american astronomical society plenary talk

67
From First Stars to First Galaxies to the Reioniza3on of the Universe: 20 Years of Computa3onal Progress Michael L Norman Director, San Diego Supercomputer Center Dis8nguished Professor of Physics UC San Diego

Upload: michael-norman

Post on 15-Apr-2017

22 views

Category:

Science


0 download

TRANSCRIPT

Page 1: American Astronomical Society plenary talk

From  First  Stars  to  First  Galaxies  to  the  Reioniza3on  of  the  Universe:    

20  Years  of  Computa3onal  Progress  

Michael  L  Norman  Director,  San  Diego  Supercomputer  Center  

Dis8nguished  Professor  of  Physics  UC  San  Diego  

Page 2: American Astronomical Society plenary talk

What    we  have  explored  computa8onally  

Page 3: American Astronomical Society plenary talk

This  talk  is  about  the    first generation of galaxies aka primeval galaxies aka protogalaxies

•  How  they  formed?  •  When  they  formed?  •  What  they  were  like?  •  Why  are  they  important?  

Page 4: American Astronomical Society plenary talk

Don’t we have observational answers?

•  Yes  and  no  •  High-­‐z  HUDF  galaxies  are  the  8p  of  the  iceberg  

•  What  we  can  see  are  the  biggest  and  brightest  galaxies  of  their  era  

•  We  are  interested  in  what  came  before  

Page 5: American Astronomical Society plenary talk

Inferences from Reionization •  Observed  high-­‐z  galaxies  don’t  provide  all  the  ionizing  photons  needed  to  explain  reioniza8on  by  z=7  (Robertson+  2015)  

•  What  does  the  faint  end  of  the  LF  look  like?  

Bouwens+    2014  

Page 6: American Astronomical Society plenary talk

Protogalaxies: An artist’s impression

A.  Schaller,  STScI  

Yikes!  Can  we  possibly  simulate  that?!!  

Answer:  YES!  

Page 7: American Astronomical Society plenary talk

Peak  Speed:  13.4  Petaflops  Total  memory:  1.5  Petabyte  Processor  cores:  362,240  

NCSA,  University  of  Illinois  

Page 8: American Astronomical Society plenary talk

Bryan  et  al.  2014,  ApJS,  211,  19  hdp://enzo-­‐project.org  

Our  analysis/viz  tool:    

Turk  et  al.  2011,  ApJS,  192,  9  hdp://yt-­‐project.org  

Powerful software (open source), developed over 22 years

Our  simula8on  tool:    

Page 9: American Astronomical Society plenary talk

Protogalaxies: A big supercomputer simulation with lots of physics

•  How  they  form  •  When  they  form  •  What  they  are  like  •  Whether  they  contribute  the  missing  photons  to  reionize  the  universe  (spoiler:  YES)  

With  these  simula8ons  we  have  discovered  

Page 10: American Astronomical Society plenary talk

Plan for the talk

1.  Numerical  cosmology  101  2.  From  primordial  density  fluctua8ons  to  the  

first  stars  3.  From  the  first  stars  to  the  first  galaxies  4.  Simula8ng  the  first  galaxy  popula8on  5.  From  the  first  galaxies  to  cosmic  reioniza8on  

Page 11: American Astronomical Society plenary talk

NUMERICAL  COSMOLOGY  101  

Page 12: American Astronomical Society plenary talk

The universe at 380,000 yr ABB  ini)al  condi)ons  for  my  simula)ons  

Fluctua8ons  have  a  well-­‐measured  power  spectrum  

Comoving  volume  

Page 13: American Astronomical Society plenary talk

Computing the Universe:  Numerical Cosmology

•  Transformation to comoving coordinates x=r/a(t)

a(t1) a(t2) a(t3)

•  Triply-periodic boundary conditions

Input  observed  fluctua8ons  

Time  step  the  laws  of  physics  in  a  computer  program  

Cosmic  web  

Page 14: American Astronomical Society plenary talk

Equations of Cosmological Hydrodynamics

Mass  cons:  

Mom  cons:  

Energy  cons:  

Species  cons:  

2-­‐body  reac8ons   photo-­‐dissoc./ioniza8on  

Mul8species  gas  dynamics  

Newton’s  law:  

Poisson  eq.  

