The Jungle Universe About scales and physics in the cosmos Simon - PowerPoint PPT Presentation

The Jungle Universe About scales and physics in the cosmos Simon Portegies Zwart Sterrewacht Leiden

Observation of the early universe (WMAP)

Abel1689

Stephen's quintuplet

The universe is multi-physics

The universe is multi-scale

Jungle scales Size scale covers anythin from: ● 13.8 billion light years to km-size ● that covers 24 orders of magnitude ● 13.8 billion years to seconds ● that covers 18 orders of magnitude

j S = ν ν F = G m 1 m 2 ( ν k ) r 2 dI = − + ν I S ν ν τ d s ∂ P Gm = − D  F − ∇ p u ∂ π 4 m 4 r Dt = 2  ρ  ν ∇ u ∂ r 1 = ∂ π ρ 2   m 4 r ( P , T , Y ) i ∂ D u u   ( ) u ∂ = + ⋅ ∇ L u = ε + ε + ε ( P , T , Y ) ( P , T , Y ) ( P , T , Y ) ∂ Dt t ν nuc i i grav i ∂ m  ∂ T GmT ∇ u ⋅ = 0 = − ∇ ( P , T , Y ) i ∂ π 2 m 4 r P

Subrahanyan Chandrasekhar Sir Isaac Newton James Clark Maxwell George Gabriel Stokes Cloude-Louise Navier Sir Arthur Eddington

Prehistoric computational astrophysics Sumerian cuneform clay tablet dated around 1,200BC explaining the periodic behavior the planet Venus around 1,600BC (compute speed ~ 1 FLOP) Abacus (500BC, compute speed ~10FLOP)

”...'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question."

von Neuman & IAS 1960 2003 ~30 000 000 times faster Jun & GRAPE-4 500BC

Radiative transport Gravity Maxwell equations hydro-dynamics gas-dynamics Stellar evolution

LGM DAS-4

Computational challenges ● High performance (desktop) computing ● Distributed (wide area) computing ● Problem solving environments (software) ● Data acquisition ● Data mining ● Visualization ● Virtual collaboration

1908-2000 10mFlops

Software operated computers Manchester mark1 (1948, 550 FLOPs) Software by Tom Kilburn

The next generation problem solving environments ● Specialization (higher resolution) ● Optimization (high-performance) ● Diversification (wide range of applications) ● Hybridization (multi physics) ● Preservation (containment of existing codes)

The Astrophysical Multipurpose Software Environment AMUSE http://amusecode.org

Scientific research and development team ● Marco Spaans ● Steve McMillan ● Gijs Nelemans ● Paul Groot ● Vincent Icke ● Eline Tolstoy ● Onno Pols ● Evert Glebbeek ● Lex Kaper ● Rien vd Weijgeart ● Rob Knop ● John Fregeau ● Breanndan O Nuaillan

AMUSE - philosophy ● Build on community codes ● Standarized interfaces ● Automate as much as possible ● Builds on lessons learned from previous generations ● Core Team: – Inti Pelupessy (post-doc) – Arjen van Elteren (software engineer) – Marcel Marosvolgi, Nathan de Vries (programmers) – David Jansen (user support)

www.amusecode.org AMUSE - design Stellar Evolution Hydrodynamics Radiative Transfer Gravity AMUSE Combining existing codes INPUT OUTPUT With an extensive support framework To provide a generic framework For doing astrophysical experiments Compare models Unit handling Data conversion Initial conditions

AMUSE http://amusecode.org ● Layers having different Python Script roles Next Level ● Written in C/C++, Java Particles Units Fortran and Python Legacy Interfaces GD HD SE RT Message Channel MPI C/C++ code Fortran Code

Pelupessy etal in prep

User script Message passing script Message passing source Community code Process 1 Process 2 Send request evolve() Send request Send answer Confirm request evolve() Send request Evolve() done Send answer Confirm request Confirm request Confirm request

Two examples ● Formation of J1903+0327 ( ApJ in press: ArXive:1103-2275 ) – Gravitational dynamics + Stellar evolution ● Evolution of young star cluster ( to be submitted ) – Gravitational dynamics + Stellar evolution + Hydro dynamics

Simulating Embedded star clusters NGC3603 cluster By HST

Numerical ingredients ● Gravitational dynamics – Direct N-body integration (PhiGRAPE) – GPU or GRAPE equipped pc ● Stellar evolution – Henyey stellar evolution (MESA) – Beowulf computer cluster ● Gas dynamics – Smoothed particles hydrodynamics (Fi) – Super computer

Evolution of a gas rich star cluster SFE=0.05 f fb =0.1 SFE=0.50 f fb =0.01

AMUSE Today ● Automated referencing ● Unit conversion ● Online documentation ● Suite of examples ● Intricate module coupling via Hamiltonian splitting

Wish-list for AMUSE ● Runtime crash-recovery ● Self-consistent code restart ● Initial conditions repository ● Extensive data mining and analysis toolbox ● High-performance AMUSE ● AMUSE on the grid (PDRA Niels Drost VU) ● Asynchronous communication support ● Load balancing on heterogeneous architectures ● Data tunneling protocol