1. Does The Universe Have A Metal Floor? Matthew Pieri The Ohio State University, 1st November, 2007 Collaborators: Hugo Martel, Joop Schaye, Anthony Aguirre, Martin Haehnelt, Cédric Grenon, Steeve Pinsonneault

3. Big question: How Widespread is Metal Enrichment?  Why care?  Formation of galaxies  Formation of Population III stars  Properties of the Intergalactic Medium Big question: How Widespread is Metal Enrichment?  Why care?  Formation of galaxies  Formation of Population III stars  Properties of the Intergalactic Medium

4. OutlineOutline Background: The Intergalactic Medium Observations of the extent of enrichment Model for galactic outflows Background: The Intergalactic Medium Observations of the extent of enrichment Model for galactic outflows

5. To Earth Heavy element absorption Emission lines from the Quasar Lyman limit Hydrogen absorption due to galaxy Quasar Observed wavelength (Å) Lyman alpha forest DLA Illustration courtesy of John Webb

6.  Filaments, pancakes & voids of structure that trace dark matter  Mostly moderately overdense, although can be  Mostly photoionized hydrogen since reionization epoch  neutral to 1 part in 10 4  at  From photoionization heating and adiabatic cooling   No in situ metal enrichment  Hence need for transport mechanism  OVI and CIV most prominent metal species  Filaments, pancakes & voids of structure that trace dark matter  Mostly moderately overdense, although can be  Mostly photoionized hydrogen since reionization epoch  neutral to 1 part in 10 4  at  From photoionization heating and adiabatic cooling   No in situ metal enrichment  Hence need for transport mechanism  OVI and CIV most prominent metal species Intergalactic Medium at High-z

7. Galaxy Formation and Evolution of the IGM Galaxy formation Feedback Gravitational instability Energy dissipation Radiative feedback Outflows IGM Galaxies Heating(Re)ionization Heating Collisional Ionization Kinetic energy Metal deposition

8. Where To Look For Widespread Enrichment  Low density regions (around mean density or lower)  Ionization species which are dominant there  Regions far from galaxies  Locations required  Low density regions (around mean density or lower)  Ionization species which are dominant there  Regions far from galaxies  Locations required Line of sight to Quasar Metal Outflow Renyue Cen

9. Which Ionization Species to Look For  Various ionization species detected  OVI  best tracer in the lowest density systems (voids)  CIV  useful for moderate density systems (filamentary structures)  Various ionization species detected  OVI  best tracer in the lowest density systems (voids)  CIV  useful for moderate density systems (filamentary structures) Rauch, Haehnelt & Steinmetz (1997 )

10. OutlineOutline Background: The Intergalactic Medium Observations of the extent of enrichment Model for galactic outflows Background: The Intergalactic Medium Observations of the extent of enrichment Model for galactic outflows

11. The Spectra I A synthetic Ly  forest in a QSO spectrum  1D Gaussian Random field  Density power spectrum filtered for pressure effects  PDF mapped to a lognormal - mimic non-linear structure  Power law equation of state  Noise, instrumental broadening and bulk absorption same as observed A synthetic Ly  forest in a QSO spectrum  1D Gaussian Random field  Density power spectrum filtered for pressure effects  PDF mapped to a lognormal - mimic non-linear structure  Power law equation of state  Noise, instrumental broadening and bulk absorption same as observed /Ang

12. The Spectra II … with an OVI forest OVI (1032,1038 ) an excellent tracer of metals in voids BUT is found in the Ly  Forest in QSO spectra … with an OVI forest OVI (1032,1038 ) an excellent tracer of metals in voids BUT is found in the Ly  Forest in QSO spectra /Ang

13. A Ly forest – and an OVI forest A Ly  forest – and an OVI forest  Binned byand equivalent apparentpixels  Binned by  Ly  and equivalent apparent  OVI pixels  Median optical depths taken  Various techniques for minimising contamination of OVI signal zpzp zpzp Pixel by Pixel Search /Ang

