Radiationinduced largescale structure

1
Radiation-induced large-scale structure
Rupert Croft (Carnegie Mellon)
2
Studing Radiation-Induced LSS Motivation We
know a lot about the growth of large-scale
structure due to gravitational instability from
small seed fluctuations. -gt seen in galaxy
surveys, Lya forest etc -gt widely studied
linear and higher order PT, halo model,
cosmic web morphology etc What about
large-scale structure caused by other
mechanisms? -gt one example is structure in
the neutral hydrogen density field caused
by sources of radiation Radiation Induced
LSS -gt how is it different in visual
appearance and statistically from
gravitationally induced LSS? -gt what can
statistics tell us about sources of radiation
(first stars, quasars, decaying DM etc)
and about cosmology?
3

• Talk plan
• Structure formation during reionization
• radiative transfer vs gravity.
• (2) Quasar light echos and how to find them.
• (with Eli Visbal, CMU)
• (3) The quasar proximity effect from the SDSS and
  joint
• constraints on the ionizing BG and baryon
  density.
• (with Taotao Fang, UCB)

(All using cosmological hydro simulations)
(All work in progress)
4
Many codes exist for doing RT around the first
stars/QSOs e.g., Abel et al 1999, Razoumov
Scott 1999, Gnedin 2000, Ciardi et al 2000,
Sokasian et al 2003, Cen 2002, Bolton
2004, Iliev et al 2005, Mellema et al 2005 Only
recently have simulations started to resolve DM
halos of mass 109 Msun which dominated the
ionizing radiation output, as well as having box
sizes large enough for bubbles of diameter 10
Mpc/h or more e.g., Iliev et al 2005, Kohler
et al 2005
5
Code Monte Carlo Radiative transfer -gt
raytraces photon paths through SPH kernels
-gt source photons and recombination photons

better than this
source
No gridding needed, so keeps the high resolution
of the simulation 10 Kpc/h vs 0.4 Mpc/h for
typical gridded sim.
6
Ingredients
(a) Gadget hydro simulation
2 x 2563 particles
40
Mpc/h box
10 Kpc/h resolution (b) RT run as
post-processing (c) Sources of radiation
associated with DM halos -gt simplest
idea we assign a mass/light ratio in a
fashion similar to Zahn et al 2006
(astr0-ph/0604177) actually 1.2
x 1042 ionizing photons/sec/M this is like
Pop II stars forming with efficiency f0.1
from a Salpeter IMF, with stellar
lifetime 5x107 yrs and escape
fraction fesc0.05 this is our fiducial case
.
.
7
the idea is to vary unknown parameters, and see
what effect this has on the LSS. -gt 9
different RT runs so far
(others have softer nu-4 spectrum)
models 5-9 are normalized to have the same total
radiation output as model 1 (fiducial) by z6
8
40 Mpc/h
1 Mpc/h thick slice
9
90 neutral by mass at z10.0
10
50 neutral by mass at z8.2
11
10 neutral by mass at z7.8
neutral remnants mostly in voids - can they be
detected/ tell us anything?
12
mass-weighted neutral fraction vs redshift

• effect of recombinations quite small for these
  late reionizing models
• (density is relatively low)
• reionization process is fastest for sources
  hosted by large halos only

13

• early on, mfp is strongly affected by
  recombinations as
• photons try to escape from dense regions around
  sources
• mfp of hard spectrum model is always gt 0.3
  Mpc/h. For a
• more realistic AGN spectrum it would be even
  more. This
• affects recombinations.

14
we will compare models when there is 50 neutral
fraction by mass
fiducial model
15
plots of neutral density(rho x neutral fraction)
16
fiducial (all halos gt 109 Msun contribute)
only halos gt 1010 Msun contribute
only halos lt 1010 Msun contribute
17
neutral fractions
18
10 ionized
30 ionized
90 ionized
70 ionized
(for fiducial case)
19
for all models 50 ionized
20

• Plot of ionized
• density instead
• of neutral
• density
• easier to see
• sources

