Energy Materials for a Low Carbon Future (Royal Society) Beyond Perovskites: Chemical Principles for Next-Generation Solar Energy Materials Prof. Aron Walsh Department of Materials, Imperial College London Materials Design Group: https://wmd-group.github.io @lonepair

Thin-Film Solar Cells Cu(In,Ga)Se2 CdTe From indirect (Si) to direct bandgap semiconductors for enhanced light absorption Light absorbing layer is a p-type semiconductor

From Materials to Devices Existing technologies benefited from decades of materials and device optimisation Complex chemical processes in CdTe solar cells: • Cu diffusion • CdCl2 annealing • Cd(S,Te) formation • (Cd,Zn)S formation http://www.nrel.gov/pv/cadmium-telluride-solar-cells.html Cu S,Te mixing Cd,Zn mixing

Lack of Structural Diversity Few materials have produced single-junction solar cells with >20% light-to-electricity conversion Face-centred cubic semiconductors: Si, GaAs, CdTe, CuInS2 Crystallographic unit cell of chalcopyrite [CuFeS2 mineral and Cu(In,Ga)(S,Se)2 ] Cu/In ions ordered in (201) planes All ions with tetrahedral coordination

Challenge for Emerging Solar Cells GaAs efficiency is close to the theoretical limit, while Si is cheap. New materials must perform! Metal oxides Cu2 O (6%) Bi2 FeCrO6 (8%) Co3 O4 (<1%) Metal sulphides SnS (5%) Cu2 ZnSnS4 (13%) FeS2 (3%) Metal halides CsSnI3 (10%) CH3 NH3 PbI3 (23%) BiI3 (1%) Examples of materials studied for thin-film photovoltaics (% sunlight to electricity conversion in photovoltaic devices) M. A. Green et al, Solar cell efficiency tables (version 52)

Talk Outline: Beyond Perovskites A. 21st Century Materials Modelling B. Metal Halide Perovskites C. Outlook for Materials Design

Chemistry on Computers “The underlying physical laws for the whole of chemistry are completely known” Dirac (1929) Boys et al, Nature 4544, 1207 (1956) EDSAC: Electronic delay storage automatic calculator o 3,000 vacuum tubes o Power consumption: 11 kW o Calculations on H2 , H2 O, BH radical

Supercomputers (1017 FLOPS) Top500.org ranking of public installations

Thousands of Interacting Electrons Density Functional Theory (DFT): Hohenberg & Kohn (1964); Kohn & Sham (1965) Source: F. Bechstedt – Many-body Approach to Electronic Excitations (2015)

First-Principles Modelling in 2018 Remove Approximations length and times scales electron-electron interactions electron-phonon interactions phonon-phonon interactions Accurate Solid-State Properties effective mass to carrier mobility phonon frequencies to lifetimes ground to excited states defects and disorder in crystals

Machine Learning Chemistry Statistical algorithms that learn from training data and build a model to make predictions 17k Machine Learning publications on Web of Science (08.18)

Machine Learning Chemistry Decision trees “Chemical robot”

Evolution of Materials Modelling CZTS Unstable CZTS Stable Quick-start machine learning guide: K. T. Butler et al, Nature 559, 547 (2018) Potential for combining first-principles predictions with experiments and databases for data-driven materials discovery

Share your Research Data https://github.com/WMD-group/hot-carrier-cooling Global movement of #openscience for transparency, reproducibility, and data-sharing

Open-Source Community PV Software http://www.solcore.solar; https://github.com/dalonsoa/solcore5

Talk Outline: Beyond Perovskites A. 21st Century Materials Modelling B. Metal Halide Perovskites C. Outlook for Materials Design

Hybrid Organic–Inorganic Perovskites Brief History (1958) – Photoconductivity in CsPbI3 (Møller) (1978) – Synthesis of CH3 NH3 PbI3 (Weber) (1994) – Metallic transition in CH3 NH3 SnI3 (Mitzi) (2009) – Perovskite dye cell (Miyasaka) (2012) – Planar thin-film solar cell (Snaith) Inorganic CsPbI3 Hybrid CH3 NH3 PbI3 or MAPI

