Accretion Physics in the
SDSS/XMM-Newton
Quasar Survey
Monica Young
with
Martin Elvis, Alan Marscher
& Guido Risaliti
SDSS/XMM Quasar Survey
• Optical: SDSS DR5 quasars
– 90,611 quasars
– 0.1 < z < 5.4
• X-ray: XMM-Newton
– Large field of view
• 1% overlap between archive and SDSS
– Large effective area  light bucket
• Result: 792 quasars with X-ray observations
– Available on HEASARC archive
3 Optical/X-ray Trends
X-ray loud
1. αox-Lopt
Green et al. 2009
X-ray quiet
2. Γ vs. Lx
3. Γ vs. L/Ledd
Steffen et al. 2006
Shemmer et al. 2008
3 Optical/X-ray Trends
X-ray loud
1. αox-Lopt
Young et al. 2009
X-ray quiet
2. Γ vs. Lx
3. Γ vs. L/Ledd
Young et al. 2009
Risaliti, Young & Elvis 2009
Monte Carlo Population Study
• Define sample: 106 quasars
– Draw (z,Lopt) randomly from quasar luminosity
function (Hopkins et al. 2007)
• Apply SDSS and XMM-Newton selection
– SDSS selection/flux limits
– XMM 6σ sensitivity: fn(Texp,θ)
• Find out which relations
are intrinsic to the
parent population
Optical/X-ray Trends
1. The αox-Lopt Relation
Is αox-Lopt Real?
αox = normally distributed around
<αox> = -1.6, σ = 0.17
αox = -0.137*log L2500 + 2.64, σ = 0.15
(Steffen+06)
 Selection effects cannot reproduce correlation!
αox
log L5000
1500 Å
5000 Å
αox
4 keV
log L1500
αox
1 keV
αox
αox-Lopt stronger effect in X-ray energy
log L1500
log L5000
Slope and scatter change strongly with X-ray energy
• Slope steepest at
low X-ray energy
• Closer to linear at
highest energies
• Change in correlation
slope is not due to
change in baseline
over which αox is
defined
Slope of αox-Lopt
Slope of αox-Lopt Relation
“Baseline Effect”
1keV
10keV
X-ray Energy (keV)
To understand why, need to understand the Γ-Lx anti-corr.
Optical/X-ray Trends
2. The Γ-Lx Relation
The Γ-Lx Relation
• Significant correlation above 2 keV
– Consistent with Green et al. 2009
– Strengthens with X-ray energy
Young+09
Green+09
2 keV
3.0σ significance
10 keV
8.6σ significance
Simulated Γ-Lx Relation: Assume Γ = f(Lbol/LEdd)
Observed slope
Simulated slope
Γ
Γ
log L2 keV
0.7σ significance
log L10 keV
6.0σ significance
• Correlation strengthens artificially with energy
• But artificial correlation not significant at L2
Simulated Γ-Lx Relation:
Assume Γ = f(Lx, Lbol/LEdd)
Observed slope
Simulated slope
4.3σ significance
9.0σ significance
• If X-ray slope is a function of Lx and Lbol/LEdd,
then observed slope, strength reproduced
Γ-Lx Correlation Due to Soft Excess?
• Lx-z correlated (flux-limited)
– Soft excess enters X-ray
spectrum at low z
• Make redshift cut: z > 1
 Γ-Lx correlation disappears
• Is soft excess strength
related to z or to Lx?
– Subject of future study
Γ-Lx Relation Steepens αox-Lopt
Simulation shows that αox-Lopt slope changes
with energy due to Γ-Lx anti-correlation
Observed
Simulated
X-ray Energy (keV)
Γ = f(L2 keV)
Slope of αox-Lopt
Slope of αox-Lopt
Γ = f(Lbol/Ledd)
X-ray Energy (keV)
αox-Lopt Independent of Baseline
Schematic Diagram
Account for effect of
Γ-Lx relation on αox-Lopt slope
 Implies constant αopt, Γ
with respect to luminosity
log νFν (ergs cm-2 s-1)
 αox-Lopt slope is independent
of optical and X-ray
reference frequencies
Opt/UV (disk)
X-rays
(corona)
log ν (Hz)
What drives αox?
• Lopt is the primary driver of αox
• BUT accretion rate is a secondary driver
– Partial correlation (αox, L/LEdd, Lopt)  7σ
X-ray bright
 Seed photon luminosity
and accretion rate both
drive X-ray efficiency
X-ray faint
log L/LEdd
αox and Comptonization Models
Thermal Comptonization Model
• Heating rate ~ lh ~ Lx/Rx
• Cooling rate ~ ls ~ Lo/Ro
lh/ls
• αox  lh/ls  geometry
Γ=1.6
T=2e9 K
lh/ls~2
Coppi 1999
lh/ls >> 2
“photon-starved”
lh
Physical Scenario (“Patchy” corona)
Low Lbol
As luminosity increases,
so does the covering
factor (i.e., more blobs).
The corona cools as it
intercepts more disk
photons.
High Lbol
The optical depth remains
constant (τ~0.1),
so Γ steepens: ΔΓ~0.2
for ΔL2~1.3 dex.
(comparable to error in Γ)
Conclusions
•
SDSS/XMM-Newton Quasar Survey (SXQS) is a powerful tool!
– 473 quasars with both optical and X-ray spectra – unprecedented sample size!
– Monte Carlo population study quantifies selection effects in the survey
•
Determine which relations are intrinsic
– Γ-Lx – not intrinsic (due to soft excess component at low z)
– αox-Lopt – intrinsic
– αox-Lopt slope constant with respect to the reference frequencies
• Implies αopt and Γ constant with respect to luminosity
•
Disk-corona structure changes with L/LEdd
– Use αox-Lopt as input to Comptonization models
– To reproduce αox-Lopt relation, the heating to cooling ratio must decrease
 covering factor of corona increases with luminosity (i.e., with L/LEdd?)
•
Next step: Defend thesis! (July 15)