recent advances in the study of black holes using x-ray...

Black Hole Binaries

Stellar-mass black holes(∼5–30M�)

Modest variability in humantimescale

Many are transient in nature

Bright outbursts can last severalmonths with up to a billion foldincrease in luminosity

AU-scale persistent jets andparsec-scale ballistic jets

X-ray QPOs (0.01–450 Hz)

Distinct spectral states(hard/intermediate/soft)

Javier Garcıa (Harvard-Smithsonian CfA) Reflection Spectroscopy of BHB FERO 8, Sep 14, 2016 2 / 35

Motivation

Dramatic spectral changes throughout the outburst!Nowak: The Microquasar Cyg X-1 203

1 10 100

10−9

10−8

Energy (keV)

νFν

(erg

s cm−2

s−1 )

Fig. 1. Cyg X-1 in its two most extreme spectralstates: the hard state dominated by a cuto↵ power-law spectrum extending to > 100 keV, and the softstate dominated by a thermal spectrum peaking be-tween 1–2 keV. The above hard state is from a multi-wavelength campaign wherein Cyg X-1 was ob-served by every flying X-ray satellite. Here we dis-play Chandra-HETG spectra (histogram), Suzaku-XIS (solid diamonds) and -GSO (hollow triangles)spectra, RXTE-PCA (circles) and -HEXTE (hol-low diamonds) spectra, and INTEGRAL -SPI (solidtriangles) spectra (see Nowak et al. 2011). Thesoft state spectra are from a simultaneous Chandra-HETG and RXTE campaign conducted in January2011. The above spectra are presented without ref-erence to any underlying spectral model, and are un-folded using only the spectral response of the detec-tors (see Nowak et al. 2005).

1; a description of these spectra can be foundin Nowak et al. (2011). Also shown are spec-tra from simultaneous Chandra-HETG/RXTE-PCA observations, covering the 0.5–40 keVrange of a spectrally soft state. In terms ofobserved flux, the 0.1–50 keV flux of the softstate is 3.9 ⇥ 10�8 erg cm�2 s�1, while the ob-served 0.5–300 keV flux of the hard state is4.9⇥10�8 erg cm�2 s�1. However, making plau-sible corrections for absorption and extrapolat-ing the spectra to determine bolometric lumi-nosities, this is at 2.2% LEdd and this hard stateis at 1.6% LEdd. (The Eddington luminosity,LEdd, used here assumes a distance of 1.86 kpcand a mass of 15 M�; Reid et al. 2011; Oroszet al. 2011.) That is, whereas we directly ob-serve most of the flux in the hard state, nearly2/3 of the soft state flux is unobserved due toboth absorption and bandpass limitations.

Fig. 2. Hardness-intensity diagram of a blackhole transient outburst (based upon GX 339�4;Belloni et al. 2005; Homan et al. 2005; adaptedfrom www.issibern.ch/teams/proaccretion).Transients begin faint, hard, and radio loud; theyevolve to brighter states while remaining hard; theysoften and become radio quiet (often preceded bya radio ejection event); they then fade, harden, andbecome radio loud once more. Cyg X-1, which ispersistently emitting in the X-ray band, occupiesonly a small portion of this diagram at fractionalEddington luminosities of ⇡ 2%, near the so-calledradio loud/radio quiet transition “jet line”.

As has been noted by previous researchers,the range of bolometric luminosities traversedby Cyg X-1 spans only a factor of ⇡ 3–4(Wilms et al. 2006, and references therein).This is somewhat narrow compared to mostblack hole transients. Furthermore, Cyg X-1also traverses a narrower range of colors thanmany black holes, never exhibiting a purelydisk-dominated spectrum without a hard tail(e.g., like the simple disk-dominated spectrumof 4U 1957+11; Nowak et al. 2008). Blackhole transients often follow color-intensity di-agrams as shown in Fig. 2, spanning & 3 or-ders of magnitude in luminosity, and wider ex-tremes of color variations. Cyg X-1, however,exists on an especially interesting portion ofthis ‘q-diagram’ — moving between the radioquiet/soft, radio loud/hard-intermediate statetransition near ⇡ 2% LEdd.