Why Halide Perovskites? Essentials for Solar Cells • Strong optical absorption (Eg ~ 1.6 eV) • Light electron and hole masses (conductive) • Easy to synthesise (cheap and scalable) Advanced Features • Dielectric screening: carrier separation (weak excitons) and transport (low scattering rates) • Slow e-h recombination: low losses, large VOC o Polar nano-domains – carrier separation o Relativistic effects – spin-orbit coupling o Phonon scattering – limit non-radiative events

Perovskites: Model vs Reality Plastic crystal behaviour probed by Quasi-Elastic Neutron Scattering (P. Barnes, DOI: 10.1038/ncomms8124); 2D IR Spectroscopy (A. Bakulin, DOI: 10.1021/acs.jpclett.5b01555); Inelastic X-ray Scattering (S. Billinge, DOI: 10.1021/acsenergylett.6b00381) with simulations

Intersection of Hard and Soft Matter Jarvist Frost CH3 NH3 PbI3 3D periodic boundary (80–640 atoms) 25 fs per frame 0.5 fs timestep based on PBEsol forces at T=300K “As soft as jelly” J. M. Frost, K. T. Butler, A. Walsh, APL Mater. 2, 081506 (2014) Combination of density functional theory, GW theory, lattice dynamics, molecular dynamics, classical Monte Carlo, continuum device models

Static Dielectric Response Standard Inorganic Dielectric Organic-Inorganic Dielectrics Microstructure Conductivity Contacts Lattice dynamics Optical response Stat. mechanics Sum of: Key role in charge carrier generation and lifetimes

Static Dielectric Response High-frequency from QSGW+SOC Low-frequency from harmonic phonons (DFT/PBEsol) With Mark van Schilfgaarde

“Giant Dielectric Constant” JPCM 20, 191001 (2008) JPCL 5, 2390 (2014) J. M. Frost and A. Walsh, Acc. Chem. Res. 49, 528 (2016)

“Giant Dielectric Constant” JPCM 20, 191001 (2008) JPCL 5, 2390 (2014) J. M. Frost and A. Walsh, Acc. Chem. Res. 49, 528 (2016)

Bananas are Lossy Dielectrics J. F. Scott, J. Phys. Conden. Matter 20, 2 (2007) Note: the analogy is limited. Mixed ionic-electronic transport in perovskites gives rise to more complex Hebb-Wagner polarization (thanks Joachim Maier)

Electrons as Large Polarons Fröhlich electron-lattice interaction ! = 1 2 1 %& − 1 %( ) ℏ+ 2,∗+ ℏ . / GaAs = 0.1 CdTe = 0.3 CH3 NH3 PbI3 = 2.4 SrTiO3 = 3.8 Frost, Butler, Walsh, APL Mater. 2, 081506 (2014); Acc. Chem. Res. 49, 528 (2016)

Mixed Ion–Electron Conductors Current-voltage hysteresis [Snaith et al, JPCL (2014); Unger et al, EES (2014)] Rapid chemical conversion between halides [Pellet et al, CoM (2015); Eperon et al, MH (2015)] Photoinduced phase separation [Hoke et al, CS (2015); Yoon et al, ACS-EL (2016)] Electric field induced phase separation [Xiao et al, NatM (2015); Yuan et al, AEM (2016)] Photo-stimulated ionic conductivity [Yang et al, AChemie (2015); Kim et al, NatM (2018)]

Mixed Ion–Electron Conductors Nature Comm. 6, 8497 (2015); ACS Energy Lett. 3, 1983 (2018) Reservoir of charged point defects (site vacancies) in thermodynamic equilibrium: V- MA , V2- Pb , V+ I A. Walsh et al, Angewandte Chemie 54, 1791 (2015) Figure 3. Iodide ion vacancy migration from DFT calculations (a) Calculated migration Vacancy Ea (eV) I- 0.6 CH3 NH3 + 0.8 Pb2+ 2.3 D ~ 10-12cm2s-1 at T = 300 K [PBEsol/DFT in 768 atom supercell with nudged-elastic band]

Mean free path of each phonon Taming Ions in Halide Perovskites Phonon glass Phys. Rev. B 94, 220301 (2016) Stimuli that can activate ionic conductivity: electrical; chemical; optical; mechanical; thermal Walsh and Stranks, ACS Energy Lett. 3, 1983 (2018) Lattice thermal conductivity is ultra-low – do mobile ions transport heat?

