Visualizing High-Resolution Simulations of Galaxy Formation and Comparing to the Latest Observations from Hubble and Other Telescopes

Transcription

1 Third Annual SRL / ISSDM Research Symposium - UCSC Systems Oktoberfest October 18-19, 2011 Visualizing High-Resolution Simulations of Galaxy Formation and Comparing to the Latest Observations from Hubble and Other Telescopes Joel Primack Physics Department, UCSC Chair, UC Computer Committee Director, UC High-Performance AstroComputer Center Scientific Basis for Galaxy Formation Simulations ΛCDM Double Dark theory CDM-only N-body simulations Hydrodynamic galaxy formation simulations Making mages with stellar evolution and dust effects Comparing simulated galaxy images with observed ones I would like to collaborate on this challenging project, involving feature extraction and machine learning, with experts in computer vision.

2 ART hydro simulations Ceverino et al. 2010, 2011 Edge-on simulated z ~ 2 galaxy Visualizing High- Resolution Simulations of Galaxy Formation and Comparing to the Latest Edge-on Observations from Hubble and Other Telescopes observed z ~ 2 galaxies Edge-on simulated z ~ 2 galaxies I would like to collaborate on this challenging project, involving feature extraction and machine learning, with experts in computer vision. Face-on Joel Primack Physics Department, UCSC Chair, UC Computer Committee Director, UC High-Performance AstroComputer Center

3 Websites related to this talk: University of California High-Performance AstroComputing Center (UC-HiPACC) Bolshoi simulations CANDELS Hubble survey Sunrise dust code Many beautiful visualizations

4

5 Funding Opportunities Calls for proposals scheduled twice annually for Fall/Winter & Spring/Summer funding Cycles. UC-HIPACC will support focused working groups of UC scientists from multiple campuses to pursue joint projects in computational astrophysics and related areas by providing funds for travel and lodging. At the heart of UC-HIPACC are working groups. 1. Small travel grants enable scientists, graduate students, and post-doctoral students to travel easily and spontaneously between Center nodes. UC-HIPACC will fund travel grant proposals submitted by faculty members, senior scientists, postdocs or graduate students up to $1000 on a first-come-first-served basis with a simple application describing the plan and purpose of the travel. 2. Grants ranging between $ $5,000 to support larger working groups or participation in scientific meetings. 3. Mini Conference grants of up to $5,000 to support collaborations of multiple UC campuses and DOE labs. 4. Grants to faculty to support astrocomputing summer research projects by undergraduates. 5. Matching grants of up to $10,000 for astrocomputing equipment. 6. Innovative initiative proposals for other purposes that are consistent with the goals of UC- HIPACC. Such purposes could include meetings or workshops, software development, or education and outreach.

6 Astro-Computation Visualization and Outreach Project lead: Prof. Joel Primack, Director, UC High-Performance AstroComputing Center UC-HIPACC Visualization and Outreach Specialist: Nina McCurdy Pleiades Supercomputer NASA Ames California Academy of Sciences Adler Planetarium Chicago HIPACC is working with the Morrison Planetarium at the California Academy of Sciences (pictured here) to show how dark matter shapes the universe. We helped prepare their show LIFE: a Cosmic Story that opened in fall 2010, and also a major planetarium show that opened the new Adler Planetarium Grainger Sky Theater July 8, 2011.

7 Galaxy Merger Simulation Run on Columbia Supercomputer at NASA Ames Research Center. Dust simulated using the Sunrise code (Patrik Jonsson, UCSC/Harvard). Astronomical observations represent snapshots of particular moments in time; it is effectively the role of astrophysical simulations to produce movies that link these snapshots together into a coherent physical theory. Showing Galaxy Merger simulations in 3D will provide a deeper, more complete picture to the public and scientists alike.

8 Hubble Space Telescope Ultra Deep Field - ACS This picture is beautiful but misleading, since it only shows about 0.5% of the cosmic density. The other 99.5% of the universe is invisible.

9 Periodic Table stardust stars

10 COSMIC DENSITY PYRAMID ViewfromtheCenter.com New-Universe.org

11 Imagine that the entire universe is an ocean of dark energy. On that ocean sail billions of ghostly ships made of dark matter...

12 DARK MATTER + DARK ENERGY = DOUBLE DARK THEORY Technical Name: Lambda Cold Dark Matter (ΛCDM)

13 Big Bang Data Agrees with Double Dark Theory! 13

14 Distribution of Matter Also Agrees with Double Dark Theory! P(k) Max Tegmark

15 Because the ΛCDM Dark Energy + Cold Dark Matter (Double Dark) theory of structure formation is now so well confirmed by observations, it is crucial to study the predictions of this theory for the formation of dark matter structure in the universe and use this to improve our understanding of the visible objects that we can study with telescopes: galaxies, clusters, and large-scale structure.

