Ultra Luminous X-ray Sources: a deeper insight into their spectral evolution

0.91 MB·

by  Fabio Pintore

Mon.Not.R.Astron.Soc.000,000–000(0000) Printed28January2014 (MNLATEXstylefilev2.2)
Ultra Luminous X-ray Sources: a deeper insight into their spectral
evolution
4 1 1 2 3
Fabio Pintore , Luca Zampieri , Anna Wolter , Tomaso Belloni
1
0 1INAF-OsservatorioAstronomicodiPadova,Vicolodell’Osservatorio5,I-35122Padova,Italy
2 2INAF,OsservatorioAstronomicodiBrera,viaBrera28,20121Milano,Italy
3INAF,OsservatorioAstronomicodiBrera,viaE.Bianchi46,I-23807Merate,Italy
n
a
J
7 28January2014
2

] ABSTRACT
E
H WeselectasampleofnearbyUltraluminousX-raysourceswithlongXMM-Newtonob-
. servationsandanalysealltheavailableXMM-NewtondatausingbothX-rayspectralfitting
h
techniquesandhardness-intensitydiagrams.ThesampleincludesIC342X-1,NGC5204X-
p
1,NGC5408X-1,HolmbergIXX-1,HolmbergIIX-1,NGC1313X-1,NGC1313X-2and
-
o NGC253X-1.Wefoundthat,althoughacommonreferencemodelcanbeusedtodescribethe
r X-rayspectra,thesourcesshowdifferentspectralevolutions,phenomenologicallydescribed
t
s intermsofvariationsinthepropertiesofasoftcomponentandahighenergytail.Variationsat
a lowenergiesareaccountedfor(mostly)bychangesinthenormalizationofthesoftcomponent
[
and/orinthecolumndensitytothesource,whilevariationsinthehighenergytailbychanges

1 intheparametersofanopticallythickcorona.Thisspectralvariabilityisratherwellcharacter-
v izedonacolour-colourandhardness-intensitydiagramintermsofsuitablydefinedhardness
5 ratios.Wesuggesttheexistenceofavariabilitypatternonthehardness-intensitydiagramand
1 weinterpretitintermsoftheswitchbetweenanear-Eddingtonandasuper-Eddingtonaccre-
8 tion regime.The transition between the two regimesseems to be driven mostly by changes
6
inthecontributionofthesoftcomponent,whichcanbeexplainedintermsoftheincreasing
.
1 importanceofwindemission.Theanalysisiscomplementedbyaninvestigationoftheshort-
0 termtime variabilityofall thesources.Ingeneral,no clearcorrelationbetweenthe spectral
4 andtemporalpropertiesisfound.
1
: Keywords: accretion,accretiondiscs–X-rays:binaries–X-Rays:galaxies–X-rays:indi-
v
viduals(IC342 X-1,NGC 5204X-1,NGC 5408X-1,HolmbergIX X-1, HolmbergII X-1
i
X andNGC253X-1)
r
a
1 INTRODUCTION 2011)ornear-EddingtonaccretionontomassivestellarBHsformed
from low metallicity stars (30-80 M⊙, e.g. Mapellietal. 2009;
Zampieri&Roberts2009;Belczynskietal.2010).
Ultra Luminous X-ray sources (ULXs) are a peculiar class
of extragalactic, point like and off-nuclear X-ray sources with Based on the early, low counting statistics XMM-Newton
isotropicluminosityinexcessof1039 ergs−1 (e.g.Feng&Soria spectraBHmassessignificantlyinexcessof100upto∼104M⊙
2011). Although they were discovered more than 30 years ago wereinferred(e.g.Milleretal.2003,2004).Howevermorerecent
and nowadays more than 450 ULXs are known and catalogued high quality XMM-Newton observations have shown that the X-
(e.g.Roberts&Warwick2000;Colbert&Ptak2002;Swartzetal. rayspectraofULXsarecharacterisedbypropertiesnotcommonly
2004;Liu&Bregman2005;Waltonetal.2011),theirnatureisstill observed in GalacticBH X-ray binary systems (XRBs) accreting
matter of debate and their observational properties are still puz- at sub-Eddington rates (Stobbartetal. 2006; Gonc¸alves&Soria
zling.Anumberofobservationalresults,includingtheX-rayvari- 2006). A roll-over at high energy, usually at 3-5 keV, is of-
ability and the existence of periodic modulations in the X-ray or ten observed coupled to a soft excess (Stobbartetal. 2006).
