A Long Look At MCG-5-23-16 With NuSTAR: I- Relativistic Reflection And Coronal Properties

1.3 MB·

by  Abderahmen Zoghbi

Draft version January 11, 2017
PreprinttypesetusingLATEXstyleemulateapjv.5/2/11
A LONG LOOK AT MCG–5-23-16 WITH NuSTAR: I- RELATIVISTIC REFLECTION AND CORONAL
PROPERTIES
Abderahmen Zoghbi1, G. Matt2, J. M. Miller1, A. M. Lohfink3, D. J. Walton4, D. R. Ballantyne5, J. A.
Garc´ıa6,9, D. Stern4, M. J. Koss7, D. Farrah8, F. A. Harrison9, S. E. Boggs10, F. E. Christensen11, W. Craig10,
C. J. Hailey12, W. W. Zhang13
1DepartmentofAstronomy,UniversityofMichigan,AnnArbor,MI48109,USA
2DipartimentodiMatematicaeFisica,UniversitadegliStudiRomaTre,viadellaVascaNavale84,I-00146Roma,Italy
3InstituteofAstronomy,UniversityofCambridge,MadingleyRoad,CambridgeCB3OHA,UK
4JetPropulsionLaboratory,CaliforniaInstituteofTechnology,Pasadena,CA91109,USA
5CenterforRelativisticAstrophysics,SchoolofPhysics,GeorgiaInstituteofTechnology,Atlanta,GA30332,USA
7 6Harvard-SmithsonianCenterforAstrophysics,60GardenStreet,Cambridge,MA02138,USA.
7InstituteforAstronomy,DepartmentofPhysics,ETHZurich,Wolfgang-Pauli-Strasse27,CH-8093Zurich,Switzerland
1
8DepartmentofPhysics,VirginiaTech,Blacksburg,VA24061,USA
0 9SpaceRadiationLaboratory,CaliforniaInstituteofTechnology,Pasadena,CA91125,USA
2 10SpaceScienceLaboratory,UniversityofCalifornia,Berkeley,California94720,USA
11DTUSpace. NationalSpaceInstitute,TechnicalUniversityofDenmark,Elektrovej327,2800Lyngby,Denmark
n 12ColumbiaAstrophysicsLaboratory,ColumbiaUniversity,NewYork,NewYork10027,USAand
a 13NASAGoddardSpaceFlightCenter,Code662,Greenbelt,MD20771,USA
J Draft version January 11, 2017

9
ABSTRACT

] MCG–5-23-16 was targeted in early 2015 with a half mega-seconds observing campaign using
E
NuSTAR. Here we present the spectral analysis of these datasets along with an earlier observa-
H
tion and study the relativistic reflection and the primary coronal source. The data show strong
. reflection features in the form of both narrow and broad iron lines plus a Compton reflection hump.
h
A cutoff energy is significantly detected in all exposures. The shape of the reflection spectrum does
p
not change in the two years spanned by the observations, suggesting a stable geometry. A strong
-
o positive correlation is found between the cutoff energy and both the hard X-ray flux and spectral
r index. The measurements imply that the coronal plasma is not at the runaway electron-positron pair
t
s limit, and instead contains mostly electrons. The observed variability in the coronal properties is
a driven by a variable optical depth. A constant heating to cooling ratio is measured implying that
[
there is a feedback mechanism in which a significant fraction of the photons cooling the corona are

1 due to reprocessed hard X-rays.
v Subject headings: AGN, X-ray reflection, X-ray coronae, Individual: MCG–5-23-16.
9
0
3 1. INTRODUCTION An additional difficulty was the requirement of simul-
2 X-ray emission from AGN provides an excellent probe taneous low energy (< 10 keV) coverage that was not
0 always available. Lubin´ski et al. (2016), for instance,
of the immediate vicinity of the central black hole. The
. compare results from different published studies with
1 spectra of Seyfert galaxies at hard energies (> 10 keV)
BeppoSAX,RXTE/HXTE,Swift-BATandINTEGRAL,
0 arecharacterizedbyapowerlawthatrollsoverexponen-
with lower ene,rgy coverage provided by several other
7 tially at energies around ∼ 100−200 keV, along with
instruments (mostly non-simultaneous), and found that
1 a Compton reflection component. The powerlaw is the
: main X-ray source, produced by Comptonization of soft estimatesofthespectralindex, thestrengthofreflection
v Randthecutoffenergyshowedsigni,ficantdifferencesin
seed photons likely produced by the viscous dissipation
i valuesandcorrelationsinthepublishedwork,evenwhen
X of accretion energy (Haardt and Maraschi 1991). Non-
analyzing the same datasets, mostly because of the low
r thermal Comptonization is not significant based on the
signal to noise ratio spectra above 30 keV.
a detection of hard X-ray cutoffs and the non-detection of
The launch of NuSTAR (Harrison et al. 2013) offered
γ-rays (Zdziarski et al. 1995; Massaro et al. 2016).
a breakthrough for these studies. With its effective area
The best spectra prior to NuSTAR were provided by
and continuous energy coverage that includes the iron
CGRO-OSSE (Johnson et al. 1993), BeppoSAX (Boella
line at 6 keV and the Compton hump peaking at ∼ 30
et al. 1997) and INTEGRAL (Winkler et al. 2003).
keV, NuSTAR provides a better handle on obtaining ac-
Modeling of the spectra obtained using these missions
curateestimates(orstringentlimits)oftheComptoniza-
(Zdziarski et al. 1993; Petrucci et al. 2001; Perola et al.
tionparameters. Manyestimateshavebeenpublishedso
2002; Molina et al. 2013; Malizia et al. 2014) provided
farforindividualSeyfertandradiogalaxiesincludingIC
someestimatesofcutoffenergiesandreflectionfractions,
4329A (Brenneman et al. 2014), Ark 120 (Matt et al.
but offered only weak constraints on the physical mod-
2014), MCG-6-30-15 (Marinucci et al. 2014b), SWIFT
els, mostly because the quality of the data was not high
J2127.4+5654(Marinuccietal.2014a),NGC4945(Puc-
enoughtobreakmodelingdegeneraciesbetweenthespec-
cetti et al. 2014), MCG–5-23-16 (Balokovi´c et al. 2015),
tral index, the reflection strength and the cutoff energy.
NGC2110(Marinuccietal.2015),NGC5506(Mattetal.