14. A Ly forest – and an OVI forest A Ly  forest – and an OVI forest  Binned byand equivalent apparentpixels  Binned by  Ly  and equivalent apparent  OVI pixels  Median optical depths taken  Various techniques for minimising contamination of OVI signal  Sensitive to weak absorption throughout the spectrum zpzp zpzp Pixel by Pixel Search /Ang

15. Overall Affect on Search

16. Real and Synthetic Spectra with best Spectra Used  Q1122-165  z=2.0-2.3  Q1442+293  z=2.5-2.6  Q1107+485  z=2.7-3.0  Q1422+231  z=3.2-3.5 Spectra Used  Q1122-165  z=2.0-2.3  Q1442+293  z=2.5-2.6  Q1107+485  z=2.7-3.0  Q1422+231  z=3.2-3.5 MP & Haehnelt (2004)

17. Statistical Significance of OVI Detection   2 test of agreement between simulation and observation  in three of the four QSOs results consistent with  for Q1422   2 test of agreement between simulation and observation  in three of the four QSOs results consistent with  for Q1422 MP & Haehnelt (2004)

18. Statistical Significance of Low Density Detection  No OVI below (95% confidence)  Limit mainly contamination by Lyman lines  No OVI below (95% confidence)  Limit mainly contamination by Lyman lines MP & Haehnelt (2004)

19. Volume Filling Factor of Metals  Fraction of universe filled by overdensities and above and above  Provides fraction of the universe that is metal enriched  Volume filling factor 4% or more  Fraction of universe filled by overdensities and above and above  Provides fraction of the universe that is metal enriched  Volume filling factor 4% or more MP & Haehnelt (2004)

20. Can We Do Better With CIV?  No Ly α forest but lower τ  New limit - noise and continuum fitting  No Ly α forest but lower τ  New limit - noise and continuum fitting Spectra Used  PKS2126-158  z=2.6-3.2  Q1422+231  z=2.9-3.5  Q0055-269  Z=2.9-3.5 Spectra Used  PKS2126-158  z=2.6-3.2  Q1422+231  z=2.9-3.5  Q0055-269  Z=2.9-3.5

21. CIV Statistical Significance  We find  No detection for (95% confidence)  Volume Filling Factor 1% or more  Schaye et al. (2003)  CIV down to  With large scatter  VFF still poorly constrained

22. Nearby Lyman Break Galaxies and the IGM  Adelberger et al. (2003, 2005)  Find strong CIV and close LBG are same systems  X-correlation of log(N CIV )>12.5 and LBGs similar to LBG autocorrelation  Claim all enrichment from superwinds at z ~ 3

23. What is the spatial distribution of metals that are seen with the pixel search? Two Samples of Pixels Line Of Sight Marker pixels in LOS Pixels in LOS within  km/s of marker pixels LBG MP, Schaye & Aguirre (2006)  Markers provided by  LBGs  Strong CIV absorption

24. Strong CIV Absorption as a proxy for Galaxies  8% of pixels in “near” sample  Nearest 30km/s discarded  Clear signal of excess enrichment close to strong CIV  Most enrichment detected far from strong CIV  Scatter in sub-samples lower but not low enough  8% of pixels in “near” sample  Nearest 30km/s discarded  Clear signal of excess enrichment close to strong CIV  Most enrichment detected far from strong CIV  Scatter in sub-samples lower but not low enough MP, Schaye & Aguirre (2006)

25. Summary of Observations  Consistent with full enrichment down to  Partial enrichment down to  Volume filling factor > 4%  Large scatter in the metallicity  Increase in enrichment near galaxies  BUT regions far from known galaxies (and most of metals by volume) still show enrichment  Inclusion of known galaxy location lowers scatter  Consistent with full enrichment down to  Partial enrichment down to  Volume filling factor > 4%  Large scatter in the metallicity  Increase in enrichment near galaxies  BUT regions far from known galaxies (and most of metals by volume) still show enrichment  Inclusion of known galaxy location lowers scatter

26. OutlineOutline Background: The Intergalactic Medium Observations of the extent of enrichment Model for galactic outflows Background: The Intergalactic Medium Observations of the extent of enrichment Model for galactic outflows