The hypothesized emission components inthe Cyg X-1 system include an accretion disk,a Comptonizing corona (with either a ther-

(Nowak+12)

Study the accretion properties of Galactic Black Holes using the RXTE archive

Detailed analysis of individual sources with physically motivated models

Dynamically track the evolution of key parameters → inner radius


Modeling X-ray Reflection: XILLVERX-ray Reflection from the Inner-DiskFirst need to solve the full photoionization problem in a optically-thick slab➜ COMPLICATED!Ionization Balance

1e−12

1e−10

1e−08

1e−06

0.0001

0.01

1

0.0001 0.001 0.01 0.1 1 10

Ion

Fra

ctio

n

Thomson Depth

HHe IHe II

J. Garcıa (CfA) X-ray Reflection Models August 29th, 2014 8 / 1

Ionization Balance

1e−12

1e−10

1e−08

1e−06

0.0001

0.01

1

0.0001 0.001 0.01 0.1 1 10

Ion F

ract

ion

Thomson Depth

O IO IIO IIIO IVO VO VIO VIIO VIII


Ionization Balance

1e−12

1e−10

1e−08

1e−06

0.0001

0.01

1

0.0001 0.001 0.01 0.1 1 10

Ion

Fra

ctio

n

Thomson Depth

Fe IFe IIFe IIIFe IVFe VFe IXFe XFe XIFe XIIFe XIIIFe XIV

Fe XVIIIFe XIXFe XXFe XXIFe XXII


XILLVER: log ⇠ = 2, � = 2, AFe = 1

1025

1026

1027

1028

0.1 1 10 100

E*F

E (

erg

s/cm

2/s

)

Energy (keV)

ReflectedIncident PL


Final spectrum depends on the illumination, geometry, and composition of the material in the disk

Calculation done with the XILLVER code (Garcia & Kallman 2010)

http://hea-www.cfa.harvard.edu/~javier/xillver/

Monday, April 20, 15


Modeling Relativistic Reflection: RELXILL

RELXILL: Relativistic reflection model that combines detailed reflection spectrafrom xillver (Garcıa & Kallman 2010), with the relline relativistic blurring code(Dauser et al. 2010).

40

20

0

-20

-40

6040200-20-40-60

-0.3 -0.2 -0.1 0.0 0.1 0.2 0.5 1.0

1.4 1.3 1.2 1.1 1 0.9 0.8 0.6 0.4100-10

10

0

100-10

10

0

β [rg]

α [rg]

Redshift z

E/Ee

Model Parametersa: Black hole spin, Rin: Disk’s inner edge

i : Inclination, ε: Emissivity index

Rf : Reflection fraction, Γ: Power-law index

Ecut : High-energy cutoff, AFe : Fe abundance

1.00

10.00

1 10 100

E *

Flu

x (

a.u

.)

Energy (keV)

Fe Kα

Fe K-edge

Compton Hump

UnblurredBlurred

(Garcıa+14b)


The Prototypical BHB GX 339-4

Is the inner radius truncated in the hard state?

10

100

1000

0 0.2 0.4 0.6 0.8 1 1.2

0.01

0.1

Norm

aliz

ed

PC

A C

ou

nt

Rate

X-r

ay In

ten

sity

(L/

L Ed

d)

X-ray Hardness (HR color)

1

2

3456

GX 339-4 + RXTE PCA: Vast amount of observations in a wide range of lumi-nosities and accretion states. Data is free of pile-up.


Fitting the Hard-State of GX 339–4

Excellent constraints on fundamental parameters of the system: BH spin(a∗ = 0.95± 0.04), inclination (i = 48± 1 deg), and Fe abundance (AFe = 5± 1)

The number of free fitting parameters is large, 52 for JF-I and47 for JF-II. The complexity of the analysis dictated ourapproach: We performed Markov Chain Monte-Carlo (MCMC)runs using the EMCEE-HAMMER Python package (Foreman-Mackey et al. 2013), which implements affine-invariantsampling. MCMC methods are powerful for high-dimensionalanalysis. Specifically, they enable an efficient exploration ofparameter space and determine a posterior probability structurefor the model of interest.