2015), NGC 4151 (Keck et al. 2015), NGC 7213 (Ursini
[email protected]
2
et al. 2015b), 3C 273 (Madsen et al. 2015a), NGC 5548 Satellite Obs. ID UTDate Exp. (ks)
(Ursini et al. 2015a), 3C 390.3 (Lohfink et al. 2015), NuSTAR 60001046002(N2) 2013-06-03 160
3C 382 (Ballantyne et al. 2014) and MRK 335 (Wilkins 60001046004(N4) 2015-02-15 210
60001046006(N6) 2015-02-21 98
et al. 2015; Keek and Ballantyne 2016). Cutoff energies
60001046008(N8) 2015-03-13 220
in the range 100−500 keV are measured, and although
Suzaku 700002010(S0) 2005-12-07 191
these values are outside the energy band of NuSTAR
708021010(S1) 2013-06-01 319
(3–79 keV), the fact that the spectra starts rolling over
708021020(S2) 2013-06-05 221
well below the cutoff energy, and the effect of the cutoff
Swift 0008042100[SW] 2015-01-28 ∼2
on the reflection spectrum that cannot be mimicked by withSW=2–11 to each
other parameters, both allow accurate measurements of 2015-03-13
the cutoff energy (Garc´ıa et al. 2015).
There are many questions to address given these new Table 1
Asummaryoftheobservationsusedinthiswork.
observations. For instance, what is the geometry of the
corona and what physical process controls the shape of et al. 2004) and Suzaku (Mitsuda et al. 2007) to obtain
the observed spectrum?. Population synthesis models a complete spectral picture between 1-79 keV. Section 3
for the X-ray background rule out significant emission presentsdetailedspectralmodeling,firstfortheironline
above ∼ 300 keV (Gilli et al. 2007; Ueda et al. 2014). band and then for the whole observed band. The impli-
Therefore,acutoffshouldbeobservableinmanysources. cations of the results are discussed in section 4. Results
Additionally, depending on its compactness, the plasma from the short time-scale variability will be published
cannotreachequilibriumatveryhightemperatureswhen separately (Zoghbi et al. in prep.).
photon-photoninteractionsbecomeimportant. Thepro-
duction of electron-positron pairs acts as a thermostat, 2. OBSERVATIONS&DATAREDUCTION
whereincreasingthepoweroftheplasmaproducesmore
2.1. Observations Log
pairs rather than increases their temperature (Svensson
1984; Zdziarski 1985). Obtaining a measure of the elec- Following the detection of both a cutoff energy
trontemperaturealongwithanestimateofthesizeofthe (Balokovi´c et al. 2015) and reverberation time delays
corona is important in assessing to what extent these ef- (Zoghbi et al. 2014) in the first observation of 2013,
fects are important and help constrain the geometry of MCG–5-23-16 was observed again with NuSTAR for a
the corona. net exposure of 528 ks in early 2015. The new obser-
CombiningenergycutoffmeasurementsfromNuSTAR vation was split into three exposures, with the first two
and size estimates from both spectral and timing infor- separated by 2 days, and the third taken 22 days later.
mation, Fabianetal.(2015)foundthattheimpliedelec- The observation ID’s, dates and exposures are shown in
trontemperaturesareclosetotheboundaryoftheregion Table1. OuranalysisalsoincludesdatatakenbySuzaku
inthecompactness-temperaturediagramwhichisforbid- in2013,simultaneouslywiththe2013NuSTAR observa-
den due to runaway pair production. This suggests that tionandalsoanearlierSuzaku observationtakenin2005
pairs are an important ingredient in AGN coronal plas- (see Table 1). Timing analysis of the 2013 Suzaku data
mas. was presented in Zoghbi et al. (2014), while the spectral
Theoreticalmodelsprovideadditionalpredictionsthat analysis is presented here for the first time.
can be tested observationally. For example, in pair-free In order to obtain spectral coverage below 3 keV si-
Compton-cooled coronae, an increase in cooling (keep- multaneous with the new NuSTAR observations, we re-
ing the power supplied to the electrons fixed) causes the quested snapshots with Swift XRT. A total of nine ex-
temperature to drop, so E , the cutoff energy is anti- posures were taken while NuSTAR was observing the
c
correlatedwithΓ, thephotonindex. Ontheotherhand, source, in addition to one that was simultaneous with
inpair-dominatedplasmas,GhiselliniandHaardt(1994) the first NuSTAR observation. The IDs of these obser-
found that E is positively correlated with the observed vations are also shown in Table 1.
c
photon index Γ for electron temperatures T < m c2
e e 2.2. Data Reduction
(where m is the electron mass and c is the speed of
e
light),whilethereversetrendispredictedaboveit. Even NuSTAR data were reduced and analyzed using the
belowthislimit,E canremainconstantfordifferentval- NuSTAR data analysis software, which is part of hea-
c
ues of Γ(e.g. Fig. 14b in Zdziarski et al. 2002). These soft v6.19. The data were reduced by running the
argumentsapplytoasinglesource,andshouldbeobserv- standardpipelinenupipeline. Sourcespectrawerethen
able in a sample of sources, as has been explored with extracted for modules A and B from circular regions 3
BeppoSAX and INTEGRAL (Lubin´ski et al. 2016), and arcmin in radius centered on the source. Background
is now being revisited with NuSTAR (Tortosa et al. in spectra where extracted from source-free regions of the
prep.). same size near the source. For the calibration files, we
In this work, we use a long-look observation of use caldb release v20160502. In the spectral analysis,
MCG–5-23-16 (z = 0.0085; M = 107.9M , Ponti et al. spectrafrommodulesAandBarefittedsimultaneously,
(cid:12)
2012) to attempt to address the question of cutoff vari- allowing for a multiplicative constant between them.
abilityanditsrelationtootherparametersdirectlyusing The XIS spectra from Suzaku were reduced also using
a single bright object. We find that the cutoff energy the relevant software in heasoft v6.19. The initial re-
is variable and shows interesting correlations with other duction was done using aepipeline, using the caldb
parameters. The rest of this paper is organized as fol- calibration release v20160607. Source spectra were ex-
lows: Section2presentsthedatareductionandanalysis, tracted using xselect from circular regions 3 arcmin
where we include data from NuSTAR, Swift (Gehrels in radius centered on the source. Background spectra
3
)10
1
-s
-2m -1V)
c7.5 e
erg -1s k0.05
x ( -2m
V Flu 5 ot. c
ke NuSTAR ph
-10 2.5 Swift XRT (E) ( N2 N8
3 F
1250 1270 1860 1880 1900 1920 2E N4 S0
N6 S1
Time (MJD-55197)
0.02
Figure 1. The long term light curve from the Swift monitor- 2 5 10 20 50
ing (blue circles) along with the NuSTAR fluxes (red diamonds).
Energy (keV)
Fluxes are obtained by fitting a powerlaw model to the 3-10 keV
band. TheS1-2datafromSuzaku havesimilarfluxestotheSwift
pointshown. Figure 2. UnfoldedE2F(E)plotforarepresentativesubsetofthe
spectraofMCG–5-23-16fromtheSuzakuandNuSTARcampaigns.