21
(No Transcript)
22
fiducial
ionized density
big galaxies only
small galaxies only
23
correlation function of mass and HI mass
clustering goes down and then up again
99 ionized
slope and amplitude can be very different
from mass xi bias gt 10 for 1 neutral field
97 ionized
90 ionized
growth factor for HI fluctuations much
larger during this short epoch than for rho (this
plot spans z from 10-7.5).
10 ionized
fiducial case
24
quarter fiducial luminosity
25
ratio of HI to matter correlation function
fiducial case
26
Many detailed semi-analytic models for
reionization exist e.g. Miralda-Escude 2000,
Furlanetto 2003, Zahn et al 2006
We will instead compare to a Very simplistic
bubble model (Babul White 1991 give analytic
form for bubbles of filling factor f and radius
r in a uniform medium)
(actually we mean intrabubble)
Assumes no correlation between density and bubbles
27
For mean neutral fraction lt 0.6, this bubble
model doesnt work (we have antibias)
28
(for all models 50 ionized)
29
neutral fraction
fiducial
10 ionized
40 ionized
70 ionized
nu-2
spectrum
30
Non-Gaussianity? e.g., how is S3 different
from under gravitational evolution?
S3 is skewness/ variance2 higher order
perturbation theory prediction for
graviational S3 is 4 for these (0.110 Mpc/h)
scales
31
PT for rho
S3 becomes constant for high ionized fraction
minimum at 30 ionized
32
all models 50 ionized
33
OK- this was structure in the HI field around
reionization - what about radiation-induced LSS
at later times? We will look at
z3. Differences from z10 . Ionizing photon
mean free path is much longer 200 Mpc/h .
Radiation field is much more uniform only very
bright rare sources(quasars) will have any
noticeable effects . Observational probe is the
Lya forest.
34
250 Mpc/h box, z3
matter
ionizing radiation
What are these weird light echos, and how can
we detect them?
35
density field around a quasar (50 Mpc/h wide box)
36
quasar light curve
37
neutral hydrogen density field around quasar
38
(No Transcript)
39
Lya forest probes of density field along line of
sight
40
Fe-t
(observable quantity)
Density of matter
For material in photoionization equilibrium (see
e.g. Hui and Gnedin 1997)
Photoionization rate (proportional to Ionizing
radiation intensity)
(Tau is prop to HI density - Gunn Peterson
1965)
41
lya spectra with uniform BG and with BGquasar
5 different sightlines
42
Simulation test put 50 quasars behind a 2
deg x 2 deg area at z3 make simulated lya
spectra (2 Angstrom res) try to detect a light
echo that we put in the noise is structure in
thedensity field
Method make a template and slide it through
the dataset. -gt need 5 parameters (x,y,z, quasar
luminosity, time
since quasar switched off) -gt very time
consuming as we need a 5 parameter grid
43
simulation test in 1D
44
What is the chance of finding something that
looks like a light echo, but is just a chance
set of density fluctuations? -gt we look for
light echos in 1000 simulations with only
density fluctuations contributing to the Lya
forest
For quasars with bolometric luminosity 5x1045
erg/s, statistical significance of detection is 1
in 1000
45
data like COSMOS survey (e.g. Impey et al 2006)
can be used to construct a grid of sightlines
through a volume
46
If we find light echos, what then? (1) They
will tell us about the lifetimes and
luminosities of long dead quasars.
(2) They are interesting objects and may be
useful for cosmology - for example,
their angular and redshift extent can
be used to construct a geometric test.
47
(c) Quqsar proximity effect from SDSS
Proximity effect (Bajtlik etal 1987) is deficit
of lya absorption close to observer
quasar. Unlike light echo case, we
know luminosity of quasar, so we can predict
what the deficit should be
Here Gammagamma for BG
gamma for this quasar
Can use this to measure gamma for BG largest
current measurement is Scott et al 2000, from
100 quasar spectra
48
Fexp(-tau)
large r depends on BG gamma because of inverse
square law
small r depends on quasar gamma
r
If we know quasar luminosity, we can in
principle constrain Gamma_BG and Omega_b
49
We use 3000 quasars from SDSS DR3 -use only
quasars above z2.4 so that mean z of lya
forest is 3.0
50
We fit continua using PCA components derived from
red side only (method of Suzuki et al 2005) -gt
the important region is that close to the Lya
emission line.
51
As a test, we can reject spectra which the PCA
doesnt fit well around the red side of the lya
emission line
52
Results
Scott et al 2000 proximity effect with 1 sigma
errors
Total seen from quasars only Haardt and Madau
1999
53

• Conclusions
• We have looked at 3 examples of RIS
• RIS during reionization encodes information about
• the radiation sources.
• Structure is qualitatively different from
  under gravity.
• There is a rich phenomenology to explore.
• (2) Light echos from bright quasars should be
  detectable.
• We can use them to tell us about the
  quasar lifetime, and
• about long-dead quasars.
• They are structures generated by the
  radiation field itself
• and may have some uses for cosmology.
• (3) The proximity effect (the RIS in the Lya
  forest very close
• to quasars) can be used to constrain the
  baryon density
• (using SDSS Lya spectra gives results
  consistent with
• BBN/CMB) and ionizing radiation BG
  intensity (need