optical flux of some sources, suggest that they can be accreting Gladstoneetal.(2009)found thatitwasalmost ubiquitousinthe
black hole (BH) binary systems. Their extreme luminosities may highest quality ULX spectra and proposed that such curvature is
beexplainedintermsofsub-EddingtonaccretionontoIntermedi- a characteristic feature of a new spectral state, the ultraluminous
ateMassBHs(IMBHs,100-104M⊙;Colbert&Mushotzky1999), state(Roberts2007). Theymodelled thesespectra intermsof an
beamed and/or super-Eddington accretion onto stellar mass BHs optically thick corona coupled to an accretion disc. In particular,
(5-20 M⊙; e.g. Kingetal. 2001; Begelman 2002; Feng&Soria they proposed the existence of a spectral sequence in which the
2 FabioPintore,Luca Zampieri,AnnaWolter,TomasoBelloni
soft component becomes more and more important as the lumi- very recently been published also by Suttonetal. (2013). Adopt-
nosity increases. For the highest luminosity sources, this compo- ingamulticolourdiscplusapowerlaw,theyclassifiedtheULXsin
nent was later interpreted as originating from outflow ejections threespectralregimes,definedasbroadeneddisc,hardultralumi-
fromthedisc(Middletonetal.2011a).Thisisconsistentwithre- nousandsoftultraluminous.Thefirstischaracterisedbyadisc-like
cent theoretical calculationsof thehydrodynamic structureof ac- spectral shapewhiletheother twobythepredominance of either
cretion discs at super-Eddington rates (e.g. Poutanenetal. 2007; thepowerlawcomponentathighenergiesorthedisccomponentat
Ohsuga&Mineshige2007;Ohsugaetal.2009). lowenergies.Whileinsomerespectstheirinvestigationiscomple-
TheincreasingnumberofobservationsofULXsthatbecame mentarytoours(thegoalandsomeoftheresultsaresimilartothose
available in the last years also prompted an investigation of their reported here), inothersour approach differsbecause wedid not
long-termspectral statevariability.IC 342X-1and X-2werethe usefluxesfromaspectralmodelforcomputinghardness-intensity
firsttwoULXsinwhichtransitionsfromalow/hardtoahigh/soft andcolour-colourdiagrams,butadoptedtheXMM-Newtoncounts
spectralstatewereobserved(Kubotaetal.2001).However,thebe- indifferentenergybands.Wewillshowthatthismethodallowsus
haviour of the soft component is intriguing. In fact, although IC to classify ULXs also in case of low quality data and weak con-
342X-2showsacorrelationbetweenthediscluminosityandtem- straintsontheparametersofthespectralmodels,andthereforecan
peraturetypicalofastandardaccretiondisc,IC342X-1showsthe beappliedtoalargersampleofsources.
oppositebehaviour(Feng&Kaaret2009).Ananti-correlationwas The plan of the paper is the following. In Section 2 we de-
observedalsoinothersources(Kajava&Poutanen2009).Changes scribetheselectionofthesampleofULXs.InSection3wepresent
reminiscent of Galactic XRB spectral transitions were also ob- theadoptedX-raydatareductionprocedureandtheresultoftheX-
servedinNGC1313X-1andX-2(Feng&Kaaret2006).Asystem- rayspectralandtemporalanalysisofallthesources.InSection4,
aticanalysisoftheirX-rayspectralvariabilityintermsofacomp- we tentatively try to interpret the spectral evolution of ULXs on
tonisationmodelplusamulticolorblackbody discwasperformed thecolour-colourandhardness-intensitydiagramsand,finally,we
byPintore&Zampieri(2012),whocharacterisedthebehaviourof discussourresultsinSection5.
thesetwoULXsintermsof variationsintheoptical thickness of
theComptonizingcorona(verythickandthickstate).Theresulting
pictureisthat ULXsshow rather peculiar spectral changes, often
differentfromsourcetosource. 2 SELECTIONOFTHESAMPLE
As shown by at least 30 years of studies of Galactic XRBs,
We selected a sample of ULXs from the catalogues of
inadditiontothespectralshape,shorttermvariabilityisveryim-
Liu&Bregman(2005)andWaltonetal.(2011)thatwereobserved
portanttounderstandtheiraccretionstates(e.g.Belloni2010).Few
byXMM-Newton.Inordertodetectthehighenergycurvatureinthe
sources(NGC5408X-1,M82X-1andX42.3+59)showquasipe-
EPICspectrum,atleast10000countsareneeded(Gladstoneetal.
riodicoscillations(QPOs),theclassificationofwhichremainsstill
2009).Inaddition,ahighnumberoftotalcountsisalsorequested
unclear (e.g. Strohmayer&Mushotzky 2003; Mucciarellietal.
to perform an analysis of the chemical abundances. Winteretal.
2006; Strohmayeretal. 2007; Strohmayer&Mushotzky 2009;
(2006) showed that the Oxygen or Iron K-shell edges can be de-
Fengetal. 2010; Middletonetal. 2011b; Dheeraj&Strohmayer
tectedwithatleast5000or40000countsintheEPICinstrument,
2012; Caballero-Garciaetal. 2013). In general, the properties of
respectively. Theobservedcount ratesof theclosestULXsinthe
the short term variabilityof ULXsarestill poorly understood. In
EPIC-pndetectorareintherange∼0.1−1.5counts−1andhence
fact, sources with similar X-ray spectra show different temporal
longexposure timesof∼ 15−100ksareneededtoprovidethe
variability.Heiletal.(2009)analysedthePowerSpectralDensities
requiredamountoftotalcounts.
(PSD)ofasampleof16brightULXsandfoundthat,irrespectively
Following these constraints we select nearby ULXs (D 6 5
of their X-rayspectra, therearetwogroups of sources: asmaller
Mpc) that have at least one long observation with XMM-Newton
group displays a well defined variability at about the same level
(∼ 10%), while in the other one the variability is almost absent. (texp >15 ks). This gives us the required ∼10000 counts in
theEPICinstrument. Theadditional requirement of atleast three
Recently, Middletonetal.(2011,a,b) showedthat alsotheanaly-
XMM-Newtondatasetsindifferentepochs(evenifsomeepochsare
sisoftheenergydependenceoftheshorttermvariabilitycanbea
ofshorterexposures)allowsustostudythespectralvariability.The
powerful tooltodiscriminateamong differentspectral models.In
finallistincludes:IC342X-1,NGC253X-1,NGC5204X-1,NGC
particular, they suggested the possibility that the short-term vari-
5408X-1,HoIXX-1andHoIIX-1,NGC1313X-1andX-2.The
abilityisproduced by turbulencesinaclumpy wind, whichfrom
sourcesinNGC1313 havebeenpresented inPintore&Zampieri
timetotimeencountersourlineofsight.