5.1. Joint Fit I: Fixed Spin and Variable Inner Radius

A principal goal of our study is to track the radius of theinner edge of the disk Rin as the luminosity varies by an orderof magnitude (i.e., over the range 1.6%–17% of Eddington;Table 1). As discussed in Section 1.1, a question of greatinterest is whether the inner disk is truncated in the hard state at

low luminosities and, if so, to what extent. In order to bemaximally sensitive to a disk that is only slightly truncated, wefix the spin to its maximum allowed value, namely a* = 0.998.In so doing, our focus is on determining how Rin trends withluminosity rather than obtaining accurate estimates of thisparameter. We note that most spin determinations in theliterature (which assume Rin = RISCO) suggest that the spin ishigh (see Section 6.1.3).Figure 6 shows the fit residuals for JF-I for two cases: (1) The

top panel shows a data-to-model ratio for a fit to Model 0, i.e.,the simple absorbed power-law model used to produce theresidual plot for a fit to Spectrum A only (which is shown in theleft-top panel of Figure 3). This simple fit prominently displaysthe reflection features, which are strong for all six spectra. Theprofile of the Fe K line shows moderate variations among thespectra. As the luminosity increases, so does the intensity of the

Figure 6. Simultaneous fit to Spectra A–F comprising a total of 106 individual spectra and 77 million counts. (Top) Ratio plots for a joint fit (JF-I) using Model 0, thesimple absorbed power-law model, which strikingly reveal the principal signatures of reflection. (Middle) Ratio plots for these same data obtained by fitting ourcanonical Model 3 (Tbabs∗(relxill+xillver)∗gabs). This model produces an excellent fit with 1.06.2c =n (Bottom) Contributions to the total χ2 (data-model), again for Model 3 and the same data.

9

The Astrophysical Journal, 813:84 (19pp), 2015 November 10 García et al.

Fit with relxill:

- 106 observations- 77 million total counts- 0.1% systematic errors!

χ2ν = 1.06

(Garcıa+15)


Detecting Geometrical Changes

For increasing luminosity, thedisk’s inner edge moves inwardand the corona cools down

For a 10M� black hole, thesechanges in inner-radius corre-spond to changing from Rin =75 km to Rin = 30 km

(Garcıa+15)


Detecting Geometrical Changes

9/12/2016 Znojemská 113, 669 02 Hnanice, Czech Republic to Weinviertler Schnellstraße, 2011, Austria - Google Maps

https://www.google.com/maps/dir/48.7976319,15.9882055/48.4612879,16.1526821/@48.630577,15.7332738,10z/am=t/data=!3m2!4b1!5s0x476da9d75c60313b:0xb... 1/3

Map data ©2016 GeoBasis-DE/BKG (©2009), Google 5 mi

Drive 45.1 km, 40 minZnojemská 113, 669 02 Hnanice, Czech Republic to

Weinviertler Schnellstraße, 2011, Austria


Accurate Estimation of the Inner Radius

Data recalibration with the PCACORR tool (Garcıa+14) allows a ten fold increasein sensitivity to the reflection features → increased accuracy!

0.92

0.96

1.00

1.04

1.08

1.12

1.16

Data

/Mod

el

Model: Power Law

Uncorrected (1% syst.)Corrected (0.1% syst.)

Response (18% resolution)

0.97

1.00

1.03

5 25

Data

/Model

Energy (keV)

Model: Relativistic Reflection 0

1

2

3

4

1 2 3 4 5 6 7 8

Δχ2

Inner-Disk Radius (RISCO)

Rin < 1.7 RISCO

Rin < 6 RISCO

Uncorrected (1% syst.)Corrected (0.1% syst.)

90% Conf.