Dark  mader  dynamics  

Friedmann  eq.  for  scale  factor  a(t)  

Metric  

Page 15: American Astronomical Society plenary talk

“HALOS”    

Gravita8onally  bound  mixtures  of  baryons  and  

dark  mader    

Where  stars  and  galaxies  form  

Evolution of gas density

(c)  Brian  O'Shea  (MSU)  and  the  Enzo  Collabora8on,  2015  

Page 16: American Astronomical Society plenary talk

FROM  LINEAR  FLUCTUATIONS  TO  THE  FIRST  STARS  

Page 17: American Astronomical Society plenary talk

Forming the First Stars (Population III)

•  The  first  genera8on  of  stars  condense  from  pris8ne  H  and  He  gas  in  very  small  dark  mader  halos  beginning  about  100  Myr  amer  big  bang    (Abel,  Bryan  &  Norman  2000,  2002;  Bromm  et    al.  2002)  

•  Rota8onal  transi8ons  in  H2  molecule  is  the  dominant  radia8ve  cooling  mechanism  that  permits  the  gravita8onally  bound  cloud  to  condense  to  a  star  

Page 18: American Astronomical Society plenary talk

Key physics: gravitational instability

 

•  What  T  and  ρ  to  use?  •  Peebles  &  Dicke  (1968)  suggested    using  mean  values  at  recombina8on  

Jeans  mass  

MJ  =  106  Ms  

Globular  cluster  

Page 19: American Astronomical Society plenary talk

Key physics: catalytic formation of H2

•  Process  starts  with  residual  electrons  amer  recombina8on  

 

H + e- è H- + hνH- + H è H2 + e-

H + H+ è H2+ + hν

H2+ + H è H2 + H

Channel  1  

Channel  2  

Page 20: American Astronomical Society plenary talk

•  Solution: Semi-implicit rate solver (Anninos et al. 1997)

Computational difficulties and solutions

•  Difficulty:  noneq.  primordial  gas  chemistry  

 

Page 21: American Astronomical Society plenary talk

Computational difficulties and solutions

•  Difficulty:  Vast  range  of  spa8o-­‐temporal  scales  

 

Gas  cloud  

protostar  

Space:  1010  Time:  1012  

Page 22: American Astronomical Society plenary talk

Computational difficulties and solutions

•  Solu8on:  recursive  adap8ve  mesh  refinement  and  hierarchical  8me-­‐stepping  (Bryan  &  Norman  1997)    

   

Page 23: American Astronomical Society plenary talk

Evolution of grid refinements

(c)  Brian  O'Shea  (MSU)  and  the  Enzo  Collabora8on,  2015  

Page 24: American Astronomical Society plenary talk

L/Δxmin=1010

density

temperature

600 pc 60 pc 6 pc

self-gravitating core

Abel, Bryan & Norman 2002

100x mass of sun

Page 25: American Astronomical Society plenary talk

924  cita8ons  and  coun8ng  

Page 26: American Astronomical Society plenary talk

Findings  and  Implica8ons  Abel,  Bryan  &  Norman  (2002;  Science)  

•  First  stars  begin  forming  about  100  Myr  ABB  •  First  stars  are  massive:  >100  x  mass  of  Sun  •  First  stars  form  in  isola8on  (one  per  pregalac8c  clump)  •  They  will  be  extraordinarily  luminous  but  only  live  for  a  few  million  years  

•  They  will  explode  as  supernovae,  and  seed  the  universe  with  heavy  elements  (C,  N,  O,  Ca,  Si,  Fe…..)  