ThenomenclatureofthelabelsisdefinedinTable1.
wereextractedfromasource-freeregionofthesamesize,
awayfromthecalibrationsource. Theresponsefileswere the following spectral groups are fitted with the same
generated using xisresp. Spectra from XIS0 and XIS3 model, allowing for a multiplicative constant to account
werecheckedtobeconsistentandthencombinedtoform forcrosscalibrationbetweeninstruments: [S0],[S1],[S2],
thefront-illuminated(FI)spectra. Comparingfront-and [N2(A,B), SW8], [N4(A,B), SW3], [N6(A,B), SW6] and
back-illuminated spectra by fitting an absorbed power- [N8(A,B), SW9]. We do not use SW2 as most of the
lawregionbetween3and5keVgivesanindexthatvaries counts are lost when correcting for pileup, removing its
by ∼5%. Energies between 1.8 and 2.3 keV are ignored abilitytoconstrainthecolumndensitybelow3keV.The
due to the calibration uncertainties associated with the cross calibration offset between the two NuSTAR mod-
CCD Si K edge. Data from other Suzaku instruments ules is < 4% is all cases, consistent with Madsen et al.
were not used because of low signal. (2015b).
The Swift XRT observations were taken in the win-
dowed timing mode except for SW2 and SW11 which 3. SPECTRALMODELING
were in imaging mode. The data were reduced using All the spectral modeling is performed in xspec v.
xrtpipeline. Sourcespectrawereextractedfromcircu- 12.9.0n (Arnaud 1996). The uncertainties quoted for
larregionsof30pixelradius(71arcsec),andbackground the parameters are 1σ confidence, corresponding to
spectra from similar regions away from the source. Ob- ∆log(Likelihood)=0.5 (or ∆W =1 where W is the W-
servations SW2 and SW11 were taken in imaging mode statistic used in xspec), unless stated otherwise. The
andsufferedsomepileup. Therefore,thespectrawereex- Galactic column in the direction of MCG–5-23-16 is
tracted from regions that excluded the central 3 pixels. N =9×1020cm−2 (Kalberla et al. 2005), and it is in-
H
We used the Swift caldb release v20160121. cluded in all subsequent fits using the model tbabs. We
Spectral channels from NuSTAR, Suzaku XIS and startthissectionbyshowingamodel-independentrepre-
Swift were grouped to have a minimum of one count per sentation of the data, then focus on the iron line first by
bin, and we use Poisson likelihood maximization in the using phenomenological models to track the long term
modeling. BackgroundspectraarehandledusingtheW- variability, then use a physical model in section 3.1.2.
statistics (Arnaud 1996), where the background counts We extend the analysis to the whole NuSTAR band in
for each bin are considered fit parameters which can be section 3.2.
solved for analytically as a function of otherparameters. Figure 2 shows the unfolded spectrum of
We use Poisson likelihood in order to be able to fit the MCG–5-23-16. Data from Suzaku observations S0
swiftlowenergyspectraandexploitthefullresolutionof and S1 (see Table 1), along with the four NuSTAR
NuSTAR at energies above 50 keV to constrain the high observations, plotted after factoring out the effective
energy turnover. Using a Gaussian likelihood requires area of the detectors. The spectra have been re-binned
the channels to be grouped, which effectively removes to a minimum signal to noise ratio of 6 for display
some energy-dependent information at those energies. clarity. The spectrum of MCG–5-23-16 is character-
ThelongtermlightcurvefromtheSwift monitoringis ized by strong iron K emission and a broad excess
shown in Figure 1. The flux changes by a factor of ∼ 3 at 30 keV, characteristic of a reflection spectrum.
onafewdaystimescale,withthefirst2013(N2)observa- MCG–5-23-16 is seen through a Compton-thin absorber
tion having the highest flux. The NuSTAR observations (N ∼ 1.4 × 1022cm−2), and little emission from the
H
sampletheupperhalfofthefluxvariationsobservedwith nuclear regions escapes below 1 keV. The observed
Swift. Given this variability and the simultaneity of ob- spectrum at these energies (below 1 keV), as revealed
servations, and for the purpose of constraining any vari- by XMM-Newton RGS spectra, is dominated by several
able column density at the host galaxy, we simultane- emission lines superimposed on an unabsorbed scattered
ously fit observations that are taken within a day or so, power-law continuum (Braito et al. 2007). They orig-
or those having roughly the same 2–10 keV flux, model- inate in a plasma in the Narrow Line Region. In the
ingtheNuSTAR andSuzaku dataseparately. Therefore, following Suzaku and Swift spectral modeling, we ignore
4
0.1 1.9 Γ lloogg((NNpp)) −1.4 7 EEnn lloogg((NNnn)) −3.8
1) N2 −1.5 6.75 −4
-eV0.08 N4 1.8 6.5 −4.2
-1s k N2-N4 1.7 −1.6 6.265 −4.4
0.06
ts 1.6 −1.7
un 1.6 nnHH σbb 7 EEbb lloogg((NNbb)) −3.6
co0.04 0.5 6.75
( 1.4 −3.8
el 0.45 6.5
od0.02 1.2 6.25 −4
m 0.4
- 6 −4.2
a
at 0 1234567 1234567 Observations1234567 1234567
d
Figure 4. Changes in the iron line complex inferred by using a
−0.02 powerlaw plus a narrow and broad Gaussian lines. Each tick in
5 10 the x-axis is a single observation ordered as: S0, S1, S2, N2, N4,
N6 and N8 respectively. The name of the parameter is shown in
Energy (keV)
the top right of each panel. Γ is the photon index. Np is the
Fexitgruemrees3.inTfluhxe),irownithlintehefriromdiffthereenNc2e.anTdheN4spoebctsrearvaarteionpslot(ttehde npanoorwdmebralrlaoizwaadtniooGrnamsu.asσlsiizaanitsioltinhn.eesEwrniedsatphnedcotfEivtbhealeyr,beratonhadedeNcnonemrgapineosdnoeNfnttb.haerneatrhroewir
aftersubtractingapowerlawmodelfittedtothe3–4and8–10keV b
bands. in section 2. The resulting parameter changes are shown
in Figure 4.
spectral energies below 1 keV since they are not part of
NuSTAR observation N2 was simultaneous with
the nuclear emission.
Suzaku S2, but there are clear systematic differences be-
tween them, both in the absolute flux values and in the
3.1. The Fe Line Complex
photonindices. Theseareduetoabsolutecalibrationun-
ThespectraofMCG–5-23-16attheironenergiesshow certainties between the two detectors and the uncertain
both a narrow and a broad components (Reeves et al. crosscalibrationbetweenthefront-andback-illuminated
2007). The goal of this section is first to investigate detectors in Suzaku. The photon index differences are
whether the broad and narrow components are variable, the main cause of the large difference in N between S2
H
and second to model the broad line with relativistic re- and N2 (which are degenerate in the fits). Nonetheless,
flection models. All fits in this section are done in the it is clear that most of the variability is in the powerlaw
1–10 keV band. continuum flux and its photon index. The energy of the
narrowcomponentisconsistentwithaconstant. Infact,
3.1.1. Long Term Variability a model where the line energy is fixed to the average is
Wefirstattempttocheckforcolumndensityvariability statistically preferred over a model where it is free to
usingtheSwift dataalone. Wefitanabsorbedpowerlaw change.
toalltheSwift spectraandtracktheirchanges. Thecol- We assess the significance of the parameter changes
umn density showed some changes. However the strong usingthesample-correctedAkaikeInformationCriterion
degeneracy between column density and powerlaw index (cAIC; Akaike 1974; Burnham et al. 2011), where we
does not allow any firm conclusions to be drawn, and compare models in which the parameter of interest is
therefore analysis of all the spectral data is required. either fixed or allowed to change between observations.