(2012)andwillbeaddedtothediscussionoftheresults.Thissam-
Theaimofthepresent workistoinvestigateinasystematic
plemightnotbefullyrepresentativeoftheULXpropertiesbutthe
way the spectral variability of ULXs on a sample of sources se-
adoptedselectioncriteriaallowustosamplearatherlargerangeof
lected to have high quality XMM-Newton observations. We com-
luminosities(L ∼1−30×1039ergs−1).
plement the spectral analysis witha careful investigation of their X
shorttermvariability.Weattempttocharacterizethespectralvari-
abilityusingalsothehardnessratiosandcolour-colourdiagrams,
that have been successfully adopted in the past to study the be-
haviourofXRBs.Consideringtherelevanceofthemetallicityfor 3 DATAANALYSISOFSPECTRALANDTIMING
someoftheproposedscenariosfortheformationofULXs,wealso PROPERTIES
put a certain effort in using the highest counting statistics X-ray
3.1 DataReduction
spectra to attempt a measurement of the chemical abundances in
theenvironmentofULXs,followingtheapproachofWinteretal. We carried out a complete spectral and temporal analysis on all
(2007)andPintore&Zampieri(2012).Asystematicanalysisofthe theavailableXMM-NewtonobservationsoftheULXslistedabove.
spectral and temporal variability of a larger sample of ULXshas Weexcludedfromtheanalysisobservationsperformedearlierthan
UltraLuminousX-raySources:a deeper insightintotheirspectralevolution 3
Table1.LogoftheobservationsoftheULXsanalysedinthiswork.
No. Obs.ID Date Expa Instr.b EPIC-pn Netcounts Off-axisangle
(ks) counts−1
IC342X-1;D=3.3Mpcc;NHGal=31.1·1020cm−2d;Ra,Dec(J2000):034555.5,+680454.2
1 0093640901 2001-02-11 4.83 pn 0.37 1761 5.08’
2 0206890101 2004-02-20 17.06 pn/M1/M2 0.42 6863,3212,3176 2.49’
3 0206890201 2004-08-17 14.36 pn/M1/M2 0.90 12473,7757,7861 4.27’
4 0206890401 2005-02-10 5.92 pn/M1/M2 1.11 6593,3718,4409 2.63’
NGC253X-1;D=3.9Mpce;NHGal=1.3·1020cm−2d;Ra,Dec(J2000):004722.56,-252051.0
1 0110900101 2000-12-13 10.97 pn/M1/M2 0.105 1154,802,893 4.57’
2 0152020101 2003-06-19 61.37 pn/M1/M2 0.271 16625,6978,7201 5.30’
3 0304851101 2005-02-10 10.93 pn/M1/M2 0.097 1057,727,694 3.13’
4 0304850901 2006-01-02 8.82 pn/M1/M2 0.141 1241,458,488 3.14’
5 0304851001 2006-01-06 8.75 pn/M1/M2 0.155 1359,530,549 3.17’
6 0304851201 2006-01-09 16.00 pn/M1/M2 0.158 2525,994,965 3.19’
7 0304851301 2006-01-11 4.34 pn/M1/M2 0.162 703,338,373 3.22’
NGC5204X-1;D=4.8Mpcf;NHGal=1.39·1020cm−2d;Ra,Dec(J2000):132938.6,+582505.7
1 0142770101 2003-01-06 15.33 pn/M1/M2 0.624 9560,3121,3211 1.14’
2 0142770301 2003-04-25 4.02 pn/M1/M2 0.862 3465,1903,1868 1.12’
3 0150650301 2003-05-01 5.26 pn/M1/M2 1.018 5357,2213,2327 1.30’
4 0405690101 2006-11-15 10.05 pn/M1/M2 1.228 12600,7752,7752 1.10’
5 0405690201 2006-11-19 31.25 pn/M1/M2 1.031 32219,11791,11957 1.08’
6 0405690501 2006-11-25 22.42 pn/M1/M2 0.775 17364,6906,7174 1.13’
NGC5408X-1;D=4.8Mpce;NHGal=5.67·1020cm−2d;Ra,Dec(J2000):140319.6,-412259.6
1 0112290501 2001-07-31 3.68 pn/M1/M2 1.441 5303,2530,2642 1.26’
2 0112290601 2001-08-08 4.50 pn/M1/M2 1.337 6023,2024,2140 1.27’
3 0112290701 2001-08-24 7.50(MOS1) M1/M2 1.018(MOS1) 2405,2482 1.23’
4 0112291201 2003-01-27 2.79 pn/M1/M2 1.228 2354,920,935 0.99’
5 0302900101 2006-01-13 92.54 pn/M1/M2 1.031 94298,25437,25331 1.09’
6 0500750101 2008-01-13 46.09 pn/M1/M2 0.775 43753,18471,17782 1.07’
7 0653380201 2010-07-17 92.68 pn/M1/M2 1.137 105377,26034,33154 1.11’
8 0653380301 2010-07-19 96.86 pn/M1/M2 1.113 107805,30145,30188 1.12’
9 0653380401 2011-01-26 87.38 pn/M1/M2 1.061 92710,29307,29126 1.12’
10 0653380501 2011-01-28 88.51 pn/M1/M2 1.024 90634,27952,27822 1.12’