•  They  will  produce  first  stellar  mass  black  holes  •  èThere  should  be  no  first  stars  around  today  

Page 27: American Astronomical Society plenary talk

A deluge of papers followed

Page 28: American Astronomical Society plenary talk

FROM FIRST STARS TO THE FIRST GALAXIES

Page 29: American Astronomical Society plenary talk

Wait! You skipped a step

•  Precise  mass  of  final  star  is  currently  not  computable  •  But  we  can  es8mate  mass  from  amount  of  collapsing  gas  and  accre8on  rate  

•  We  have  good  Pop  III  stellar  models  which  give  their  life8mes,  luminosi8es,  and  fates  as  a  func8on  of  mass  (Heger  &  Woosley  2002,  Shaerer  2003)  

•  We  parameterize  our  ignorance  with  a  primordial  ini8al  mass  func8on  (PIMF)  

   

Page 30: American Astronomical Society plenary talk

Assembly of the first galaxies

•  First  galaxies  assemble  from  gravita8onal  merger  of  lower  mass  halos  that  previously  contained  first  stars  and  were  processed  by  their  radia8ve,  chemical,  and  kine8c  feedback  

•  A  typical  “first  galaxy”  will  incorporate  O(10-­‐100)  such  systems  

•  Need  to  simulate  a  larger  volume  with  “AMR  everywhere”,  tracking  the  detailed  star  forma8on  and  feedback  

Page 31: American Astronomical Society plenary talk

Assembly of the first galaxies •  Key  physics:  hierarchical  structure  forma8on  driven  by  dark  mader  clustering  

•  Key  computa8onal  physics:  subgrid  models  for  star  forma8on  and  feedback  (Pop  III  stars  and  metal-­‐enriched  star  clusters)  

Lacey  &  Cole  1993  

Page 32: American Astronomical Society plenary talk

Forming  a  Numerical  Star/Star  Cluster  Wise  &  Abel  2008;  Wise  et  al.  2012a  

Z>10-­‐4  

Metal  enriched  star  cluster  

Test  for  collapse  

Test  for  metallicity  

Pop  III  star  

PIMF   Salpeter  

Life8mes,  yields,  endpoints  

Time-­‐dependent  feedback  

yes  no  

Create  star  par3cle  

Feedback  &  Pop  III  remnants  

Page 33: American Astronomical Society plenary talk

Assembly of the first galaxies •  Key  physics:  transport  of  ionizing  and  photodissocia8ng  radia8on  from  young  massive  stars  

•  Key  computa8onal  physics:  adap8ve  ray  tracing  (Abel  &  Wandelt  2002;  Wise  &  Abel  2011)    

Page 34: American Astronomical Society plenary talk

Population III Star Formation Fireworks (Cox/Patterson, NCSA)

Page 35: American Astronomical Society plenary talk

Birth of a Galaxy Wise  et  al.  2012a,b;  2014      

(c)  John  Wise  (GIT)  and  the  Enzo  Collabora8on,  2012  

Page 36: American Astronomical Society plenary talk
Page 37: American Astronomical Society plenary talk

Galaxy  counts  (luminosity  fcn.)  

Volume  averaged  star  forma3on  history  

Page 38: American Astronomical Society plenary talk

STATISTICS  OF  THE  FIRST  GALAXIES:  THE  RENAISSANCE  SIMULATIONS  

Page 39: American Astronomical Society plenary talk

The Renaissance Simulations

40  cMpc  

Each  zoom-­‐in  region  is  200  8mes  the  volume  of  the  “First  Galaxy”  simula8on  but  is  simulated  at  nearly  equivalent  resolu8on  è  Massive  amounts  of  computer  power  required  (~50  M  cpu-­‐hrs)  

Page 40: American Astronomical Society plenary talk

Peak  Speed:  13.4  Petaflops  Total  memory:  1.5  Petabyte  Processor  cores:  362,240  

NCSA,  University  of  Illinois  

Page 41: American Astronomical Society plenary talk

A virtual tour of the Renaissance Simulation (Cox/Patterson, NCSA)