To explore the changes in the iron line complex us- For the narrow line, we find that fixing the line flux be-
ing a phenomenological approach, we show in Figure 3 tween observation gives a lower cAIC than allowing it to
the spectrum of the iron line from the highest and low- change, with ∆cAIC = −2 from each observation. This
est flux NuSTAR observations (N2 and N4, respectively, impliesthatthemodelwithafixedlinefluxispreferred.
which are separated by almost two years). The differ- Forthelineenergy,∆cAIC<0fortheSuzaku-onlydata
ence between the spectra is also shown. The spectrum and ∆cAIC > 10 when the NuSTAR data are included.
in the iron line complex is clearly variable, and most of In other words, the line appears to change only between
the variability is in the low energy wing of the line, not differentinstrumentsbutisconstantwithinthesamein-
thecore. ThecentroidenergyfortheN2andN4spectra strument. This suggests that the line flux is physically
are 6.30±0.01 and 6.36±0.01 keV, respectively, while constant and the observed changes are caused by instru-
the centroid from the difference spectrum is 6.17±0.05 mentabsolutecalibration. Forthecolumndensity, there
keV.Itappearsthereforethatthestrongestchangesinthe appears to be significant changes between observations,
iron line are in the broad component and not the narrow even when using the same instrument (∆cAIC>15 cor-
component. respondingtoa>3.3σ significance). Changesinthepa-
To investigate this in a more systematic way, we fit rametersofthebroadcomponentaresignificantatthe3σ
the 1–10 keV band of the spectral groups using a pow- level when the narrow component is assumed constant.
erlaw, a narrow and a broad Gaussian line. The width Degeneracies between the two reflection components are
of the narrow line is fixed at the instrument resolution addressedinsection3.1.2whereweusefullphysicalmod-
as neither NuSTAR nor Suzaku data are able to resolve els.
it. Constraints to the column density for the NuSTAR
3.1.2. Relativistic Model
data are provided by including Swift data, as discussed
5
Obs NH Γ R θ Rin q log(ξ) Nr Nx Ec
Fitstoindividualspectrabetween1–10 keV
S0 1.49(1) 1.84(3) 0.3(1) 57(7) <28 2(1) 2.6(1) -3.25(1) -3.38(6) ...
S1 1.41(1) 1.89(2) 0.2(1) 39(8) <40 3(1) 2.6(1) -3.10(1) -3.4(1) ...
S2 1.41(1) 1.90(1) 0.21(3) 47(7) 77(20) >6 2.70(5) -3.19(1) -3.4(1) ...
N2 1.24(5) 1.91(1) 0.29(8) 34(7) 41(18) 5(2) 2.3(1) -3.18(1) -3.6(1) ...
N4 1.26(5) 1.83(2) 0.5(1) 64(6) <41 <7 2.5(9) -3.34(1) -3.5(1) ...
N6 1.46(5) 1.88(2) 0.4(2) 52(13) <83 2(1) 2.3(8) -3.29(1) -3.6(5) ...
N8 1.34(5) 1.84(1) 0.29(7) 45(8) <33 3(2) 2.3(8) -3.26(1) -3.5(1) ...
Fitstoindividualspectrabetween1–79 keV
N2 1.39(1) 1.87(1) 0.2(1) 51 >50 8(2) 2.5(5) -3.23(1) -3.45(1) 140(10)
N4 1.26(4) 1.77(1) 0.34(3) ... <68 1(1) 2.7(1) -3.41(2) -3.84(5) 120(8)
N6 1.41(5) 1.82(1) 0.25(7) ... <59 1(2) 2.5(1) -3.33(1) -3.7(1) 146(18)
N8 1.35(4) 1.77(1) 0.18(3) ... <73 0.2(7) 2.7(1) -3.33(1) -3.66(5) 124(7)
Jointfittoallthespectrabetween1–79 keV
NH Γ Np θ Rin q log(ξ) Nr Nx Ec
N2 1.40(1) 1.85(1) -1.47(1) 51 47(23) 1.6(6) 2.66(3) -3.80(1) -3.60(2) 152(8)
N4 1.25(4) 1.78(1) -1.68(1) ... ... ... ... -4.21(6) ... 107(7)
N6 1.45(4) 1.83(1) -1.59(1) ... ... ... ... -4.12(7) ... 149(14)
N8 1.41(4) 1.79(1) -1.60(1) ... ... ... ... -4.11(1) ... 130(6)
Table 2
Fitparameterfortherelxill+xillverforthe1-10keVfitsfortheSuzaku andNuSTAR datasets,andfor1-79keVfortheNuSTAR
datasets. The1σ statisticaluncertaintyinthelastsignificantfigureisshowninbrackets. N isthenormalizationinunitsoflog(photons
cm−2 s−1). Subscriptsr,pandxareforrelxill,cutoffplandxillverrespectively. Risthereflectionfractionmeasuredastheratioof
directtoreflectedfluxesbetween20–40keV.Rin isinunitsofgravitationalradiusrg andtheinclinationθ isindegrees. q isthe
emissivityindex,ξ istheionizationparameter,ΓisthephotonindexandEc isthecutoffenergy. Theobservationsinthefirstcolumnare
definedinTable1.
The narrow Gaussian line is due to distant reflec- whenfittingthewholeNuSTARbandinsection3.2. The
tion from the Broad Line Region or the torus. To abundance of the inner and outer reflectors are assumed
model it more physically, we replace the narrow Gaus- the same. We fit all seven spectra independently, and
sian with the reflection model xillver (Garc´ıa et al. the results are summarized in Table 2.
2014). xillver includes self-consistently emission from The reflection parameters are generally better con-
Fe Kβ and allows for the reflection to be ionized. Us- strained in the Suzaku spectra. This is because the res-
ing the xillver+powerlaw model accounts for most of olution of NuSTAR in the iron K band does not allow
the residuals around 6.4 keV, including the Fe Kβ line the narrow and broad components of the line to be un-
present in MCG–5-23-16, also seen in previous observa- ambiguously separated. One effect of this is that the
tions(Reevesetal.2007). Thexillvermodelprovidesa values of the iron abundance are not consistent between
better fit compared to pexmon (Nandra et al. 2007), for Suzaku and NuSTAR. The three Suzaku spectra sug-
example, with ∆W ∼ 28 per observation for the same gestedavalueofabout1, while itiswasaround0.5(the
number of degrees of freedom for Suzaku spectra where model minimum) in the NuSTAR spectra. We there-
the Kβ line is most clearly seen. This corresponds to fore fixed the NuSTAR value at the Suzaku value. The
∆cAIC = 17 and a significance of > 3.5σ. The best uncertaintiesinTable2arecalculatedfromMonteCarlo
fit ionization parameter of the reflector in this case is MarkovChains(MCMC)asthe1σstandarddeviationof
log(ξ)=0, consistent with neutral reflection1. thechains. Wefoundwhenexploringthelikelihoodspace
Continuing the analysisof the spectra in the1–10 keV thatitismulti-modal,withmultipleparametercombina-
band, we next model the broad component of the iron tions having close likelihood values near the maximum.