HoIIX-1;D=4.5Mpce;NHGal=3.42·1020cm−2d;Ra,Dec(J2000):081929.0,+704219.3
1 0112520601 2002-04-10 4.64 pn/M1/M2 3.054 14164,8316,8886 1.13’
2 0112520701 2002-04-16 3.77 pn/M1/M2 2.751 10377,4796,4994 1.11’
3 0112520901 2002-09-18 4.33 pn/M1/M2 0.815 3527,1317,1434 1.11’
4 0200470101 2004-04-15 40.75 pn/M1/M2 3.056 124532,49487,50578 1.14’
5 0561580401 2010-03-26 23.19 pn/M1/M2 1.238 28709,14098,13993 1.14’
HoIXX-1;D=3.55Mpcg;NHGal=4.06·1020cm−2d;Ra,Dec(J2000):095753.2,+690348.3
1 0112521001 2002-04-10 7.05 pn/M1/M2 1.895 13350,5806,5782 1.11’
2 0112521101 2002-04-16 7.64 pn/M1/M2 2.173 16610,7041,7338 1.13’
3 0200980101 2004-09-26 83.17 pn/M1/M2 1.495 124339,46782,47339 1.13’
4 0657801601 2011-04-17 0.96 pn/M1/M2 0.760 733,2851,2807 7.40’
5 0657801801 2011-09-26 7.40 pn/M1/M2 2.477 13830,11969,13487 5.29’
6 0657802001 2011-03-24 3.20 pn/M1/M2 1.369 4381,2882,3672 7.32’
7 0657802201 2011-11-23 13.10 pn/M1/M2 2.211 28964,16186,15303 5.21’
aGTIofEPIC-pn;bpn=EPIC-pncamera;M1/M2=EPIC-MOS1/MOS2camera;cSahaetal.(2002);dDickey&Lockman(1990);eKarachentsevetal.
(2003);f Stobbartetal.(2006);gFreedmanetal.(1994);
December2000,becausethecalibrationbeforethisdatemaybein- solar flares. In order to avoid distortions induced by background
complete.DatawerereducedusingSASv.11.0.0,extractingspec- particles, EPIC-MOSand EPIC-pnspectra were extracted select-
tra and lightcurves from events with PATTERN6 4 for EPIC-pn inggoodtimeintervalswithabackgroundcountrateintheentire
(whichallowsforsingleanddoublepixelevents)and PATTERN6 fieldofviewnothigherthan0.7counts−1inthe10−12keVen-
12 for EPIC-MOS (which allows for single, double, triple and ergyrange.Afewobservations ofNGC5408 X-1andHoIIX-1
quadruplepixelevents).Weset‘FLAG=0’inordertoexcludebad wereexcludedfromtheanalysisbecausetheeventlistwasempty.
pixels and events coming from the CCD edge. We also excluded The log of observations and relevant information on the sources,
fromtheanalysistheobservationsinwhichthesourceswereinthe includingtotalnetcountsinthevariousinstruments,isreportedin
CCDgapandanumberofobservationsseverelyaffectedbystrong Table1.
4 FabioPintore,Luca Zampieri,AnnaWolter,TomasoBelloni
SpectraandlightcurvesofIC342X-1,NGC253X-1,HoII ofNGC5408X-1weaddanunderlyingplasmacomponent(APEC
X-1,HoIXX-1and NGC5408 X-1wereobtained selectingcir- inXSPEC)withatemperatureof∼0.9keVthatleadstoasignifi-
cularextractionregionsof30”and65”forsourceandbackground cantimprovementinthefit.Thisemissioncouldbeproducedeither
(whenpossible,onthesameCCDwherethesourceislocated),re- bythediffusegasinthehostgalaxyordirectlyfromtheenviron-
spectively. NGC 5204 X-1 was often very close to the CCD gap ment around the source (Strohmayeretal. 2007; Middletonetal.
andhencetheextractionregionwasdifferentfromobservationto 2011b).
observation(rangingfrom24”to31”forthesourceandfrom50” The (unabsorbed) luminosities of the sources of the sample
to65”forthebackground). spanfrom∼7·1038ergs−1to∼3·1040ergs−1(Figure2-top),
Allthespectrawererebinnedwith25countsperbininorder andthereforesampleawiderangeofobservedULXluminosities.
to apply the χ2 statistics. The spectral fitswere performed using Most of the sources exhibit intermediate luminosities, clustering
XSPECv. 12.6.0(Arnaud 1996).Toimprove thecounting statis- around ∼ (7−8)·1039 erg s−1, but with significant long-term
tics we fittedEPIC-pn and EPIC-MOSspectra simultaneously in variability, which in some cases is higher than a factor of 3 (i.e.
the 0.3-10 keV energy range. In all fits, a multiplicativeconstant HolmbergIXX-1,NGC1313X-2andIC342X-1).