Page 42: American Astronomical Society plenary talk

How do the first galaxies form? Ans:  Mergers  and  Acquisi8ons  

z=25  

First  stars  

First  galaxies  

8me  

z=15  

Page 43: American Astronomical Society plenary talk

When did the first galaxies form? Ans:  Immediately  amer  the  first  stars  

Page 44: American Astronomical Society plenary talk

What are they like? Sta8s8cs  from  ~3000  halos  

Stellar  mass  v.  halo  mass  

UV  luminosity  v.  halo  mass  

O’Shea  et  al.  (2015)  

Page 45: American Astronomical Society plenary talk

How Many Are There? UV  Luminosity  Func8on  of  First  Galaxies  

 

Faintest  galaxy  Hubble  can  see  Faintest  galaxy  JWST  can  see  

O’Shea  et  al.  (2015)  

Brighter   Fainter  

Numerous  faint  galaxies  dominate  photon  budget  for  reioniza8on    

Page 46: American Astronomical Society plenary talk

FROM  FIRST  GALAXIES  TO  COSMIC  REIONIZATION  

Page 47: American Astronomical Society plenary talk

Escape  frac3on  vs.  halo  mass  

Frac3on  of  halos  ac3vely  forming  stars  vs.  halo  mass  

Xu  et  al.  (2016)  

Escape of ionizing photons

Page 48: American Astronomical Society plenary talk

A Calibrated Simulation of Reionization

Page 49: American Astronomical Society plenary talk

IGM completes reionization at z=7.3

zrei(100%)=7.3  

Planck  1σ

Page 50: American Astronomical Society plenary talk

A Calibrated Simulation of Reionization

•  Early  stages  driven  by  intermident  star  forma8on  in  smallest  galaxies  –  Leads  to  recombining  HII  regions  

•  Later  stages  driven  by  steady  star  forma8on  in  more  massive  galaxies  

•  Reioniza8on  starts  and  ends  consistent  with  observa8ons  

•   τes  agrees  with  Planck  data  within  error  bars  

Chen,  Norman  &  Xu  (in  prep)  

Page 51: American Astronomical Society plenary talk

Predic8ons  about  the    first generation of galaxies Ques3on   Predic3on  How  they  formed?   From  the  ashes  of  Pop  III  stars  When  they  formed?   As  early  as  z=25  (earlier  in  

rare  peaks  of  the  density  field)  

What  were  they  like?   Like  ultra-­‐faint  dwarfs  Why  were  they  important?   -­‐Galaxy  building  blocks  

-­‐Contributed  to  reioniza8on  -­‐May  be  ancestors  of  modern-­‐day  ultra-­‐faint  dwarfs  

Page 52: American Astronomical Society plenary talk

James  Webb  Space  Telescope  Launch  2018  

Page 53: American Astronomical Society plenary talk

Ongoing Investigations •  When  does  the  last  Pop  III  star  form  and  under  what  condi8ons?  –  Proximity  to  first  galaxies  (Xu  et  al.  2016)  

•  How  do  X-­‐rays  from  accre8ng  stellar  and  supermassive  black  holes  modify  this  picture?  –  Preheat/preionize  the  IGM  everywhere  (Xu  et  al.  2013,  2015)  

•  How  to  test  our  calibrated  reioniza8on  model?  –  21  cm  cosmology  (Ahn  et  al.  2015)  

•  How  to  connect  with  proper8es  of  Local  Group  dwarfs?  (Wise  et  al.  2014;  O’Shea  et  al.  in  prep)  

•  Observable  proper8es  for  JWST  (Barrow  &  Wise,  in  prep)  

Page 54: American Astronomical Society plenary talk

Acknowledgements to former students, postdocs, and

collaborators •  Tom  Abel  (Stanford)  •  Kyungin  Ahn  (Korea)  •  Marcello  Alvarez  (CITA)  •  Peter  Anninos  (LLNL)  •  Greg  Bryan  (Columbia)  •  James  Bordner  (UCSD)  •  Pengfei  Chen  (UCSD)  

