Computational Science for Energy Wanda Andreoni Centre de Calcul Atomique et Moleculaire (CECAM) Ecole Polytechnique Federale – Lausanne .

Computational Science for Energy

Wanda AndreoniCentre de Calcul Atomique et Moleculaire

(CECAM) Ecole Polytechnique Federale – Lausanne

www.cecam.org Trieste, May 31 2010

Computational Science: main domains of application

New powerful algorithms, better software (& hardware)

are needed… to help advancement of knowledge (basic science) to design new solutions: from materials & processes to device architectures to establish an intelligent management of power

generation and distribution systems to monitor/control/forecast “green” operations.

Outline

Materials science and chemistry Solar Energy Hydrogen and Water Novel batteries Nuclear CO2 capture and sequestration

Modeling and simulations : other applications

The Sun as Source of Energy

MODELLING & SIMULATIONS needed

(i) to improve on and design new materials;

(ii) to monitor, improve on and guide materials processing;

(iii) to optimize system-device integration & architecture;

(iv) to optimize performance of PV power supply systems (e.g. sizing).

The Science of Photovoltaics

Class Issues Status Needs (examples) of modeling for simulations

a-Si defects rich literature large-size long-time degradation

CIGS/CdTe defects, doping.. Ab initio improved algorithms structure growth physical insights HPC Organic fundamental mechs model calculations new algorithmsHybrid exciton & carriers and theories generation & migration degradation (polymer)

QD Multi Exciton Generation debate on mechanism ? Combination with experiment: crucial for model development.

a-Si based solar cells Advantages mature technology material is abundant low cost Disadvantages - less efficient than Xtal - degrades easily under

illumination (Staebler-Wronski effect)

a-Si:H

Why are experiments not sufficient? Hydrogen often eludes experiments; fast dynamics

What can simulations provide? Proofs of models; new possible scenarios; dynamics

What type of simulations? Classical molecular dynamics, electronic structure

calculations, and synergy (Car-Parrinello method and alike) Potentials? Sampling?

CdTe- and CIGS-based PV Advantages relatively cheap thin-film technology Disadvantages complex structure scarcity of core elements

IssuesDefects & impurities: generation,

diffusionNature of interfaces (dependence on

deposition method) Carrier transport also through

interfacesRole of Grain Boundaries

What can atomistic simulations do more?

• More accurate prediction of energy gaps, defect levels • Study of interfaces is lacking structure and composition inter-diffusion• Study of grain boundaries formation and role• Study of the effect of temperature & stress conditionsModels must be of relatively large sizes (at least 1000 atoms)Methods:Combination of classical MD and ab initio simulations Difficulty to obtain reliable interatomic potentials Efficient intelligent sampling of atomic configurations (REMD; MetaDynamics etc)Accurate and efficient algorithms for high-performance computing

PV Materials Processing

Modeling is complicated ; it may require multiscale (from atomistic to continuum) but also sophisticated optimization procedures.

Need for robust algorithms development (simulations and analysis)

System:Integration & Design

New design problems for PV require the combination of tools and methodologies from electronic and photonic technologies.

Maxwell equations & models of the electronic behavior (carrier generation, collection and transport) –

Technology-Computer-Aided-Design

Algorithm development required for integration of different methodologies for

hierarchical optimization (multi-parameter)

Photocatalysis I

for hydrogen production via water splitting(also for air and water purification; surface self-cleaning and self-sterilizing…)

Typical catalyst: TiO2

Challenges & need for simulations• Catalysts in the visible• Avoid modification of the “catalyst”• Avoid use of sacrificial reductants or oxidants (see Kohl et al. Science 324, 74 (09))

Photocatalysis IIDo we really understand what happens at the water/TiO2 interface?

K. Onda et al., Science 308, 1154 (05)“wet electrons”

Innovative Batteries

Li-air aprotic batteries

• Light, small, cheap• No self-discharge• Long-time storage

Oxygen through an air cathode: an “unlimited” cathode reactant !

Non-aqueous electrolyte avoids corrosion

Li/Air: Research Questions & Topics

Computational Models and Tools

Battery research combines the three most challenging aspects of computational physics: ** non-equilibrium, multiphase and multiscale (in space and in time) **=> A complete model may require 100’s of Petaflops (Exascale) computing.

Nuclear power: safety issues

ExamplesReactors : materials under extreme conditions; agingFuel cycle : recycling of minor actinidesNuclear waste : safe storage

Structural materials : Understanding interaction of dislocations with irradiation defects (e.g. the microstructure) is necessary to predict steel hardening under irradiation.

Fuels : Understanding the chemistry of actinides is vital to optimize actinide extraction and complexation

Reactor materials aging : Corrosion, fatigue, fracture…

MD simulation of radiation damage

Hierarchical multi-scale simulation of nuclear fuel

Atomistically-informedphase-field approachfor void nucleation andgrowth & fission-gasbehavior

Continuum mechanics,PDEs, constitutive laws

Atomic/electronic level (Newton’s laws) Radiation damage,micro-structural mechanismsand materials parameters

‘Mesoscale’ (viscous force laws)Effect of microstructural processes (fission gas, voids, cracks, diffusion, …) on thermo-mechanical properties

Materials science -engineering scalelinkage

Continuum level

CO2 : capture & sequestration (CCS)

Challenges (examples)

I. Find new solvents and additives for wet CO2 capture by scrubbing.

Amine absorption not amenable to large scale deployment in power plants e.g. high rate of degradation due to oxidation and salt formation; high energy penalty for amine regeneration.