was introduced for the three instruments to account for possible In all the observations the comptonizing medium (corona)
residualdifferencesincalibration.TheEPIC-pnconstantwasfixed is optically thick and cold (Figure 2-bottom). The temperatures
to1,whilefortheMOStheywereallowedtovary.Ingeneral,the and optical depths are in the range kT ∼ 1 − 6 keV and
cor
differenceamongthethreeinstrumentswaslessthan10%. τ ∼ 3−30, significantly lower the former and higher the latter
thanthose seeninGalacticXRBs(kT > 50 keV and τ 6 1,
cor
e.g. McClintock&Remillard 2006)1. This is consistent with the
3.2 Spectralfits
findingsofGladstoneetal.(2009),whointerprettheultraluminous
A combination of a multicolor disc plus a comptonisation model stateintermsofanopticallythick,coldcoronaphysicallycoupled
has been shown to offer a reasonable description of both poor totheinnerregionsofanaccretiondisc.However,thefitofobser-
and high quality ULX spectra and to provide a better fit vation#2ofNGC5408X-1showsamorepronounceddegeneracy
than simple models as a multicolour blackbody disc (diskbb in intheparametersofthecoronawhichisalsoconsistentwithbeing
XSPEC, Mitsudaetal. 1984), powerlaw, a slim disc (diskpbb warm(84keV) and marginallyopticallythin(τ ∼ 0.9). Forthis
in XSPEC, Mineshigeetal. 1994) or a diskbb+powerlaw (i.e. reason, thisobservation isnot shown intheplot τ −kTcor (Fig-
Stobbartetal.2006;Gladstoneetal.2009;Middletonetal.2011a; ure2-bottom).
Pintore&Zampieri2012).Followingthesefindingsandguidedby Atsuchhighopticaldepthsandlowtemperatures,thephysical
theideaofdescribingthespectralevolutionofULXswithinacom- conditionsinthecoronaareratherdifferentfromthoseinGalactic
monframework,wethenfittedthespectraofoursampleofULXs XRBs.Ifthegasispurehydrogen,thebremsstrahlungluminosity
adopting as reference model a multicolor accretion disc (diskbb ofthecoronaisLbrem,cor =2.5·1021·Tc1o/r2·τes·rcorergs−1,
in XSPEC) plus a comptonising component (comptt in XSPEC; whereτesistheelectron-scatteringopticaldepthandrcoristhera-
Titarchuk1994).Wenotethatforsomesourcesthespectralfitswith diusofthecorona.AssumingthatL isnegligible(<1038
brem,cor
thistwo-componentmodeloftendisplayseverallocalminimawith erg s−1), we find rcor . 7×1010 cm for kTcor ≈ 1 keV and
verysimilarvaluesoftheχ2,sometimeswithevidenceforbotha τes ≈ 5.Atthesametime,askingthattheeffectiveopticaldepth
strong/warmandaweak/cool(orno)disc.InTable2wequotethe τeff = [(τabs +τes)·τabs]1/2 < 1 (τabs is the true emission-
valuesoftheparametersandformalstatisticalerrors(at90%con- absorption optical depth) and assuming τabs << τes, we obtain
fidencelevel)fortheabsoluteminimafoundwiththismodel,but rcor & 15 cm (again for kTcor ≈ 1 keV, τes ≈ 5). Therefore,
westressthat,forsomepoorqualityspectra,theactualuncertainty physicalconditionsareconsistentwiththeexistenceofbothcom-
on the spectral parameters caused by the topology of the χ2 sur- pactandmoderatelyextendedopticallythickcoronae.
facemaybelarger.Becauseofthelowstatistics,inthemajorityof Wenotethatthesourcesoccupyboththeverythickandthick
thecaseswearenotabletoindependentlyvarythetemperatureof regionsoftheτ −kTcorplane(dividedbytheboundaryatτ ∼9,
thedisc(Tdisc)andthatoftheseedphotons(T0)andhencewetie as nominally defined in Pintore&Zampieri 2012). However, the
themtogether.Althoughthisisnotfullyconsistentfromaphysical tworegionsdonotappeartobecompletelydetachedonthisplane
pointofview(thecoronaisoftenopticallythickandthentheinner butrather smoothlyconnected. Somesources arealsoresidingin
disc ishidden), leavingthem free to vary independently withina eitheroneortheotherstateatdifferenttimes(seebelow).
factor of afew does not leadto significant qualitative changes in Concerningthesoftcomponent,thetemperaturesareconsis-
theresults(seealsoPintore&Zampieri2012). tentwiththosefoundinpreviousworks(∼ 0.15−0.30keV,e.g.
Two absorption components (tbabs, Wilmsetal. 2000, in Stobbartetal.2006).OnlyinsomeobservationsofNGC253X-1,
XSPEC)wereconsideredforallspectralmodels:onefixedatthe we found that temperature of the soft component increases up to
Galacticcolumndensityalongthedirectionofthesourceandthe 0.8keV,althoughitshowsequallygoodfitswithbothacoldand
second one, free tovary inorder to account for local absorption. a warm disc component (Barnard 2010). We emphasize that the
The adopted Galactic N and distances are listed in Table 1, in softcomponentmaycontribute∼ 20−30%ofthetotalemission
H
theheadersforeachsource.InFigure1weshowtheresultsofthe inthe0.3-10keV energyrangeforthelessluminoussourcesand
spectralfitsobtainedatthelowest,mediumandhighestluminosity up to∼ 50% for the most luminous ones. Thissuggests either a
levelforeachsource.Theygiveavisualimpressionoftheobserved
spectralandintensityvariability.Thesourcesareorderedaccord-
ingtotheir positionalong thesequence onthehardness-intensity 1 Wenotethat,fortemperatures ofthecorona& 2keV,thetrendinthe
diagramofFigure6-left(seeSection4.2).Ingeneral,thesources kTcor−τplane(Figure2-bottom)isnotverydifferentfromthatexpected
show atrendinwhichtherelativeimportance ofthesoftcompo- forconstantspectralindex/Comptonparameterandmaythenreflectinpart
nent increases with the luminosity of the source. In Table 2 we somedegeneracy inthedata.However, below≃ 2keV,thebehaviouris
reportthebestfitparameters.Forthehighestqualityobservations realbecausekTcoriswellconstrainedwithintheXMM-Newtonbandpass.
