Computational Astrophysics: Research to Teaching and Beyond Adam Frank University of Rochester

A Cast of Many AstroBEAR MHD / Clumpy Flows Sorin Mitran UNC *Andrew Cunningham (UR, UC Berkeley, LLNL) Alexei Poludnenko (UR, NRL) Kris Yirak (UR Grad Student) Jonathan Carroll (UR Grad Student) Thanks to: NSF, DOE, NASA, UR Laboratory for Laser Energetics

Boundary conditions Initial conditions Simulations: Numerical Experiments with Diff. Eq & PDEs space time

3 rd Age of Simulations: 1 st Age: 1-D calculations with computers 1 st Age: 1-D calculations with computers –CDC 7600, etc 2 nd Age: Moderate rez 2-D calculations 2 nd Age: Moderate rez 2-D calculations –Supercomputers: Cray, etc 3 rd Age: High rez Multi-D, Multi-physics calc. 3 rd Age: High rez Multi-D, Multi-physics calc. –Grid Computing, Massively Parallel, Clusters We are talking about “Virtual Reality” on a scientific level. We are talking about “Virtual Reality” on a scientific level. Sim. data sets now as “rich” as real data sets Sim. data sets now as “rich” as real data sets –Petabytes/flops = one quadrillion bytes. Not just simulation but “Cyberscience” Not just simulation but “Cyberscience” –Integrate IT on all levels of science practice

History: Research Simulation Shocked Clumps Astrophysical environments very heterogeneous. Winds, Blastwaves & ISM are all “clumpy” Interaction of 1-clump with passing wind is a well studied problem. Critical Parameter t cc : Cloud Crushing Time Woodward 1976

Frontiers of Algorithm Development Adaptive Mesh Refinement – code automatically places grid cells where needed. Adaptive Mesh Refinement – code automatically places grid cells where needed. Multi-physics – code simulates many physical processes simultaneously Multi-physics – code simulates many physical processes simultaneously –Magnetic Fields, gravity, radiation transport, chemistry, ionization dynamics.

UR Computational Group: AstroBEAR Multiyear, Multi grad-student effort Began 2002 Now on 4 th generation of student

Simulation in Age of CyberScience How to maintain 10 6 line code base across N years M revisions K Student Generations Static documentation doesn’t work. Our solution: Wiki, self- compiling documentation

Adaptive Mesh Refinement AMR R. Deiterding Different AMR Methods: Different AMR Methods: Grid based regridding Cell based regridding University of Rochester Code University of Rochester Code AstroBEAR: grid Hierarchy of Grids Hierarchy of Grids Require Prolongation/Restriction Operators Require Prolongation/Restriction Operators Carry data from one grid level to another Prolongation (Corse to Fine) Prolongation (Corse to Fine) Restriction (Fine to Corse) Restriction (Fine to Corse)

Mach 10 radiatively cooled bullet AMR grid generation in the system Pre-Planetary Nebula: CRL 618 “Explosion” from dying Solar-type star

AMR MHD Div B = 0 Hydro: Need conservative prolongation/ restriction operators. Hydro: Need conservative prolongation/ restriction operators. MHD: Maintain solenoidal condition. Need divergence free operators on “staggered mesh”. MHD: Maintain solenoidal condition. Need divergence free operators on “staggered mesh”. Cunningham et al 2007

Results I MHD Shocked Clump “Standard” Test M = 10  = n c /n w =   =4  = 4

Many Clumps: Radiative MHD Shocks in Heterogeneous Media Cunningham et al 2007 M = 10  = n c /n w =   =10  = 10

How AMR Changes Game: Resolution and Convergence Convergence formally defined as approach to known analytic solution. For complex non-linear problem these rarely exist Define convergence as change relative to highest resolution simulation possible. For adiabatic shock clump convergence appears at N = 120 cells/R cFor adiabatic shock clump convergence appears at N = 120 cells/R c

Resolution and Convergence Radiative Clumps Radiative cooling allows post shock flows to collapse – but how far? AMR allows us to run highest resolution radiative clump simulations to date: N = 1500/R c Highest resolution radiative clump simulations to date: N ~ 1500/R c

Resolution and Convergence Radiative Clumps Radiative cooling allows post shock flows to collapse – but how far? Increases in resolution show qualitatively new behaviors as  x < crit

Resolution and Convergence Radiative Clumps What measures, metrics can we trust at given resolution? ester.edu/~yirak/for adam/vorticities.png

Radiative MHD Clumps with Self-consistent Fields

How many MHD-AMR Codes  Not many  AstroBEAR  Flash  ENZO  Orion  AMR VAC  Athena

Computation in Teaching My adventure in E-ed-biz 2000 NSF Career Award AstroFlow – Simulation outreach tool Planetarium asks to buy copy (?!?) Create Truth-N-Beauty LLC with UR E-education digital media company Produce simulation based modules for: McGraw-Hill, Prentice Hall etc

Second Avenue Software The e-biz goes it alone 2006 Truth-N-Beauty becomes 2nd Ave

What we built Celestial Sphere Phases of Moon Seasons Solar System Builder Planetary Atmospheres

Going Beyond Teaching Outreach: Use partners in new media to create interactives for websites. Discover/Astronomy/SciAmerican “Serious Games” – twitch games with a science theme.

Going Beyond Teaching

Star Formation

Conclusions Computation Computation –Advanced AMR/multi-physics codes allow new era of simulation –“Weather vs. Climate”: what to do with Petaflops/bytes Teaching by Simulation Teaching by Simulation –New opportunities if done well (graphics, pedagogy) –New opportunities outside of classroom.