Location of the Inner Radius

100

101

102

103

0.1 1 10

Rin

(R

g)

L/LEdd (%)

Plant+15

Koleh+14

Petrucci+14

Tomsick+08

Miller+06

Reis+08

Basak+Zdz16

Garcia+15

Garcıa+15


XMM-Newton Epic Pn (TM) vs. RXTE PCA

0.95

1.00

1.05

1.10

1.15

4 20 45 10

Data

/Mod

el R

ati

o

Energy (keV)

XMM-Newton Timing-ModeRXTE PCA

XMM Timing Mode data much narrower Fe K line than simltaenous RXTE PCAdata! → Possible calibration issues?


Including the Disk Emissionpossible variation with luminosity. The reflection analysis is fully analogous to our earlier study ofGX 339–4 (except that we will employ the self-consistent model variant from Figure 3).

1 10 100Energy (keV)

0.01

0.10

1.00

F ν [k

eV (P

hoto

ns c

m-2

s-1 k

eV-1

)]

seed thermal diskComptonized disk

relativistic reflectiondistant reflection

total

-50

5

RXTE: self-consistent thermal disk

10Energy (keV)

-50

5

∆ χ

3 10 30

-50

5

RXTE: no thermal-disk

Figure 3: (Top) A self-consistent analysis links the coro-nal power-law component, (seed) thermal disk emission,and relativistic reflection components, for the most lu-minous of GX 339–4’s hard state spectra observed byRXTE. An additional component of unblurred reflectionis also in evidence, from the corona irradiating the flaredouter lip of the disk. The 44-million count spectrumfrom Garcıa et al. (2015) is shown in black. (Bottom)Residuals illustrate the quality of fit to RXTE data us-ing our self consistent model (upper panel, black), orwith a standard model that ignores thermal disk emis-sion (lower panel, red). Because RXTE’s range begins at⇠ 3 keV, it is insensitive to the predicted disk emission.

Most crucially, Chandra will open a new win-dow into the thermal disk emission and providea cross-check on the reflection prediction. Adirect detection of thermal emission would beproof positive of a small inner radius. BecausekTdisk / R�3/4, a large truncation radius, even10s of times the size of RISCO would be essentiallyundetectable in direct thermal emission.

Because the time for the peak of the bright-hard state is variable, we employ a strategy usinga TOO plus two followup observations (which isalso useful for tracking any changes with lumi-nosity). Our target sensitivity is > 4 sigma for a0.15 keV disk in a 100 mCrab BH (for GX 339–4,this corresponds to 6% Eddington and R ⇡ 6Rg).Ideally, we will actually reach > 10� given su�-ciently bright data, and a low reflection Rin.

For our analysis we will employ our suite ofrelativistic reflection models relxill, and a newvariant of our Compton-scattering kernel simpl(Steiner et al. 2009), simplcomp which uses themore rigorous Comptonization kernel given bynthcomp (Zdziarski et al. 1996).

4 Technical Feasibility

4.1 Trigger Condition

Using RXTE’s full archive of GX 339–4 outburstsas a template for outbursting BH behavior (no-tably GX 339–4 is itself the single most likely system to have a triggering outburst), we analyzedthe outburst patterns of GX 339–4 to establish our TOO strategy. From its outbursts on record,GX 339–4 is always first detected in the hard state while its flux increases. The full durationof this initial hard-state phase can range from ⇠ 1 � 10 months. However, upon crossing aboveF = 100 mCrab, the bright hard state peaks ⇠ 1 month later. This holds for both GX 339–4’sfainter and brighter outbursts.

Our plan upon trigger is scheduled accordingly: The BH transient must be newly in outburst,hard (� < 2), and rising in flux. Any such system reaching 50 mCrab will be in our sights, and atrigger will be activated upon crossing 100 mCrab (hardness and intensity will be monitored usingeither or both of Swift BAT and MAXI). The first observation is to take place anytime between 5and 15 days after trigger, with the two followup observations to be spaced apart by at least 5 days.All observations are to be completed within 1 month of the trigger.