•  Brian  O’Shea  (MSU)  •  Dan  Reynolds  (SMU)  •  Geoffrey  So  (Intel)  •  Bridon  Smith  (Edinburgh)  •  Mad  Turk  (UIUC)  •  John  Wise  (GA  Tech)  •  Hao  Xu  (UCSD)  

And  too  many  NSF  and  NASA  grants  to  men8on  

Page 55: American Astronomical Society plenary talk

RESERVE  SLIDES  

Page 56: American Astronomical Society plenary talk

Cosmological  Parameters  

Page 57: American Astronomical Society plenary talk

Mass in stars and remnants

Page 58: American Astronomical Society plenary talk

Star formation rate densities

Page 59: American Astronomical Society plenary talk
Page 60: American Astronomical Society plenary talk
Page 61: American Astronomical Society plenary talk
Page 62: American Astronomical Society plenary talk
Page 63: American Astronomical Society plenary talk

Renaissance  Simula)ons  Publica8ons  

Reference   Topic  

Xu  et  al.  (2013)   Pop  III  stars  and  remnants  

Xu  et  al.  (2014)   X-­‐ray  feedback  from  Pop  III  black  holes  

Chen  et  al.  (2014)   Scaling  rela8ons  for  SAMs  

Ahn  et  al.  (2015)   21  cm  signal  from  X-­‐ray  prehea8ng  

O’Shea  et  al.  (2015)   UV  luminosity  func8on  

Xu  et  al.  (2016a   Late  Pop  III  star  forma8on    

Xu  et  al.  (2016b,  submided)   Galaxy  proper8es  and  escape  frac8ons  

Xu  et  al.  (2016c,  in  prep)   X-­‐ray  background  from  Pop  III  stars  

Page 64: American Astronomical Society plenary talk

Renaissance  Simula8ons  Fact  Sheet  Configura3on  

L_periodic  (cMpc)   40  

L_refined  (cMpc)   6.6  

N_p  (effec8ve)   40963  

m_p  (solar  mass)   2.9  x  104  

AMR  levels     12  

Δx  min  (pc)   19/(1+z)  

z_init   99  

Physics  

Cosmology   WMAP7  

ICs   MUSIC    

Code   ENZO  

gas  dynamics   9-­‐species  primord.  2  metal  fields  

Chemistry/cooling   9-­‐species  noneq.  metal  line  

Radia8ve  transfer   EUV,  LW  

Lyman-­‐Werner  bkgd  

Yes  

Pop  III  SF+FB   Wise+  2012b  

Pop  II  SF+FB   Wise+  2012b  

Simula3ons  

Runs   7  

Core-­‐hrs   ~100  M  

Data  (TB)   ~70    

M.  Norman,  Aspen  EoR  2016  

Page 65: American Astronomical Society plenary talk

Star  Forma8on  Prescrip8ons  Wise  et  al.  (2012a,b;  2014)  

Pop  III   [Z/H]  <=  -­‐4  

Par8cle   Individual  Pop  III  star  

Mass   IMF  w/Mchar=40  Msol  

thresholds   fH2>5x10-­‐4,  δb>5x105,  div(V)<0  

Star  proper8es  

Schaerer  (2002)  

SN  yields   Heger  &  Woosley  (2003)  

Metal-­‐enriched   [Z/H]  >  -­‐4  

Par8cle   Molecular  cloud/star  cluster  

Mass   >  1000  Msol  

thresholds   Τ<1000Κ, δb>5x105  div(V)<0  

SF  efficiency   0.07  fcold  inside  MC  radius  

Radia8ve  FB   6000  γ/baryon  over  20  Myr  

SN  FB   6.8x1048  erg/s/Msol  amer  4  Myr  

Mass  recycling  &  enrichment  

Pop  III  IMF  

Page 66: American Astronomical Society plenary talk

FLD  versus  MORAY  

M.  Norman,  Aspen  EoR  2016   Norman  et  al.  (2015)  

Page 67: American Astronomical Society plenary talk

M.  Norman,  Aspen  EoR  2016  

FLD  versus  MORAY  

Norman  et  al.  (2015)