27. Galactic Outflows Many simultaneous Type II supernovae (SNe II) in the starburst phase  Coherent Extragalactic Outflows Outflows may be necessary to explain many observations and solve many problems: Metallicity of the IGM Metallicity of the IGM Entropy content of the IGM Entropy content of the IGM Abundance of Local Group dwarf galaxies Abundance of Local Group dwarf galaxies M/L ratio of dwarf galaxies M/L ratio of dwarf galaxies Overcooling problem Overcooling problem …

28. Value of Anisotropic Outflows  Travel preferentially in to low-density regions  Not require large volume filling factors  Travel further  Provide a source of metallicity scatter

29. Observations of Anisotropic Outflows Near IR and visible ACS-HST mosaic of M82 ( Gallagher, Mountain & Puxley) H  emissio n Blue, star forming region

30. For well-formed disks (Mac Low & Ferrara, 1999) Dark matter halo (NFW, MIS) Gaseous disk Simulations of Individual Objects I - Disk scale effects

31.  Better description of  Larger scale effect of many randomly orientated disks  Forming galaxies in starburst phase  Off centre explosions Blue: cold, dense gas Red: hot gas (outflow) Simulation of Individual Objects II - Halo scale effects  Superposition of 3 intersecting plane-wave density perturbations  Galaxy at intersection of 2 filaments inside a cosmological pancake

32. Analytical Model for Anisotropic Outflows L SN R(t) M L SN R(t ) M  e  e.g. Tegmark, Silk, & Evrard (1993), Scannapieco & Broadhurst (2001) MP, Martel & Grenon (2007) The isotropic case An anisotropic case

33. Choice of Opening Angle M82 obs:  ~ 75 o MacLow & Ferrara sims:  ~ 55 o Martel & Shapiro sims:  ~ 100 o Hence, we treat opening angle as a free parameter

34.  3D Gaussian random field of volume (12 h -1 Mpc) 3  Filtered on 10 mass scales to reproduce halo collapse on different scales Unfiltered grid Filtered grids + R 1 = 50.4 kpc M 1= 7.6 x 10 7 M  R 10 =1.86 Mpc M 10 =3.8 x 10 12 M  For each filtered grid:  Find density peaks  Calculate direction of least resistance  Calculate collapse redshift (peak reaches  and forms halo)  Check for mergers and unphysical halos Monte Carlo Simulation

35. Galaxy Formation  Halo gas heated to the T vir during collapse  Cools due to atomic line cooling  which is more efficient for metal enriched gas  Once gas is cooled star-formation begins  10% of gas turned into stars  Neglect life time of the most massive stars and SNe II begin  Galactic outflow quenches star formation  Burst length 50 Myrs  10 51 ergs released by each SN II  1 SNII per 89.7 M sol of stars  Outflow energy escapes galaxy with mass dependent efficiency  Halo gas heated to the T vir during collapse  Cools due to atomic line cooling  which is more efficient for metal enriched gas  Once gas is cooled star-formation begins  10% of gas turned into stars  Neglect life time of the most massive stars and SNe II begin  Galactic outflow quenches star formation  Burst length 50 Myrs  10 51 ergs released by each SN II  1 SNII per 89.7 M sol of stars  Outflow energy escapes galaxy with mass dependent efficiency

36. Density structure around a density peak on the halo smoothing scale:   peak  Ax 2  By 2  Cz 2  2Dxy  2Exz  2Fyz Largest of (A’,B’,C’)  Direction of least resistance  Determine A, B, C, D, E, F by least-square fit.  Rotate coordinate axes to eliminate cross-terms (D,E & F) (x, y, z)  (x’, y’, z’) and   peak  A’x’ 2  B’y’ 2  C’z’ 2 x z’ x’ y z y’ Halo Smoothing Scale The Direction of the Outflow

37. Driving pressure (energy injection and expansion) (Energy deposition rate by Supernovae and dissipation rate by Compton drag) Drag due to sweeping up IGM Gravitational deceleration from the enclosed matter and the halo The Expansion of the Outflow