Each of the three observations will be coordinated between Chandra and NuSTAR teams. Be-cause evolution through the hard state generally takes several weeks, strict simultaneity betweenChandra and NuSTAR is not necessary (though it is preferred); it will be acceptable to have bothmissions collect data within a 24-hour window.

3

Steiner+16

* Self-consistent model including diskemission* Parameters linked to the inner radiusand the accretion rate from the soft-state data:

kTdisk = kTsoft

(M˙Msoft

)3/5 (Rin

RISCO

)−6/5

Ndisk = Nsoft

(M˙Msoft

)−4/5 (Rin

RISCO

)18/5

RXTE insensitive to the predicteddisk emission→ Needs low-energycoverage


The Peculiar Hard State of XTE J1752–322

A total of 60 single RXTE pointings spanning ∼1 month for 300 ks of exposure

101

102

103

0 0.2 0.4 0.6 0.8 1

PC

U-2

Cou

nt

Rate

X-ray Hardness

GX 339-4XTE J1752-223

(Garcıa+16)


XTE J1752–322: An Outstanding Hard-State Spectrum

102

103

104

105

106

107

10 100

Cou

nts

keV

-1

Energy (keV)

XTE J1752-223

PCA (100 million cts)HEXTE (10.4 million cts)

(Garcıa+16)

Two new methodologies are implemented:1) Re-calibration of the HEXTE instruments2) New reflection models with a Comptonization continuum


Fits to XTE J1752–322

Rapidly spinning BH (a∗ ∼ 0.992), also with super-solar abundance (AFe ∼ 4)

10-3

10-2

10-1

100

101

keV

2 (

Photo

ns

cm-2

s-1

keV

-1)

Total Model

Coronal

Distant Reflection

XTE J1752-223PCA + HEXTE

Emission

Relativistic Reflection

-3

0

3

1 10 100

Δχ

Energy (keV) (Garcıa+16)


Other Sources: XTE J1550–564

XTE J1550–564 has good determinations of mass, distance, inclination and thusspin via the continuum-fitting method (Steiner+11) → Would reflection spec-troscopy provide similar results?


A lot more data to analyze!

The RXTE PCA Archive contains over 15,000 spectra from 29 black hole binaries,corresponding to 30 Msec of total exposure


(Bonus Track)High-Density Effects in the

Reflected Spectra


High-Density Effects

High-density models produce a remarkable excess of flux at soft energies. Free-freeemission becomes important (∝ n2

e)!

1022

1023

1024

1025

1026

1027

1028

100 101 102 103 104 105

E*F

E (

erg

cm

-2 s

-1)

Energy (eV)

Γ=2.3, ξ=50

log(ne)=15log(ne)=16log(ne)=17log(ne)=18log(ne)=19

Incident

(Garcıa+16b)


Limitation in Atomic Data

Current atomic data is only applicable up to ne ∼ 1018 − 1019 cm−3, but we needto go a lot higher! (ne ∼ 1022 cm−3 at least)

0.01 0.03 0.1 0.3 1

1−f

10

102

103

104

105

106

107

108

109

M(M

¯)

13

15

17

19

21

23

25

27

0.99 0.97 0.9 0.7 0f

1015

1016

1017

1018

1019

1020

1021

1022

1023

5 5.5 6 6.5 7 7.5 8 8.5 9

Lum

inosi

ty (

erg

/s/e

rg)

Energy (keV)

log(ne)=10log(ne)=22

(Garcıa+16b)

Could high-density effects solve the large Fe abundances problem?


Final Remarks

The bright hard state shows signs of mild truncation (∼few Rg ),with the disk moving closer to the black hole as the luminosityincreases.

However, large truncation (∼hundreds of Rg ) would contradictthe observed continuum and would imply much lower disk temper-atures (kTdisk ∝ R−3/4).

The large Fe abundance predicted by the models is not yet wellunderstood.

New atomic data for high-density plasmas is needed to properlymodel the reflected spectra in this regime.

Reproducibility is one of the main principles of the scientificmethod.


recent advances in the study of black holes using x-ray...

Documents