16

17

18 z=49.0 t=49 Myr Expansion... z=12.0 t=374 Myr z=2.95 t=2.23 Gyr

19 t= 6.66 Gyr t= 13.7 Gyr (today) End of expansion for this halo Wild Space Tame Tame Space Space

20 dark matter simulation - expanding with the universe Text same simulation - not showing expansion 20 Andrey Kravtsov

21 CONSTRAINED LOCAL UNIVERSE SIMULATION

22 22

23 Virgo Cluster MWy & M31 Fornax Cluster

24 σ8 = 0.82 h = 0.70 Cosmological parameters are consistent with the latest observations Force and Mass Resolution are nearly an order of magnitude better than Millennium-I Force resolution is the same as Millennium-II, in a volume 16x larger Halo finding is complete to Vcirc > 50 km/s, using both BDM and ROCKSTAR halo finders Bolshoi and MultiDark halo catalogs will be released September 2011 at Astro Inst Potsdam and Stanford; Merger Trees will also soon be available

25 BOLSHOI SIMULATION FLY-THROUGH <10-3 of the Bolshoi Simulation Volume 100 million light years

26 Sloan Video Ends with sphere of CBR and two astronomers looking at it as thought they are on the outside

27 Simulations of Interacting Galaxies Including Dust The Antennae HST image of The Antennae

28 Sunrise Radiative Transfer Code For every simulation snapshot: SED calculation Adaptive grid construction Monte Carlo radiative transfer Polychromatic rays save CPU time Photons are emitted and scattered/ absorbed stochastically Patrik Jonsson

29 Spectral Energy Distribution w/o dust face on edge on Ultraviolet Infrared Visible Light Patrik Jonsson

30 Accelerating Dust Temperature Calculations with Graphics Processing Units Patrik Jonsson, Joel R. Primack New Astronomy 15, 509 (2010) (arxiv: ) When calculating the infrared spectral energy distributions (SEDs) of galaxies in radiation-transfer models, the calculation of dust grain temperatures is generally the most time-consuming part of the calculation. Because of its highly parallel nature, this calculation is perfectly suited for massively parallel general-purpose Graphics Processing Units (GPUs). This paper presents an implementation of the calculation of dust grain equilibrium temperatures on GPUs in the Monte-Carlo radiation transfer code Sunrise, using the CUDA API. The Nvidia Tesla GPU can perform this calculation 55 times faster than the 8 CPU cores, showing great potential for accelerating calculations of galaxy SEDs.

31 Galaxy Merger Simulation run on the Columbia Supercomputer This image and the following videos show a merger between two Sbc galaxies, each simulated with 1.7 million particles. The images are realistic color composites of u, r, and z-band images. Galaxy mergers like this one trigger gigantic starbursts in which millions of stars form. But dust absorbs about 90% of the light, and reradiates the energy in the far infrared. We calculate this radiative transfer using ~10 6 light rays per image. The simulation was run by Greg Novak, and the visualization is by Patrik Jonsson.

32 Galaxy Merger Simulation Patrik Jonsson, Greg Novak, Joel Primack music by Nancy Abrams

33 Gas inflows to massive halos along DM filaments 320 kpc RAMSES simulation by Romain Teyssier on Mare Nostrum supercomputer, Barcelona Dekel et al. Nature 2009

34 Galaxy Hydrodynamic Simulations Physical phenomena included in ART simulations analyzed thus far: Dark matter and baryons, metal and molecular hydrogen cooling to ~100K, UV background with self-shielding, star formation, energy input from stellar winds and supernovae, advection of metals. See Ceverino & Klypin 2009; Dekel, Sari, & Ceverino 2009; Ceverino, Dekel, & Bournaud 2010; Goerdt, Dekel, Sternberg, Ceverino, Teyssier, & Primack 2010; Fumagalli, Prochaska, Kasen, Dekel, Ceverino, & Primack 2011; Ceverino, Dekel, Mandelker, Bournaud, Burkert, Genzel, & Primack 2011; Kasen et al and other papers in preparation. Additional phenomena included in ART simulations currently running: Radiative feedback from luminous stars and from AGN.

35 Stars

36

37 simulated z ~ 2 galaxy Bassi computer, NERSC ART hydro sims. Ceverino et al Face-on Edge-on now running on NAS Schirra computer simulated z ~ 2 galaxy observed z ~ 2 galaxies simulated z ~ 2 galaxies Ly alpha blobs from same simulation Goerdt et al. 2010

38

39 CANDELS Simulation edge-on CANDELS Simulation face-on w/ Dust w/o Dust w/ Dust w/o Dust Simulation shown is MW3 at z=2.33 imaged to match the CANDELS observations in ACS-Vband and WFC3-Hband Pixel scale - convolved with simulated psfs - noise and background derived from ERS observations (same field as examples shown) MW3 was imaged at face-on and edge-on viewing angles both with and without including dust models Mark Mozena

40 Sunrise Visualizations UCSC grad student Chris Moody (working with Primack, Matt Turk, and Patrik Jonsson) has created a pipeline to process ART simulation outputs efficiently to create multiwavelength Sunrise images. UCSC grad student Mark Mozena (working with Faber, Dekel, Koo, Lotz, and Primack) has perfected methods to convolve with appropriate PSFs and add noise to these Sunrise images, so that they can be compared directly with observations. We know that Sunrise with standard dust assumptions matches SEDs of nearby galaxies. We are runnning a number of dust models on hydro simulations of z ~2 to 3 galaxies to see what dust models will best agree with observed SEDs for similar galaxies. We plan to generate simulated images in ~5 wavebands of ~10 orientations of ~10 timesteps of the ~20 simulated galaxies that we have now (leading to about 10 4 images). We plan to expand this by an order of magnitude over the next year or so (producing ~10 5 images). We are also looking into machine classification of real and simulated galaxy images, initially as a senior thesis project with UCSC astrophysics major Andrew Breslin. We would like to collaborate with experts on Computer Vision and Feature Extraction.

41 Websites related to this talk: University of California High-Performance AstroComputing Center (UC-HiPACC) Bolshoi simulations CANDELS Hubble survey Sunrise dust code Many beautiful visualizations