II. Accelerate mineral carbonation for permanent CO2 fixation as carbonate.

Increase the reaction rate is crucial to obtain an industrial viable process. E.g. aqueous mineral carbonation: accelerate the rate of CO2 hydration and of silicate dissolution

CCS: a multi-scale multi-physics problem

Basic and general needs for C.S.

Higher accuracy Electron excitation spectrum Defect energies Rates of chemical reactions Rates for diffusion in complex systems… More realistic models of complex systems Multi-scale methodologies High Performance Computing often crucial ! Close collaboration with experimental research

Advanced modeling & simulations for …

future technologies of power generation & distribution (e.g. smart GRID)

Powerful and novel algorithms to optimize planning, to characterize behaviour & forecast

response (short- and long-term) under various scenarios (multiple temporal and spatial scales).

Better software and visualization capabilities to transform grid management to real-time automated state. New demands to technology will require the aid of computer-

aided design. Examples: for large-scale energy storage and low-loss transmission.

Designing and simulating a network so that it works in real time represents a grand computational challenge on an unprecedented scale.

PV-based Power Supply Systems

PV Stand-alone, Grid-connected or Hybrid Note: HPV includes other RE sources (typically wind, hydrogen, diesel)

Need: Optimize system engineering Modeling of single components Control and coordination System sizing Prediction of maximum-power point

Methods: Conventional approaches: empiric, analytic, numeric, statistical Innovative approaches: Artificial Intelligence methods (ANN, GA, FL…) ANN=artificial neural network GA=genetic algorithm FL=fuzzy logic

CECAM and C.S. for Energy

Our activities First workshop on “Critical materials

issues in inorganic photovoltaics” W.A., Claudia Felser,Tanja Shilling, June

2008 Brainstorm meeting on

“Computational Science for Energy ” W.A. and Claude Guet, Divonne, May 09

CECAM Workshops on ENERGY & ENVIRONMENT (2010)

2010 Materials modelling in nuclear energy environments: state of

the art and beyond M. Samaras, R. Stoller, R. Schaeublin, M. Bertolus, April 26-29 (Zurich) Gas separation & gas storage using porous materials L. Valenzano, C.O. Arean, C.M. Zicovich-Wilson, May 17-19

(Lausanne) Electronic-structure challenges in materials modeling for

energy applications N. Marzari and A. Rubio, June 1-4 (Lausanne) Ab initio electrochemistry M. Sprik and M. Koper , July 12-14 (Lausanne) Actinides: Correlated electrons and nuclear materials L. Petit, B. Amadon, S. Miller, June 14-16 (Manchester) Computational carbon capture B. Smit, S. Calero, T.J.H. Vlugt, July 26-28 (Lausanne) Simulations and Experiments on Materials for Hydrogen

Storage S. Meloni, S. Bonella, G. Schenter, October 11-14 (Dublin)

THANK YOU FOR YOUR ATTENTION

Knowledge advancement and design of new

solutions There is a strong need for advanced materials, novel processing

routes and innovative devices in the generation and exploitation of alternative energies. Control and design imply substantial progress in understanding. Simulations (computational materials science and chemistry) using accurate methodologies and HPC are often invoked as critical auxiliary tools to experiment.

Examples: increase lifetime of nuclear reactors; tailor materials properties for better performance, guide materials processing to lower cost & help system-level integration. New methods for carbon sequestration rely on understanding that only HPC

simulations can provide Use of bio-fuels relies on the understanding of bio-energy conversion

mechanisms (plant and microbial processes) for which HPC simulations are mandatory

Coupling climate and environmental modeling is a must to make a step forward.

Free Energy Diagram of Metal-Oxide catalyzed Recharge

G

Li2O2 (s)Li+ + LiO2

-

O2 + 2(Li+ + e)

- 2 e U0

- e U0

O2Li

I + (Li+ + e) M-O-M-O-

= U – U0Start

Goal: Oxygen Gas + Li Metal

Time

Computational Example 1 – Redox Reaction on Cathode

Realistic, ab-initio modeling of Oxygen Redox Reaction in aprotic environments the challenge is the reverse (recharge) reaction

Realistic, ab-initio modeling of Oxygen Redox Reaction in aqueous environments similar to fuel-cells, but Lithium replaces

Hydrogen

Purposes understand reaction kinetics and

rate limiting steps understand overvoltages and hence

energy efficiency design low-cost (metal-oxide) based

catalysts

Computational Approach various forms of (dynamic) Density

Functional Theory Ab-inito calculation of possible reaction pathway for the oxygen reduction reaction on a catalytic surface. By Manos Mavrikakis, U.Wisconsin et.al., performed at NCSA and SDSC TeraGrid systems. (Fuel Cell)

Computational Example 2 – Interfaces and Transport

Realistic modeling of electrolyte/electrode interfaces

Purpose Model the solvent and ion

transport mechanisms which is a very different problem

than the one posed in (1)

Computational Approach Molecular Mechanics Model of

electrolyte Quantum Mechanical Model of

electrode Establish a boundary region Probably limited to flat, 2D

geometries (w/current super- computers)

Need to invent ad-hoc methods to add 3D nano-morphology effectsCombined Quantum-mechanical and molecular

Mechanical Model of electrolyte/electrode interfaceModel by T.Jacob, Univ. Ulm

Experiments

MetaDynamics: A. Laio and F.L. Gervasio, Rep. Prog. Phys. 71 (2008)

Materials Science