38. Rate of Energy Deposition/Dissipation … from IMF (Kropa 2001), energy per SN and burst length f w - energy escape fraction f * - star formation efficiency Cooling due to Compton drag against CMB photons: Total rate of driving by SNe: Equations solved numerically for radius, R, at each time-step

39. Radius vs. Opening Angle

40. Largest outflow  = 180 o Halo Mass = 2.8 x 10 9 M  Formed at z=8.1 Example Isotropic Outflow MP, Martel & Grenon (2007)

41. Two possibilities: 1.Ram-pressure stripping (prevents galaxy formation) when 2.Metal deposition: need to recalculate cooling time (leads to earlier galaxy formation) 1.Calc based on volume of overlap and metal content of outflows: Density Peaks Hits Before Collapse Neighbouring Collapsing peak Source Halo 2 M  of metals per SN and IMF f esc - mass escape fraction

42. Our Simulation at End (z=2)  Case of  = 40 o  Red wedges  outflows  Black circles  pre-collapse radius (smoothing scale) of halos with galaxies  Galaxies in a common filament produce aligned outflows MP, Martel & Grenon (2007)

43. Volume Filling Factor Statistics N - Total # of grid points N  - # of grid points at density  N’  - # of enriched grid points at density  - Gas overdensity - Gas overdensity where N’  ’<  - cumulative version (densities below  ) MP, Martel & Grenon (2007)

44. Impact of Reionization on Enrichment of IGM Remember that observations of 4% volume filling factor Cumulative Volume Filling Factor FOC MP & Martel (2007)

45. Impact on Enriching Galaxies FOC Volume Filling Factor MP & Martel (2007)

46. ConclusionsConclusions  Observations consistent with  Full enrichment down to  Partial enrichment down to  Volume filling factor of metal enrichment > 4%  Large unexplained scatter in metallicity  More enrichment near known galaxies but still clear signal of metals far  Anisotropic outflows due to large scale structures  Motivated by observations and simulations  Travel further into low-density regions, away from filaments and sheets of structure  With  in  dramatic  in enrichment of high density systems  Can enrich 10% more of the underdense Universe and 40% more of Universe below  The extent of enrichment is sensitive to epoch of reionization

47. Future Work  Switch to N-body  Produce fake spectra and compare with observations  Consider impact of reionization on PopIII star formation at z<5  More sophisticated models of reionization Simulations Analysis of data Extent of enrichment? More Answers! N-body Analytic

48. CIV pixels recovered

49. LBGs as Markers  LBGs same sample as Adelberger et al. (2003)  Region claimed to contain all CIV absorption  but CIV detected in both samples  LBGs same sample as Adelberger et al. (2003)  Region claimed to contain all CIV absorption  but CIV detected in both samples Pieri, Schaye, Aguirre & Haehnelt (2005) - in preparation

50. Testing the Method  Trivial consequence of method? No (!)  Same as for previous figure but spectrum from hydro simulation  Metals distributed according to Schaye et al. (2003) including scatter  Trivial consequence of method? No (!)  Same as for previous figure but spectrum from hydro simulation  Metals distributed according to Schaye et al. (2003) including scatter

51. Scatter in Abundance  Scatter determined as in Schaye et al. (2003)  Significantly less scatter in near sample  Some evidence that far sample has less scatter  Scatter determined as in Schaye et al. (2003)  Significantly less scatter in near sample  Some evidence that far sample has less scatter MP, Schaye & Aguirre (2006)

52. OVI Near to and Far from Strong CIV  Again excess enrichment close to strong CIV  Little or no evidence for OVI in far sample  Mostly likely since OVI too weak to detect  Better examples of OVI detection at lower z  Again excess enrichment close to strong CIV  Little or no evidence for OVI in far sample  Mostly likely since OVI too weak to detect  Better examples of OVI detection at lower z

53. Comparison of Real and Synthetic Spectra

54. Cosmological Simulations Gaussian random density fields on a cubic grid, filtered on various mass scales  CDM model (WMAP) Peak finder (calculates direction of least resistance) Press-Schechter formalism List of halos: M, r, z coll, e M, z coll  T vir Cooling model z gf Outflow model Evolution and feedback