UltraLuminousX-raySources:a deeper insightintotheirspectralevolution 5
Table2.Bestfittingspectralparametersobtainedwiththeabsorbeddiskbb+compttmodel.Theerrorsareat90%foreachparameterofinterest.
No. NHa kTdiscb,c kTcord τe APECf LX[0.3-10keV]g Ldisc[0.3-10keV]h χ2/dof
(1021cm2) (keV) (keV) keV (1039ergs−1) (1039ergs−1)
IC342X-1
1 6.7+0.5 0.213+0.007 3.2+0.1 6.6+2 5.6+1.2 1.2+0.6 57.17/58
−0.5 −0.007 −1.2 −2 −1.1 −0.4
2 5.5+0.2 0.390+0.003 2.21+0.04 9.0+0.2 5.7+0.5 1.7+0.2 443.53/429
−0.2 −0.003 −0.04 −0.2 −0.5 −0.3
3 6.2+0.1 0.494+0.003 1.73+0.02 9.9+0.1 11.3+0.4 3.8+0.2 753.10/788
−0.1 −0.003 −0.02 −0.1 −1.1 −0.3
4 5.3+0.2 0.348+0.007 1.98+0.02 8.5+0.1 13.3+1.2 1.1+0.3 479.56/485
−0.2 −0.007 −0.02 −0.1 −1.1 −0.2
NGC253X-1
1 0.97+0.01 0.236+0.004 2.8+0.1 5.6+0.4 0.8+0.2 0.3+0.04 93.88/97
−0.01 −0.004 −0.1 −0.3 −0.2 −0.05
2 0.82+0.04 0.794+0.002 1.25+0.03 27+5 2.4+0.1 1.7−0.2 852.44/810
−0.04 −0.002 −0.03 −4 −0.1 −0.5
3 0.64+0.01 0.104+0.007 0.94+0.02 13.8+0.4 0.72+0.19 <0.26 88.57/91
−0.01 −0.007 −0.02 −0.4 −0.14
4 0.20+0.02 0.464+0.009 0.97+0.04 20+3 0.99+0.31 0.33+0.11 75.117/73
−0.02 −0.009 −0.04 −2 −0.11 −0.07
5 0.45+0.02 0.35+0.01 1.62+0.06 9.5+0.6 1.2+0.3 0.21+0.06 88.229/83
−0.02 −0.01 −0.06 −0.6 −0.23 −0.06
6 0.49+0.01 0.715+0.008 1.61+0.08 10+1 1.30+0.3 0.64+0.08 174.57/161
−0.01 −0.008 −0.08 −1 −0.2 −0.07
7 0.40+0.03 0.79+0.002 1.80+0.02 9+3 1.30+0.5 0.68+0.16 44.325/45
−0.03 −0.002 −0.02 −3 −0.44 −0.13
HoIXX-1
1 1.26+0.05 0.240+0.002 2.75+0.03 7.2+0.1 13.8+1 2.0+0.2 685.17/706
−0.05 −0.002 −0.03 −0.1 −0.9 −0.2
2 1.23+0.05 0.219+0.002 2.85+0.03 6.66+0.08 15.8+1 1.3+0.2 749.82/817
−0.05 −0.002 −0.03 −0.08 −0.9 −0.1
3 1.35+0.02 0.245+0.001 2.38+0.01 9.07+0.05 12.1+0.3 2.19+0.06 2182.62/2044
−0.02 −0.001 −0.01 −0.05 −0.3 −0.05
4 0.60+0.1 0.320+0.010 3.03+0.08 6.4+0.2 22+2 2.3+0.5 159.82/195
−0.1 −0.010 −0.07 −0.2 −2 −0.5
5 1.09+0.04 0.258+0.003 1.91+0.01 8.45+0.08 23+1 1.7+0.2 907.47/967
−0.04 −0.003 −0.01 −0.08 −2 −0.1
6 1.60+0.08 0.250+0.002 4.5+0.1 6.4+0.2 16+2 2.9+0.4 340.03/350
−0.08 −0.002 −0.1 −0.2 −1 −0.3
7 1.22+0.04 0.266+0.002 1.86+0.01 8.74+0.07 28+2 2.2+0.2 1122.67/1154
−0.04 −0.002 −0.01 −0.07 −1 −0.2
NGC5204X-1
1 0.04+0.04 0.306+0.002 1.39+0.02 12.7+0.4 5.1+0.3 1.90+1.2 398.45/453
−0.04 −0.002 −0.02 −0.4 −0.3 −0.4
2 0.67+0.06 0.203+0.002 6.20+0.2 3.5+0.2 7.7+1 2.3+0.3 238.65/221
−0.06 −0.002 −0.2 −0.1 −0.9 −0.3
3 0.31+0.05 0.283+0.002 1.55+0.03 8.8+0.3 7.7+0.9 3.1+0.3 301.49/271
−0.05 −0.002 −0.03 −0.3 −0.8 −0.2
4 0.33+0.03 0.216+0.001 1.44+0.01 7.2+0.1 8.7+0.6 2.3+0.2 717.93/722
−0.03 −0.001 −0.01 −0.1 −0.6 −0.2
5 0.47+0.02 0.265+0.001 2.35+0.02 5.72+0.08 8.1+0.4 2.9+0.1 838.98/822
−0.02 −0.001 −0.02 −0.02 −0.4 −0.1
6 0.32+0.03 0.231+0.001 4.31+0.05 4.6+0.09 6.6+0.4 1.8+0.1 642.82/676
−0.03 −0.001 −0.05 −0.05 −0.4 −0.1
NGC5408X-1
1 0.11+0.03 0.180+0.001 2.00+0.04 5.0+0.2 0 10.0+0.2 6.6+0.5 280.79/271
−0.03 −0.001 −0.04 −0.2 −0.5 −0.4
2 0.09+0.03 0.192+0.001 84+2 0.11+0.02 0 9.8+1 6.2+0.5 235.86/273
−0.03 −0.001 −2 −0.01 −0.9 −0.5
3 0.28+0.05 0.177+0.001 1.03+0.03 8.9+0.3 0 11.0+0.1 7.4+0.7 93.49/133
−0.05 −0.001 −0.03 −0.3 −0.1 −0.6
4 0.72+0.07 0.171+0.001 1.56+0.04 6.8+0.3 0 8.2+1.3 4.7−0.6 118.3/127
