

18 18 Screening procedure 1. Candidates generation : DFT calculations 3,483 candidates 2. Phase stability Formation energy vs. SnO and MO q/2 / / 3. Band gap (E g ) by GGA(PBE) Criterion: E g > 2 ev 4. Band edge position (ε CBM and ε VBM ) by GGA(PBE) Criterion: ε CBM > H + /H 2 level and ε VBM < O 2 /H 2 O level

19 19 Phase stability different structures Energy Phase separation Convex hull A AB Composition B

20 20 Formation energy SnO WO 3 pseudo binary system Convex hull Included in ICSD SnO WO 3

21 21 Formation energy SnO MoO 3 pseudo binary system Convex hull as yet unknown SnO MoO 3

22 22 Formation energy Convex hull of SnO MO q/2 pseudo binary systems Reported oxides in ICSD (Red characters) are located on the convex hull. Band gap screening

23 23 Band gap Band gap of actual photocatalysts 2 ev (GGA) Band gap 0 ~ 1 ev 1 ~2 ev 2 ~ 3 ev over 3 ev SnO Ta 2 O 5 SnO WO 3 SnO MoO 3

24 Band edge position 24

25 25 Experimental results Synthesis of β SnMoO 4 Mixture of SnCl 2 and K 2 MoO 4 powders 1 hour annealing in Ar gas Washed and dried

26 26 Crystal structure of SnMoO 4 Newly discovered compound 498 K synthesized sample Space group type: P2 1 3 (Cubic) Lattice constant: a = 7.26 Å a c Sn b Mo O Trigonal prism which is characteristic of Sn(II)

27 27 Photocatalytic activity of SnMoO 4 Degradation of methylene blue under simulated daylight Newly discovered SnMoO 4 powder exhibits clear photocatalytic activity.

Calculations and Bayesian Optimization Atsuto Seko,

28 Discovery of new low thermal conductivity materials First-Principles Anharmonic Lattice-Dynamics Calculations and Bayesian Optimization Atsuto Seko, Atsushi Togo, Hiroyuki Hayashi, Koji Tsuda, Laurent Chaput, and IT 28

29 Find ultra-low thermal conductivity materials of 0.1W/m K level in Mat. Proj. database (55,000 crystals) diamond >2000 Thermal conductivity (W/m K) AlN Thermal conductivity 29

30 Background Thermal Conductivity (TC) κ = κ electonic + κ lattice Lattice Thermal Conductivity (LTC) Reliable experimental dataset is limited ( < 100 crystals). Reliable first principles calculations are very expensive. (CC5 class : 1 day/100 cores for 1 simple crystal.) Little knowledge to predict LTC deductively. (Simple rule to determine LTC is not clear.) Materials search has been made through modification of known compounds showing high/low LTC. 30

Cost for First Principles Calculations CC0 1min Single shot GGA

Lattice thermal conductivity Diffusivity by FPMD simulation

31 Cost for First Principles Calculations CC0 1min Single shot GGA CC1 Structure opt. Computational Cost with 1 core CC2 CC3 CC4 CC5 1hour 1day 1 week 1 year 2x2x2 supercell Single shot GGA Phonon frequency Lattice thermal conductivity Diffusivity by FPMD simulation 1000 atoms cell Single shot GGA Computational Cost with 100 cores CC5 CC6 CC7 1min 1hour 1day 1 week 1 year 10 nodes 31

32 Thermoelectric materials Essential for utilizing otherwise waste heat. Power Low T Electric current Heat flux Electric current n-type p-type Electric current High T Figure of Merit Seebeck coefficient : electrical conductivity : thermal conductivity 32

33 Physical origin of thermal resistivity Phonon-phonon scattering (phonon anharmonicity) Harmonic phonons do not interact. q phonon q'' q' phonon 33

34 phonopy: open source for ab-initio phonon calcs Phonon dispersion Entropy C V Free Energy Atsushi TOGO Kyoto Univ. Free Energy Volume Thermal Expansion Number of citations Web of Science 2016 August 34

35 First principles phonon database Structure Database e.g. ICSD, Atomwork,... Phonon Database An open database DFT Database e.g. Mat. Proj., OQMD, aflowlib phonopy First principles engine e.g. VASP, Quantum Espresso Abinit, etc... 35

36 Phonon Kyoto University d: phonon-dos t: thermal properties at constant volume g: mode-grüneisen parameters q: physical properties at 0GPa 36

37 Phonon Kyoto University

38 Ab-initio Lattice Thermal Conductivity (LTC) Atsushi TOGO

39 Lattice Thermal Conductivity First principles calculation vs. Experimental data Reliable calculations whose accuracy are comparable to experiments! 39

40 Virtual Screening 40

41 Candidates of descriptors for LTC Model 1: Volume V and Density ρ Model 2: Model 1 + primitive elemental descriptors H Li Be B C N O F LiH LiF BeO BN

42 Ranking of LTC for 54,779 compounds in MPD library Z score logκ The higher Z-score, the lower predicted LTC! Virtual screening 42

43 Top 10 lowest LTC compounds among 54,779 Virtual screening of 54,779 compounds in MPD library ab initio LTC 43

44 Lattice Thermal Conductivity (LTC) Newly discovered!! 44

45 Thermoelectric materials Essential for utilizing otherwise waste heat. Power Low T Electric current Heat flux Electric current n-type p-type Electric current High T Figure of Merit Seebeck coefficient : electrical conductivity : thermal conductivity 45

46 Top 10 lowest LTC compounds among 54,779 Virtual screening of 54,779 compounds in MPD library ab initio LTC 46

47 Newly discovered candidates for thermoelectrics with low LTC of < 0.5 W/mK K) and narrow band gap of < 1 ev. A. Seko, A. Togo, H. Hayashi, K. Tsuda, L. Chaput, and IT PRL (2015) 115,

48 Summary Materials discovery through first principles calculations 48

49 Fin

Fumiyasu; Tanaka, Isao The Authors. Published by WI

Fumiyasu; Tanaka, Isao The Authors. Published by WI Title Discovery of a Novel Sn(II)-Based O Daylight-Driven Photocatalysis Hayashi, Hiroyuki; Katayama, Shota; Author(s) Hinuma, Yoyo; Yokoyama, Tomoyasu; M Fumiyasu; Tanaka, Isao Citation Advanced Science