line with a full relativistic reflection model. We use MCMC is therefore well suited to explore this multi-
the relxill model (version v0.4a) (Dauser et al. 2010; modality. All the chains reported here were generated
Garc´ıa et al. 2014). The reflection spectrum is a result using the affine invariant ensemble sampler (Goodman
of a hard X-ray source illuminating a constant density and Weare 2010). All the chains were run several times
disk. The observer sees emission from both the illumi- andlongenoughtoensureconvergence. Theconvergence
nating source and the reflector. The reflection spectrum is assessed with both the autocorrelation of the chains
isconvolvedwitharelativistickerneltomodelthestrong and the stability of the chain variances.
gravityeffectsoftheblackhole. Aswewillshow, thein- Table2quotesonlytheaveragevalues. Therearehow-
ner radius of the disk is > 10r , and because the inner ever two general solutions in the modeling with different
g
radius is degenerate with the black hole spin, we fix the values for the ionization parameter ξ of the relativistic
latter at the maximum. The fact that the inner radius reflectioncomponent(log(ξ)∼0andlog(ξ)∼2.7). This
is > 10r , the exact fixed spin value has little effect on is illustrated in the top panel of Figure 5, where we plot
g
thefits. Weuseasinglepowerlawemissivityprofile, and the confidence contours for the best fit parameters for
assume that the directly observed component has the (the log of) the ionization parameter versus the photon
same spectral index as that illuminating the disk. The indexforalltheobservations. Althoughthebestfitvalue
high energy cutoff is fixed at 300 keV, and it is modeled for the ionization parameter is log(ξ) ∼ 2.3−2.7, lower
values(≤1)arealsopossiblewithaslightlyhighervalue
1 ξ insubsequentdiscussionsisinunitsofergs−1 cm−1 to Γ. Two parameters which are of interest are Rin and
6
3 S0 S1 S2 N2 N4 N6 N8 ofindividualcomponentsandallowforthecutoffenergy
to be a free parameter. The model has an xspec form:
ξ)2 tbabs*ztbabs* (relxill+cutoffpl+xillver). Allow-
g(
o ingallparameterstobefreeshowedthesamestrongcor-
l1
relations between the inclination and distant reflection
0 flux discussed in section 3.1.2, with the highly inclined
1.81.851.9 1.8 1.9 1.8 1.9 1.8 1.9 1.8 1.9 1.8 1.9 1.8 1.9
Γ diskbeingthebestfit. Becausethisisphysicallyunlikely,
we fix the inclination at θ = 51◦, the weighted average
100
from modeling the Suzaku spectra in section 3.1.2. We
75
found that the exact value for the inclination does not
Rin50 affect the following results significantly. The main effect
is that a lower (higher) value causes the distant reflec-
25
tor to be weaker (stronger), as already suggested by the
2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 bottom panel of Figure 5.
q
The best fit parameters from modeling the whole
NuSTAR band are shown in Table 2. The best fit model
6 and residuals are shown in Figure 6. The residuals in
the middle set of panels in Figure 6 have been binned
Nx
4 so that the signal to noise ratio of every spectral bin is
at least 6. This is done so the residual plot is meaning-
2
ful. The residuals at the high energy part of the spectra
30 50 70 30 50 70 30 50 70 30 50 70 30 50 70 406080 30 50 70
θ are shown in the bottom set of panels, plotted as the
residualsofintegrated(unbinned)datatotheintegrated
Figure 5. Bestfitconfidencecontoursfortheinnerradiusofthe model. This is a convenient way of plotting to account
disk and emissivity profile of the disk as measured from relxill forthefactthatbinsattheseenergieshavesmallnumber
when fitted the spectrum below 10 keV. The top panel shows the
of counts. The bottom set of panels show that the devi-
confidence contours for the ionization parameters versus photon
index,themiddlepanelisfortheinnerradiusversustheemissivity ations between the model and the data are comparable
index, and the bottom panel is for normalization of the xillver tothedeviationsbetweenthetwoNuSTAR modulesdue
componentversustheinclinationangle. to counting noise or cross calibration uncertainties.
the emissivity index, which locate the emitting region The goodness-of-fit statistic is estimated using Monte
relative to the central object. These two parameters are Carlo simulations. We start from the best fit parame-
highly correlated in the modeling as shown by the mid- ters and generate a large number of parameters drawn
dle panel of Figure 5. Although the best fit values are fromtheMCMCchains. Foreachparameterset, aspec-
at Rin ∼20 gravitational radii (rg =GM/c2) and q ∼3, trumisfakedusingfakeitinxspec,takingintoaccount
lowerandhighervaluesforRin arealsosupportedbythe counting noise. The faked spectra are then refitted with
data. the model and a distribution of fit statistics is produced
The confidence contours of the inclination θ and the from the resulting fits. The fraction of simulated data
normalization of the distant reflector Nx are shown in thathaveafitstatisticthatisatleastasgoodastheob-
the bottom panel of Figure 5. The best fit inclination served value are 0.96, 0.59, 0.21 and 0.4 for observations
value in this case is θ ∼ 50◦, but higher inclinations N2,N4,N6andN8,respectively. Thesegoodnessparam-
with stronger distant reflection are also supported by eters correspond to the probability of rejecting the null
the data. The data in this case supports either a low hypothesiscorrespondingtothebestfitmodel. Thesefits
inclination disk (θ < 60◦) with a relatively weak dis- areverygoodgiventhehighqualitydata. Thehighvalue
tant reflector or a highly inclined disk (θ ∼ 88◦) and fortheN2-S2combinationisduetocrosscalibrationun-
strongerreflection. Giventhatthisobjectisseenthrough certainties between NuSTAR and Suzaku and between
a Compton-thin rather than thick absorber, suggests an the front- and back-illuminated detectors in Suzaku.
intermediate inclination, making the first solution more Most of the parameters remained similar to those in
physically plausible. We note here that parameters re- section 3.1 and Figure 5. R , q and the parameters of
in
ported in the analysis of the first NuSTAR observation xillverwereallconsistentwithbeingconstantbetween
(Balokovi´c et al. 2015) are consistent with one of the observations. Therefore, and in order to obtain further
local minima in the fit, with log(ξ) ∼ 0. Our best fit constraints on the variable parameters, we fit all four
has a value of log(ξ) ∼ 2.7. Other parameters change NuSTAR observations (and the matching Suzaku and
accordingly, with reflection parameters not differing sig- Swift datasets) together, allowing only parameters that
nificantly (apart from the normalization). showed variability in the individual fits to vary. These
parameters are: N , Γ, E and the normalizations of
3.2. Full Band Relativistic Model H c
relxill (N ) and cutoffpl (N ). The best fit parame-
r p
Here, we extend the analysis to higher energies (up ters in this case are presented in Table 2.