−0.07 −0.001 −0.04 −0.3 −1.2 +0.6
5 0.62+0.01 0.153+0.002 1.656+0.008 6.04+0.04 0.92+0.06 8.6+0.2 4.5+0.1 968.60/913
−0.01 −0.002 −0.008 −0.04 −0.06 −0.2 −0.2
6 0.61+0.01 0.154+0.003 1.506+0.009 6.77+0.05 0.84+0.04 7.9+0.3 3.7+0.1 868.60/796
−0.01 −0.003 −0.009 −0.05 −0.03 −0.3 −0.1
7 0.59+0.01 0.156+0.002 1.813+0.008 5.88+0.03 0.95+0.03 9.6+0.2 4.3+0.1 1085.20/1030
−0.01 −0.002 −0.008 −0.03 −0.03 −0.2 −0.1
8 0.43+0.01 0.163+0.001 1.882+0.008 5.73+0.03 0.96+0.04 8.9+0.3 3.8+0.1 1158.99/1023
−0.01 −0.001 −0.008 −0.03 −0.04 −0.2 −0.1
9 0.64+0.01 0.154+0.001 1.728+0.007 5.97+0.03 0.902+0.04 9.1+0.2 4.2+0.1 1045.72/981
−0.01 −0.001 −0.007 −0.03 −0.04 −0.3 −0.1
10 0.55+0.01 0.160+0.04 2.047+0.009 5.43+0.03 0.87+0.04 8.5+0.6 3.8+0.1 1055.82/995
−0.01 −0.03 −0.009 −0.03 −0.03 −0.3 −0.1
HoIIX-1
1 0.39+0.03 0.189+0.001 4.69+0.04 3.10+0.04 21+1 3.2+0.3 576.28/629
−0.03 −0.001 −0.04 −0.04 −1 −0.3
2 0.77+0.04 0.200+0.004 2.16+0.03 6.2+0.1 22+1 7.4+0.3 507.21/532
−0.04 −0.004 −0.03 −0.1 −1 −0.3
3 0.75+0.06 0.167+0.001 1.00+0.02 8.4+0.2 5.7+0.5 2.5+0.3 191.13/185
−0.06 −0.001 −0.02 −0.2 −0.7 −0.3
4 0.48+0.01 0.203+0.001 2.63+0.01 4.66+0.02 23+1 5.5+0.1 1326.82/1318
−0.01 −0.001 −0.01 −0.02 −1 −0.2
5 0.57+0.02 0.200+0.001 1.51+0.01 7.17+0.07 8.6+0.2 3.6+0.2 819.61/760
−0.02 −0.001 −0.01 −0.07 −0.3 −0.1
aIntrinsiccolumndensityinexcesstheGalacticone;bInnerdisctemperature;cTheseedphotonstemperatureT0isassumedtobeequaltoTdisc;d
Temperatureofthecorona;eOpticaldepthofthecorona;f Temperatureoftheplasmacomponent;gUnabsorbedtotalX-rayluminosity;hUnabsorbeddisc
luminosity.
6 FabioPintore,Luca Zampieri,AnnaWolter,TomasoBelloni
NGC 253 X−1 IC 342 X−1
V)−1 L ~ 1039 erg s−1 V)−1 L ~ 9*1039 erg s−1
e aver e aver
k k
s −1 s −1 10−3
m −2 10−4 m −2
s c s c 10−4
n n
o o
ot ot
h h
P P 10−5
V (2 10−5 V (2
e e
k 2 k
2
0
χ χ 0
−2
−2
0.5 1 2 5 0.5 1 2 5
Energy (keV) Energy (keV)
Ho IX X−1 NGC 5204 X−1
s keV)−1−1 0.01 L aver ~ 2*1040 erg s−1 s keV)−1−1 10−3 L aver ~ 7*1039 erg s−1
m −2 10−3 m −25×10−4
c c
s s
n n
oto 10−4 oto2×10−4
h h
P P
V (2 V (2 10−4
e e
k 4 k
2
2
0
χ χ
0
−2
−2
0.5 1 2 5 0.5 1 2 5
Energy (keV) Energy (keV)
NGC 5408 X−1 Holmberg II X−1
V)−1 2×10−3 V)−1 L ~ 2*1040 erg s−1
e L ~ 9*1039 erg s−1 e aver
s k−1 10−3 aver s k−1 2×10−3
m −2 5×10−4 m −2 10−3
c c5×10−4
ns 2×10−4 ns
o o
hot 10−4 hot2×10−4
P P
V (2 5×10−5 V (2 10−4
e e
k k
2 2
χ 0 χ 0
−2 −2
0.5 1 2 5 0.5 1 2 5
Energy (keV) Energy (keV)
Figure1.ComparisonofEPIC-pnunfolded (E2f(E))spectra,fittedwiththediskbb+comptt model(Table2).Fordisplaypurposes,thespectraandthe
residualswererebinnedataminimumof5σ.Onlythehighest(red),lowest(black)andmediumluminosity(green)spectraforeachsourceareshownfor
clarity.Thedashedlinesarethediskbbandcompttcomponents,whilethesolidlineisthesumofthetwo(inNGC5408,theAPECcomponentistakeninto
accountforthefitbutnotshownintheplot).Goingfromtop-lefttobottom-right,thesourcesareorderedaccordingtotheirpositionalongthesequenceon
thehardness-intensitydiagramofFigure6(seeSection4.2).Thelegendshowstheaverage(0.3-10keV)unabsorbedluminosity.Thereisatrendinwhichthe
relativeimportanceofthesoftcomponentincreaseswiththeluminosityofthesource.