55. Radio observations e.g. of Sculptor Dwarf Galaxy Evidence of anisotropic outflow or gas torus? HI radio contours Optical Carignan et al. 1998 Observations of Anisotropic Outflows

56. The Numerical Algorithm Initial conditions: list of peaks. For each peak: r, M, z coll, z gf, z mg, e. z i = 20, z f = 0,  z =  0.005 rM t gf Rhalos outflows For each iteration For each peak z = z coll ? For each halo z = z gf ? z = z mg ? x x For each outflow Update R Check for hits

57. z   peak  Ax 2  By 2  Cz 2  2Dxy  2Exz  2Fyz   peak  A’x’ 2  B’y’ 2  C’z’ 2 z’ y’ x’ x y A D E D B F E F C 1, 2, 3 1, 2, 3 e 1, e 2, e 3 The Direction of the Outflow II

58. An SPH simulation - Theuns et al (2002) log (temperature) log(density) log(HI abundance) Stars & Supernovae log(Z) HeII(abundance) 5 Mpc/h

59. Changes in Scale for Direction Determination Changes in direction  for all halos  switching from Halo smoothing scale to the minimum possible scale (12 pixels)

60. Simulations of Galactic Outflows in Cosmological Volumes (Semi-)Analytical Methods Numerical Simulation Outflow described analytically (e.g. Scannapieco & Broadhurst 2001) All but Aguirre et al. (2001) do not have anisotropic outflows  Cosmological Volume (~ (10 Mpc) 3 )  Thousands of galaxies  Anisotropic outflows have not been described sufficiently if at all Hydrosimulations with energy deposited into particles or arranging particles into a shell and giving them a radial velocity (e.g. Theuns et al, 2002, Scannapeico, Thacker & Davis 2001) Often do not have enough resolution per galaxy to produce anisotropic outflows.

61. Aguirre et al. (2001)  Combination of SPH and analytical outflow  Use completed SPH simulations and the outflows are a post-process  Starts isotropic and each radial direction calculated individually  Fixed radial trajectories and so becomes anisotropic based on the failure of the outflow to propagate in a given direction  Disadvantages  Combination of SPH and analytical outflow  Use completed SPH simulations and the outflows are a post-process  Starts isotropic and each radial direction calculated individually  Fixed radial trajectories and so becomes anisotropic based on the failure of the outflow to propagate in a given direction  Disadvantages  Real outflows are likely to acquire a tangential component and continue to contribute to the outflow  No feedback effects are included as galaxies have already been simulated  No investigation of the significance of varying opening angle

62. Problem: the different filtered grids represent the same region of the universe. Two different peaks on different grids will contain the same matter if they are at similar locations. They cannot both form a galaxy: violation of mass conservation. M a (z coll ) a M b (z coll ) b 1 st case: (z coll ) b > (z coll ) a (rare) 2 nd case: (z coll ) a > (z coll ) b (frequent) The matter that was going to end up inside halo a is removed by the formation of halo b. Suppression: Halo a will never form. Halo a forms at redshift (z coll ) a. Later, at redshift (z coll ) b, halo a is destroyed by a merger event that leads to the formation of halo b. Suppression and Merging

63. Number Density Statistics

64. End Point Results MP, Martel & Grenon (2007)

65. Isotropic Anisotropic, large  Anisotropic, small  Outflow from galaxy located inside a filament

66. The Cooling Time M cool r cool Starts as an isothermal sphere Density  -2 Cooling rate higher at center Cooling proceeds “inside-out” r > r cool : gas is still hot r < r cool : gas has cooled (White & Frenk 1991, ApJ, 379, 52) R cool and M cool increase with time When M cool = M, cooling is completed t = t gf z gf

67. Threshold Threshold Points from Dijkstra et al 2004 F C

68. Star Formation Efficiency But an increase in star formation leads to earlier reionization & more suppression to counter this

69. Anisotropic Outflows and Photoionization Suppression