to 79 keV) and focus on the full NuSTAR data. Col- The column density appears to change between obser-
umn density constraints are provided by Swift XRT for vations in a way that is not directly related to the ob-
observations N4, N6 and N8, while for N2 we use the served flux. The variability is however only marginally
Suzaku observation S2 as it is of higher quality and it is significant. The 99.5% confidence limits on N are con-
H
simultaneous with N2. We model the spectrum with a sistent with a constant column. The remaining param-
model similar to that in section 3.1.2, except that we fit eters change significantly between observations. The re-
for a powerlaw explicitly so we can track flux variations sultsoftheirvariabilityissummarizedinFigure7,which
-1V) 7
e
k
-1s N2 N4 N6 N8
-2m 0.1
c
2
V
e
k
) ( 0.01
E
(
F
2
E
del 10−3
o
M
10 100 10 100 10 100 10 100
N2 N4 N6 N8
or 5
r
r
e
/
el) 0
d
o
m
-
a
t −5
a
d
(
1 10 1 10 1 10 1 10
)
el
d N2 N4 N6 N8
dmo 10-3
ted
ae
grat 0
er
tg
ne
als (i- int -10-3 FFMMPPAB
us
sidunt 50 70 50 70 50 70 50 70
eo
Rc Energy (keV)
Figure 6. Top: ThemodelsfittedtothefourNuSTARepochs. Themodelconsistsofapowerlaw(blue,dashed),distantreflection(green,
dotted) and relativistic reflection (red, dot-dashed), with their sum shown in solid black. Middle: Fit residuals produced after rebinning
the spectra. Red and blue colors are from the two NuSTAR modules, and the green is the low energy spectra from Suzaku (for N2) and
Swift (for N4, N6 and N8). Bottom: Residuals of integrated counts relative to the integrated model, a convenient way of visualizing the
residualwhenthenumberofcountsperbinissmall.
N2 r = 0.90(6) r = 0.0(3) r = 0.7(1)
N4
200 N6
N8
Ec 150
100
1.75 1.8 1.85 0.2 0.25 0.3 −10.2 −10.1 −10
Γ R log(PL Flux)
Figure 7. Changes in the parameters describing the primary X-ray continuum. The contours are the 1, 2 and 3 σ confidence intervals
foreachparameter. Risthereflectionfraction. Thesolidlineintheleftandright-mostpanelsisthebestfitlinearmodelandthedashed
lines are the 1σ uncertainty in linear model. The Pearson correlation coefficient is quoted in the top-right corner of each panel. Ec is in
keV, the powerlaw flux Fp is in units of ergs cm−2 s−1 and is measured between 2–10 keV. The blue points show a single measurement
fromaBeppoSAX takenin1998(Perolaetal.2002). AlinearfittotheEc−ΓrelationgivesEc =(665±104)Γ+(−1067±186)anda
linearfittotheEc−Fp relationgivesEc=(267±58)Fp+(2822±586). TheobservationlabelingN2-N8isdefinedinFigure1.
8
shows the best fit parameter confidence contours for Γ, value of 3. There is, however, a strong degeneracy be-
reflectionfraction2Randthe2–10keVfluxofthepower- tween the inner radius and emissivity index parameters,
lawcomponent(F ),plottedagainstthehighenergycut- such that the data also supports (at the 99% confidence
p
off. Althoughfluxfromboththepowerlawandreflection level) a disk that extends down to R < 6r , with a
in g
components change, their ratio remains relatively con- flatteremissivityindex(q <2). X-rayreverberationlags
stant. The difference between high- and low-flux model detected in this source, taken at face value suggest that
spectrainthiscaseresemblestheshapeoftherelativistic the latter solution is preferred (Zoghbi et al. 2014).
reflection (plus a powerlaw), and it directly explains the The reflection component, producing the narrow core
observed difference spectrum between N2 and N4 shown of the iron line and a considerable fraction of the reflec-
in Figure 3. tion hump at 30 keV, appears to be constant across the
The photon index is correlated with the high energy two year period spanned by the data, unless the inclina-
cutoff E within individual spectra. This is a conse- tion of the disk changes, which is unlikely. This is not
c
quence of the model parameterization when the data surprising if emission comes from material far from the
above ∼ 50 keV have a relatively low signal to noise ra- central source, thereby smoothing out any variability. A
tio. Additionally, both Γ and the continuum flux appear consequenceofthisobservationisthatthereflectionfrac-
to be correlated with the cutoff energy when all four ob- tionfromthiscomponentvaries, evenwhenthefluxand
servationsareconsidered. Thereflectionfractionisinde- index of the illuminating source are the only parameters
pendent of cutoff energy. To quantify these correlations, that change, not the geometry of the system. A corre-
weusetheMCMCchainsalreadycalculatedtocalculate lation of the distant reflection strength R with the flux
thePearsoncorrelationcoefficientr. Wefindcorrelation in a single object is therefore naturally explained (e.g.
coefficients of 0.90(6), 0.0(3) and 0.7(1) for the relations Malzac and Petrucci 2002).
between E and Γ, R and F , respectively. The number The flux from the inner reflector on the other hand,
c p
in bracket is the uncertainty in the last significant digit tracks closely variations of the illuminating source. The
taken as the 1σ sample standard deviation. strengthofthereflection(measuredastheratiooffluxes
WeemphasizethatR inFigure7isthereflectionfrac- between 20 and 40 keV) remains constant, a result also
tion from the relativistic reflection. The reflection frac- seen in other objects (e.g. Lohfink et al. 2016). Combin-
tion from the distant reflector (not plotted), increases ing this with the fact that the column density changes
with E , driven by the F −E correlation and the fact very little, and also the lack of significant changes in
c p c
that the flux from the distant reflector is constant. Fig- the relativistic reflection parameter, gives a picture in
ure 7 also shows one measurement from a BeppoSAX whichthefluxchangesseeninthissource(e.g. Figure1)
taken in 1998 (Perola et al. 2002). That measurement are driven by intrinsic flux fluctuations in the primary
appearstofollowthesametrendsweobserve,albeitwith source, which are closely matched, on days to months
larger uncertainties. timescales, byvariationsinthe fluxofthe relativistic re-
We note here that the two, low and high ξ, solutions flection component. This seems to be the long term ex-
found when fitting data below 10 keV (section 3.1) no tension of the relativistic reverberation signatures seen
longer give a comparable goodness of fit. The higher ξ on short time-scales in this object (Zoghbi et al. 2014).
solutiongivesabetterfitby∆W >30,correspondingto
4.2. Plasma properties
a significance of >3.7σ per observation. This solution is
therefore preferred over that reported in Balokovi´c et al. Modelingthereflectionspectrumproperlyallowsusto
(2015) for observation N2. Forcing the ξ ∼ 0 solution extract information about the Comptonization process.
gives,inadditiontoaworsefit,lowercutoffenergiesbut We find that most of the variability between observa-
the correlations in Figure 7 hold. tions is due to changes in the flux of the primary power-
We also note the low N value for N4. It is un- law component. Its photon index and cutoff energy are
H
likely that the column density changes significantly in alsofoundtobesignificantlyvariable,whiletherelativis-
days time-scale. We therefore tested tying the N val- tic reflection component remains constant in shape and
H
ues between observations. We found a slightly worse fit constant in flux relative to the primary component. The
(∆W ∼ 14 or a significance of ∼ 2σ), with no large primary continuum changes in a structured manner, as
changes in E . The reason is that NuSTAR data qual- indicatedbythehighcorrelationcoefficientsbetweenthe
c
ity is much better than Swift, so forcing a new N does photon index and flux with the cutoff energy.
