Masters Presentation

Transcript

1. Imperial College London Screening earth-abundant materials for photovoltaics Emad Uddin Supervised by Dr Alex Ganose 03/06/2024
2. Imperial College London 2 Background: Why focus on photovoltaics (PV)? Why use earth-abundant elements? Methodology: Screening process Crystal structure prediction Results: Electronic properties of a candidate material Overview I am a solar panel made from 100% earth abundant elements!
3. Imperial College London Global warming 3 Mauna Loa, Hawaii Observatory
4. Imperial College London Solar energy availability 4 Solar Energy (720,000,000 TWh) Human Energy Consumption (160,000 TWh)
5. Imperial College London Principles of PV semiconductors 5
6. Imperial College London Limitation of current semiconductors 6 • Silicon: • Indirect band gap. • energy- intensive fabrication. • Gallium Arsenide: • Direct band gap. • More expensive to fabricate relative to Silicon. • Perovskite: • Direct band gap. • Contains toxic lead. • Degrades under environmental conditions. A. Polman, et al, Science 352, 307 (2016). DOI: 10.1126/science.aad4424.
7. Imperial College London Desired PV semiconductor properties § Direct band gap in the range 1.1 – 1.55 eV. § Low electron and hole effective masses. § High optical absorption. § Minimise defects (less radiative recombination). § Good band alignment with charge collection contacts in the device. § Composed of low-cost earth-abundant elements. 7
8. Imperial College London Scalable, non-toxic, earth abundant elements 9 Vesborg, P. et al . (2012). Addressing the terawatt challenge: scalability in the supply of chemical elements for renewable energy
9. Imperial College London Hypothesis 10 Can we find a stable earth-abundant chemical composition, so that its ground state crystal structure exhibits ideal solar cell properties? Steps: § Find all the chemically sensible compositions. § Thorough screening process to identify candidate compositions. § Predict the ground state crystal structure. § Determine the electronic properties to validate their suitability as PV semiconductors.
10. Imperial College London Finding neutral compositions § We have chosen 12 elements: Si, O, Mg, Al, K, Na, F, S, Cl, P, Ca, Ti. 𝑆𝑒𝑎𝑟𝑐ℎ 𝑠𝑝𝑎𝑐𝑒 = 𝑛! 𝑘! 𝑛! − 𝑘! (𝑟!) 11 § All compositions must be neutral:
11. Imperial College London Finding neutral compositions § We have chosen 12 elements: Si, O, Mg, Al, K, Na, F, S, Cl, P, Ca, Ti. 𝑆𝑒𝑎𝑟𝑐ℎ 𝑠𝑝𝑎𝑐𝑒 = 𝑛! 𝑘! 𝑛! − 𝑘! (𝑟!) 12 140,800 Initial size of the combinatorial search space
12. Imperial College London Finding neutral compositions 13 140,800 § All compositions must be neutral: 𝐴" 𝐵# 𝐶$ : 𝑥 𝑞% + 𝑦 𝑞& + 𝑧(𝑞' ) = 0 § Electronegativity must follow: 𝜒()*+,- > 𝜒-./)*+0. § Apply these constraints using SMACT. § Specific cations exhibit specific oxidation states in the presence of specific anions. SMACT filtered search space 13,633
13. Imperial College London Finding unique compositions § We use the Materials Project dataset to filter out known compositions. 14 § We use the Materials Project dataset to filter out known compositions. 140,800 SMACT filtered search space 13,633
14. Imperial College London Finding unique compositions § We use the Materials Project dataset to filter out known compositions. 15 § We use the Materials Project dataset to filter out known compositions. § Earth Mover’s Distance to measure chemical similarity 140,800 13,633 13,095 Compositions unrecorded in MP database
15. Imperial College London Finding unique compositions § Earth Mover’s Distance to measure chemical similarity 16 140,800 13,633 13,095 Compositions unrecorded in MP database Hargreaves, C.J.,(2020). Chemistry of Materials, 32(24), pp.10610–10620.
16. Imperial College London Compositional search space 18 140,800 13,633 13,095 Compositions ranked with probability and distance 1496
17. Imperial College London Property prediction 19 140,800 13,633 13,095 Compositions ranked with probability and distance 1496 § Crabnet uses a self-attention- based transformer to calculate composition properties.
18. Imperial College London Property prediction 20 140,800 13,633 13,095 Compositions with ideal predicted band gap 1496 § We use Crabnet to predict band gap without crystal structure. 765
19. Imperial College London Property prediction 21 140,800 13,633 13,095 Stable Compositions 1496 § We also predict thermodynamic stability 765 81
20. Imperial College London Finding the crystal structure 25 Pickard, C. J, “Ephemeral data derived potentials for random structure search.” PRB (2022).
21. Imperial College London Ab Initio Random Structure Searching (AIRSS) 26 Noble, E. (2021). Ab initio random structure searching empowers cathode material discovery for batteries. [online] FutureCat. Available at: https://futurecat.ac.uk/abinitiorandomstructuresearch/ [Accessed 30 May 2024]. P4 Cl2 O K4 P2 S3 K4 SiS4 P3 Cl2 F Na3 TiP3 S4
22. Imperial College London Summary of AIRSS results 27 Composition Number of relaxations P4 Cl2 O 955 K4 P2 S3 556 K4 SiS4 833 P3 Cl2 F 1585 Na3 TiP3 S4 1082
23. Imperial College London Electronic properties of P3 Cl2 F 28
24. Imperial College London Summary § Screened over 140,000 compositions implementing simple chemical theory, machine learning and statistical analysis. § Identified crystal structure for our candidate compositions. § Validated electronic properties through DFT. § The chosen compositions were not suitable for photovoltaic cells. 30
25. Imperial College London Future work Must incorporate further validation steps: § Calculate thermodynamic stability using DFT. § Calculate electronic properties using DFT with a hybrid exchange-correlation functional. § Calculate phonon dispersion for dynamic stability. § Explore defect properties to understand recombination. § Calculating spectroscopic limited maximum efficiency. 31
26. Imperial College London Acknowledgements § Dr Alex Ganose § The Virtual Atoms group including Ruiqi Wu, Leo Lou, Shirui Wang § Imperial HPC. 32