Talk Outline: Beyond Perovskites A. 21st Century Materials Modelling B. Metal Halide Perovskites C. Outlook for Materials Design

Solar Absorber Shopping List • Low-cost and non-toxic elements • Direct optical bandgap (1–2 eV) • Easy to deposit and scale-up production • Semiconductor with low carrier concentrations • Tolerant to impurities and microstructure • Chemically stable at interfaces • Workfunction matched to electrical contacts

Solar Absorber Shopping List • Low-cost and non-toxic elements • Direct optical bandgap (1–2 eV) • Easy to deposit and scale-up production • Semiconductor with low carrier concentrations • Tolerant to impurities and microstructure • Chemically stable at interfaces • Workfunction matched to electrical contacts What can we reliably calculate using materials modelling?

Does it Absorb Sunlight? Simple descriptor: Eg . Advanced: optical absorption and detailed-balance for a thin-film SLME metric: Yu and Zunger, Phys. Rev. Lett. 108, 068701 (2012) Detailed balance: Blank et al, Phys. Rev. App. 8, 024032 (2017) Materials Chemistry Structure/Composition o Orbital character o Band widths o Band degeneracy o Selection rules

Is it Stable? Experiment (Scragg) Kesterite Unstable CZTS Stable Simple descriptor: ΔHf . Advanced: secondary phases under realistic growth conditions (ΔGf ) Materials Chemistry Structure/Composition o Choice of elements o Stoichiometry o Synthetic routes o Metastability Kesterites: Jackson and Walsh, J. Mat. Chem A 2, 7829 (2014) Perovskites: Zhang et al, Chin. Phys. Lett. 3, 036104 (2018) [arXiv 2014] Cu2 ZnSnS4 à Cu2 S + ZnS + SnS +S(g)

Does it Conduct (p–i–n)? CZTS Stable Simple descriptor: me * and ΔH(q,q’) for key defects. Advanced: self-consistent EF analysis Materials Chemistry Structure/Composition o Band energies o Growth conditions o (Co-)dopants o Solid-solutions Analysis of SnS: Y. Kumagai et al, Phys. Rev. Appl. 6, 014009 (2016) Defect review: J. Park et al, Nature Rev. Mater. 3, 194 (2018) Self-consistent defect cycle

Will Carriers Survive? Defect tolerance : Walsh and Zunger, Nature Mater. 16, 964 (2017) Defect review: J. Park et al, Nature Rev. Mater. 3, 194 (2018) Simple descriptor: ΔH(q,q’) point defect levels. Advanced: prediction of carrier capture rates Materials Chemistry Structure/Composition o Redox active ions o Lattice vibrations o Passivation o Post-processing

How to Extract Charge? CZTS Unstable CZTS Stable Simple descriptor: ɸ (workfunction). Advanced: structure matching and interfacial effects Materials Chemistry Structure/Composition o Atomic levels o Coordination o Reactivity o Interface layers QM/MM band energies: D. O. Scanlon et al, Nature Mater. 12, 798 (2013) ELS for perovskites: K. T. Butler et al, J. Mater. Chem. C 4, 1129 (2016) Electronic matching Lattice strain Site overlap for A find B with low Schottky barrier & small lattice mismatch & high atomic overlap

Progress Beyond Perovskites Ganose, Savory and Scanlon, Chem. Commun. 53, 20 (2017)

Conclusions: Beyond Perovskites • First-principles modelling is having impact on emerging photovoltaic technologies • Many simulation approaches required to describe length and time scales • Developments needed for quantitative predictions to support materials discovery Slides: https://speakerdeck.com/aronwalsh Thanks to past and present group members including Federico Brivio, Jarvist Frost, Jonathan Skelton, Lucy Whalley and collaboration network incl. P. Barnes, M. Toney, S. Stranks

Final Note: Chemist vs Machine Test your skills against a machine learning algorithm at predicting if a material is insulating Wolverton group: http://palestrina.northwestern.edu/metal-detection