H
not affect the NuSTAR data significantly but makes the It is known that in the simple cutoff powerlaw model,
Swift fit slightly worse. the flux, index and cutoff energy can be correlated by
construction. This is apparent in the elongated contours
4. DISCUSSION
showninFigure7fromindividualfits. Itshouldbenoted
4.1. X-ray reflection and the inner disk however, that the correlations between observations are
The most detailed analysis of the broad component of robust, in a sense that if the parameters from one ob-
theironlineinMCG–5-23-16priortothisworkwaspre- servation are fixed at the best fit parameters of another
sentedinReevesetal.(2007). Theresultspresentedhere observation, the fit significantly worsens. Another way
regarding the reflection spectrum are consistent with to see this is to observe that in the Γ−Ec and Fp−Ec
thatanalysis. Takingthebestfitparametersfortherela- plotsinFigure7, theelongatedcontoursarenot parallel
tivisticreflectionsuggestatruncateddiskat∼40r with totheobservedcorrelations,indicatingthatalthoughthe
g
a emissivity index close to a standard non-relativistic parametersmightbecorrelatedwithin asinglespectrum,
their correlations between observations are robust.
2Whatwerefertoasreflectionfractionhereisreflectionstrength
4.2.1. Γ−E Correlation
inthenomenclatureofDauseretal.(2016). c
9
PreviousresultsonpossibleΓ−E correlationsinsam- tween Γ and the flux is well established in bright AGN
c
ples ofobjectswerenotconclusive. Piroetal.(1999)first andGalacticblackholes(e.g.SobolewskaandPapadakis
notedapossibleΓ−E relationusingtwoBeppoSAX ob- 2009; Yang et al. 2015).
c
servations of NGC 4151. Petrucci et al. (2001) found a
4.2.3. Physical Interpretation
weak relation with six Seyfert Galaxies, and a relatively
stronger relation was found by Perola et al. (2002) using Beforediscussingthedetailsoftheplasmaphysics,itis
aslightlylargersample. UsingINTEGRALdata,Molina worth mentioning the possibility that the changes in Ec
etal.(2013)reportedaweakrelationwhileMaliziaetal. may not be due to changes in the intrinsic election tem-
(2014) reported no relation. It appears therefore that as perature, but rather due to changes in the gravitational
far as a sample of AGN is concerned, there is at most a redshift of a constant spectrum (e.g. Nied´zwiecki et al.
weak relation between Γ and E . 2016). Such changes in the gravitational redshift of the
c
The data for individual objects is less clear. Mostly emittedphotons(duetochangesinthesizeofthecorona
becauseofthedifficultyofobtaininghighqualitydatain for instance) could artificially introduce variations in Ec
single epoch observations, with low energy coverage. We without the plasma properties changing. This, however,
note that from the INTEGRAL study of Molina et al. also produces changes in the reflection fraction due to
(2013), who analyzed separate observations of individ- the focusing of light rays into and away from the disk.
ual objects, in almost all cases of objects with multiple The fact that the observed reflection fraction is constant
observations, flatter spectra are accompanied by small suggeststhatthegeometrydoesnotchangesignificantly,
cutoff energies. The uncertainties in the parameters are, strengthening the interpretation in which the Ec varia-
however, large. Using NuSTAR, Ballantyne et al. (2014) tions are intrinsic to the plasma. Also, the inner radius
analyzed two observations of the radio galaxy 3C 382 in we measure is relatively large, so GR effects are present
two flux intervals. The low flux observation had a flat- butnotextreme,andthereforethediscussionofrelativis-
terspectrumandahighercutoffenergycomparedtothe ticmodelinginNied´zwieckietal.2016arenotapplicable
higher flux observation (i.e. opposite the trend seen in in this work.
INTEGRAL data and seen here in MCG–5-23-16). Also Inthesimplestconsiderationsofapurethermalplasma
using NuSTAR data, Keek and Ballantyne (2016) found (Sunyaev and Titarchuk 1980) (no electron-positron
a positive correlation between Γ and E in Mrk 335, al- pairs), an increase in the soft flux impinging on the
c
though we note that both the photon indices and cut- coronaleads tosofter Comptonizationspectraand lower
off energies (< 50 keV) found there are small, differing temperatures as the electrons are cooled efficiently (e.g.
substantially from other studies using the same datasets Zdziarski et al. 2002). This picture cannot be applied
(Parker et al. 2014; Wilkins et al. 2015). The result we directly here, first because pairs are not included and
foundheresuggestsastrongpositivecorrelationbetween their effect could be important (Fabian et al. 2015), and
the photon index Γ similar to Mrk 335. secondbecausethecorrelationwemeasureisinthehard
flux(emittedbythecorona)ratherthanthesoftfluxim-
4.2.2. Luminosity−Ec Correlation pingingonit. Therefore,wewouldlikefirsttoassessthe
Variability of the cutoff energy (or electron tempera- importance of pair production given the measurements
ture) with flux or luminosity has not been explored in we have.
detailextensivelyinAGN,unlikeblackholebinaries. As We start by comparing the observations to predictions
we pointed out in section 4.2.1, results from AGN are ofpair-dominatedplasmas. Thetemperatureofaplasma
notconclusiveyet,wherebothacorrelationandananti- cannot be arbitrarily high for a given size. The key
correlation of E and flux have been reported for 3C 382 parameter that is often used is the compactness l =
c
and Mrk 335 respectively (Ballantyne et al. 2014; Keek 4π(mp/me)(rg/r)(L/Ledd),whichmeasurestheluminos-
and Ballantyne 2016). ity to source size ratio. As the compactness increases,
ForX-raybinaries,CygX-1showedanincreaseincut- photon-photon interactions become important, and any
off energy when the luminosity in the hard X-rays drops extraheatinggoesintoproducingelectron-positronpairs
during the hard state (Gierlinski et al. 1997; Ibragimov ratherthanheatingtheplasma,causingthetemperature
et al. 2005). A strong anti-correlation of the cutoff en- to saturate (Guilbert et al. 1983; Zdziarski 1985).
