Simulation of Optical Waves

Transcription

1 Simulation of Optical Waves K. Hertel, Prof. Dr. Ch. Pflaum 1 Department Informatik Friedrich-Alexander-Universität Erlangen-Nürnberg 2 School for Advanced Optical Technologies Friedrich-Alexander-Universität Erlangen-Nürnberg KONWIHR Berichtsworkshop / 19

2 1 Applications 2 Problem domain 3 Solver 4 Implementation and Optimization 5 Future 2 / 19

3 The Solar Panel Industry Classical photovoltaics are increasingly supplanted by modern thin-film technologies: Reduction of material and energy in production Printable photovoltaic panels scale up easily to mass production Flexible substrates allow for previously unseen deployment scenarios Integration into clothing and building facades 3 / 19

4 Current R&D Efforts Challenges are Competitiveness compared to classical panel technologies in terms of quantum efficiency 4 / 19

5 Current R&D Efforts Challenges are High production cost for organic solar cells despite decreased material usage 5 / 19

6 New Projects: ECN (starting 2012) Simulations for analyzing up-conversion and light-trapping properties of thin-film solar cells in the framework of the Energie Campus Nürnberg Brings together work groups in the field of renewable energy research The Solarfabrik project within this framework is targeted at improving printable photovoltaics Its objective is the reduction of production cost for solar panels based on organic and inorganic thin-film technologies Friedrich-Alexander-Universität (FAU), Georg-Simon-Ohm-Hochschule Nürnberg, Fraunhofer Institute für integrierte Schaltungen (IIS) und für integrierte Systeme und Bauelementtechnologie (IISB), Bayerisches Zentrum für angewandte Energieforschung (ZAE) 6 / 19

7 Industrial Project: LIST (BMU, since ) Simulations targeted at optimizing structures in terms of their light-trapping properties in thin-film solar cells Primary objective is an analysis of the causes that lead to discrepancies in TCO quality in lab vs high-volume production environments Evaluation of the diffusive properties of surfaces and their impact on the quantum efficiency of thin-film solar cells Forschungszentrum Jülich, Fraunhofer-Institut für Silicatforschung (ISC) und für Schicht- und Oberflächentechnik (IST), Malibu, Laserzentrum Hannover (LZH), Euroglas and other glas and panel production companies 7 / 19

8 Industrial Project: SiSoFlex (BMU, BMBF, from ) Couples simulation of electrical and optical properties of solar cells Evaluates new cell concepts based on aluminum Aims at developing silicon based solar cells on top of flexible substrates Bosch, Next Energy, Alanod, Fraunhofer Institut für photonische Mikrosysteme (IPMS) and their glas and panel production sub-contractors 8 / 19

9 Geometric setting Box shaped domain consisting of layers of different materials Surface properties are essential in understanding resulting quantum efficiencies Modeling of the surfaces based on synthetic or experimentally scanned structures of produced cells (typical structure sizes range from 1 10µm) 9 / 19

10 Geometric setting Nano particals and nano wires made of silver can vastly improve the diffusion of incoming light Leads to better light-trapping and absorption properties of the resulting cells Electro-magnetic properties of the materials depend highly on the respective frequency within the spectrum of interest ( nm) 10 / 19

11 Discretization Orthogonally structured Yee cells Aligned as a set of structured staggered grids Hold the degrees of freedom of the discretized components of the elctro-magnetic field (and coefficients determined from physical properties of the materials involved) EM field is discretized with degrees of freedom per local wave length within the media (leading to a mesh size of about 5nm) 11 / 19

12 Computational Effort Domain with many complex-valued degrees of freedom in the orders of magnitude of per component of the EM field e.g. 5µm 5nm 5µm 5nm 2.5µm 5nm = Restriction to numerically simple explicit schemes in order to gain results in a reasonable amount of time High demand of memory resources the example above takes about 300GiB of main memory Partitioning the domain for distributed computing is simple due to the structured grids involved 12 / 19

13 Iterative Scheme Frequency dependent material properties and seeking a time harmonic solution suggest solving the problem in frequency space Explicit iteration method based on the Finite Difference Frequency Domain method solving Maxwell s equations Adaptations for material properties of silver (THIIM for negative permittivity) Absorbing boundary conditions (Perfectly Matched Layer): Split field approach 13 / 19

14 Simulation Code Development of an efficient, object oriented, highly parallel simulation code High maintainability by using expression templates Automatic load balancing in order to make better use of available compute resources taking into account varying solver times over the simulated spectrum Solar Cell Test Cases Problem Maxwell Artificial Structures AFM Scans others Expression Template Library Operators Data Decomposition Communication Discretization Domain 14 / 19

15 Scaling and Performance Work conducted within the KONWIHR project yields the following improvements: Reduction of memory consumption and size of the computational domain while keeping the size of the simulated physical domain Clever use of boundary conditions (phase shift, boundary symmetries for enforcing periodicity) Targeted optimization of performance critical sections Strong and weak scaling results on a cluster based on the Nehalem microarchitecture 15 / 19

16 Scaling and Performance Work conducted within the KONWIHR project yields the following improvements: Code is bandwidth limited due to the simple scheme and high amount of state variables involved in solving Maxwell s equations Excellent scaling behavior in accordance with expectations for stencil based codes Strong and weak scaling results on a cluster based on the Nehalem microarchitecture 16 / 19

17 Time dependent problems Application for a DFG project: Penetration of solar cells with laser pulses Developments of a measurement tool set for evaluating the light trapping properties of thin-film solar cells qualitatively Behavior of solar cell structures in the presence of time dependent light pulses Existing Laser simulation code with Gaussian mode analysis Universität Bielefeld and Malibu 17 / 19

18 Thank you! Thanks very much for your attention! 18 / 19

19 References For main references confer KONWIHR-II-Abschlußbericht Pictures courtesy of: FVEE ( BINE Informationsdienst ( DeutscheSolar ( MIT ( Yee cell by Steven G. Johnson, licensed under the GFDL 19 / 19