ergyandluminositywasalsoobservedinGX339-4when Following Zdziarski (1985) and Ghisellini and Haardt
theluminositywasabove∼10%theEddingtonluminos- (1994),wecalculatethespectralindexα(α=Γ−1)and
ity,anditremainedconstantbelowthat(Miyakawaetal. cutoff energy predicted from models in thermal and pair
2008). AsimilarresultwasfoundbyMottaetal.(2009), equilibria, for different values of the compactness lh and
who additionally observed the reverse trend when the lh/ls,wherelh isthecompactnessoftheComptonization
source softened before transiting to the soft state. Simi- plasma (i.e. the power heating the corona), and ls is the
larbehaviorisseeninotherobjects,includingV404Cyg compactness of the soft source producing the seed pho-
in its recent outburst as observed with Fermi (Jenke tons. Weusethemodeleqpair(Coppi1999)togenerate
et al. 2016), and possibly also Cyg X-1 in the soft state spectra for a grid of parameters for lh (between 10 and
from NuSTAR observations (Walton et al. 2016). We 5×104) and l /l (between 1 and 100). We assume the
h s
note that when E is correlated with luminosity, so is soft source is a blackbody with a temperature of 10 eV
c
Γ, and when the spectra soften during the hard inter- and the plasma is spherical and contains no background
mediate state, both Γ and E reverse their dependence electron plasma (τ = 0). The generated spectra are
c p
on luminosity. The result we find for MCG–5-23-16 ap- thenfitted witha cutoffpowerlaw modeltosimulatethe
pears to match the behavior of black hole binaries not fitting procedure and obtain the energy spectral index α
in the hard state where there E is anti-correlated with andE . TheresultsareshownintheleftpanelofFigure
c c
L, but during the intermediate state. A correlation be- 8. Contours of α and E are shown, and the 1σ mea-
c
10
100
00 0.3 0.2 No Pair
8 1
0.4 00 1 Balance
0
2
0 0.5 1
0
6
2
0.6 c
e
0 1 0 m
l/lhs 10 50 400350300 25000..87180 150 120 00 8 Θ = kT/e No Pairs
0.9200 0.1
1.0
Slab data
1.1 Hemisphere R = 1.23 r
in g
1.2 Sphere
1
10 100 1000 104 1 10 100
l l
h h
Figure 8. Left: Contours of constant spectral index α = Γ−1 (solid lines) and cutoff energy Ec (in keV; dashed lines) for different
parametersofcompactnesslh andhardtosoftcompactnessratio(lh/ls). Thegreenboxesarethe1σmeasurementerrorinαandEc from
thefourobservations. ThisplotisessentiallyaconversionplotfromtheobservedΓ(orα)andEc totheplasmaparameterslh andlh/ls
usingtheeqpairmodelandassumingaplasmainpairequilibrium. Right: Themaximumtemperature(inunitsofmec2)thatcanreached
byaplasmadominatedbyrunawaypairproductionforthreegeometries. Upperlimitstol andmeasurementsofΘareshown.
h
suredvaluesofαandE areshownwiththegreenboxes. off measurements. Using the nthcomp (Zdziarski et al.
c
This plot shows that, first, the inferred compactness ra- 1996) model (as implemented in the reflection model
tiol /l ∼5−6,whichisnotatypicalofAGN.Secondly relxillCpwhichcalculatesthereflectionspectrumwhen
h s
and most importantly, it shows that, if the plasma is illuminated by nthcomp; Garcia et al. in prep.) instead
dominated by pairs, then the inferred l is large (for ref- ofthecutoffplshiftstheelectrontemperaturesuponly
h
erence, a source radiating at the Eddington limit that is by ∼10%, and our conclusion about the plasma content
1 gravitational radius in size has l ∼ 105). Therefore, is not altered. The points can of course be shifted to
based just on the the measured α and E , if the plasma the right (i.e increasing l ) if the black hole mass is er-
c h
isdominatedbypairs,thesourcehastobeverycompact roneous. We find that in order for the plasma to be in
suggesting a small size and/or high luminosity. pair balance for the slab geometry, the black hole mass
Furtherinformationisprovidedbyincludingthesource needs to be smaller by ∼ two orders of magnitude (i.e.
size measurement we have from the reflection spectrum M ∼ 106M ). We note that the black hole mass in
(cid:12)
and the observed luminosity. The results in this case Ponti et al. (2012) is uncertain. A mass estimate using
are shown in the right panel of Figure 8, showing the the fundamental plane of black hole activity (Gu¨ltekin
temperature-l relations at the pair limit from the mod- et al. 2009) using the radio fluxes from (Mundell et al.
eling of Stern et al. (1995) for three geometries of the 2009)andtheX-rayfluxesfromourobservationssuggest
corona. For a given compactness and a geometry, a alowermassof107M . Otherestimatessuggestsimilar
(cid:12)
sourcecannothaveatemperatureabovethelinesshown. smaller values (e.g. Zhou and Zhao 2010), but not low
Below the lines, the effect of pairs decreases and the enough to alter our conclusions about the plasma con-
plasma contains only electrons. As we have discussed tent. NotealsothatthecorrelationbetweenE andl is
c h
in section 3, the reflection spectrum constrains the inner weaker than the E −F correlation in Fig. 7. That is
c p
radiusofthedisktobeRin =47±23rg. Ifweassumethe because we use the wider energy range to calculated the
coronaisofthesamesize(otherwise,relativisticfeatures flux and also because of the uncertainties in the radius
inthereflectionwillbewashedout),weobtainthegreen measurements.
circles shown in Figure 8. We used the cutoff powerlaw
flux in the range 0.1–200 keV to measure the luminos- 4.2.4. Cutoff Variability
ity assuming a standard cosmology (Ωm = 0.3,Λ = 0.7) The additional information provided by the four mea-
with H = 68kms−1Mpc−1, and a black hole mass of surements of the plasma properties provide further con-
0
M =107.9M (Pontietal.2012)3. Theredarrowsshow straints. The cutoff energy in pair-dominated plasmas
(cid:12)
the upper limits on the observed l obtained by setting scales inversely with compactness (or with luminosity
h
R = 1.23r , the innermost stable circular orbit for a whenthesourcesizeisconstant,asisthecaseherewhere
in g
maximally spinning black hole. The electron tempera- we measure a constant reflection fraction and reflection
ture is estimated as kT ∼E /2 (Petrucci et al. 2001). parameters): E ∝ l−1 (Svensson 1994). This comes
e c c h
We can see that all the measurements fall below fromthefactthatincreasingl producesmorepairs,and
h
the pair limit lines for the three geometries, indicating for balance to hold, the same energy is now distributed
that the plasma in MCG–5-23-16 is not dominated by to more particles so the energy per particle drops. This
pairs and consists mostly of electrons. This is a ro- trend is not what we observe in MCG–5-23-16. Instead
bust statement given the small uncertainties in the cut- we find that the cutoff energy is higher for higher fluxes,
providing further evidence that the plasma is not domi-
3 WefollowasimilarproceduretoFabianetal.(2015). nated by pairs. In electron plasmas on the other hand,