UltraLuminousX-raySources:a deeper insightintotheirspectralevolution 7
Table3.AbundancesinferredfromtheOxygenK-shellphotoionizationedge(0.538keV).
12+log(O/H)a
No.Observation #2 #3 #4
1) 10 IC342X-1 8.75±0.13 8.75±0.15 8.63±0.15
-g s No.Observation #2
er
9 NGC253X-1 8.8±0.1
3
0
1 No.Observation #7 #8 #9 #10
(X IC 342 X-1
L NGC 253 X-1 NGC5408X-1 8.71−+00..711 8.72−+00..1106 8.64−+00..1028 8.63+−00..1074
NGC 5204 X-1
NGC 5408 X-1 No.Observation #3 #7
1 NGC 1313 X-2
NGC 1313 X-1 HoIXX-1 8.73+0.05 8.68+0.11
Holmberg II X-1 −0.04 −0.16
Holmberg IX X-1
No.Observation #5
52000 53000 54000 55000 56000
Time (MJD) HoIIX-1 8.63−+00..1046
a For the solar abundance we assume 12 + log(O/H) = 8.69
30 IC 342 X-1 (Asplundetal.2009).
NGC 253 X-1
NGC 5204 X-1
NGC 5408 X-1
20 NGC 1313 X-2 8.69dex;Asplundetal.2009).WeselectedonlyEPIC-pnspectra
NGC 1313 X-1 withmorethan5000counts(Winteretal.2007),i.e.observations
Holmberg II X-1
Holmberg IX X-1 #2,3,4forIC342X-1,#2forNGC253X-1,#3,7forHolm-
bergIXX-1,#1,4,5,6forNGC5204X-1,#7,8,9and10forNGC
τ 10 5408X-1and#5forHolmbergIIX-1.ForNGC5204X-1,theX-
9
8 rayspectrumisratherinsensitivetovariationsoftheOabundance
7 sothatnodefiniteconclusioncanbedrawn.InTable3weshowthe
6 estimatedchemicalabundancesandtheiruncertainties.Theabun-
5 dances are consistent with solar with the exception of NGC 253
4 X-1whichismarginallysuper-solar (8.8dex). Themetallicityof
Ho II X-1 obtained by Goadetal. (2006) differs from ours pos-
3
siblybecauseofthedifferentreferencevalueadoptedfor[O/H]⊙
1
(8.93dex,Wilmsetal.2000).
kTcor (keV) Finally,noclearevidence ofanironL-shelledgewasfound
inanyofthesourcesofoursample.Thisisinlinewiththelackof
evidenceoftheironK-shellfeaturesinULXspectracomingfrom
Figure2.top:Unabsorbedtotalluminositiesinthe0.3-10keVenergyband
anionisedwind(e.g.Waltonetal.2013).
as a function oftime; bottom: optical depth τ versus temperature ofthe
coronakTcor(diskbb+compttmodel).WeaddedalsoNGC1313X-1and
X-2 forcomparison (see Pintore&Zampieri 2012). Different colors and
symbolsrefertothesources,aslistedinset. 3.4 Temporalanalysis
Wecomplementedthespectralanalysiswithaninvestigationofthe
temporalpropertiesoftheULXsofoursample.Foreachsourcewe
higherobscurationofthehardcomponentoranincreaseintheim-
computedthefractionalrootmeansquare(RMS)variabilityampli-
portanceofthesoftcomponent,bothlikelyassociatedtotheonset
tude(F ),thatmeasuresthevarianceofasourceoverthePoisso-
ofstrongerwinds. var
niannoiseinthetimedomainandisusuallynormalizedtotheaver-
agecount-rate(Edelsonetal.2002;Vaughanetal.2003).Thefrac-
tionalvariabilitycanbeusedalsotostudytheenergydependence
3.3 Chemicalabundances
oftheshort-termvariabilityofthesourceanditisapowerfultoolin
Wetentatively tried to determine the chemical abundances in the ordertocharacterisethepropertiesofthespectralcomponents(e.g.
ULXenvironmentslookingforOxygenandIronedgesintheirX- Middletonetal.2011a).WeevaluatedF frombackgroundsub-
var
ray spectra. In the spectral fits the tbabs absorption model is re- tractedlightcurvesinseveral energybands−0.3-10 keV, 0.3-2.0
placed with tbvarabs that allows variation in the chemical abun- keVand2.0-10keV−binningtheminintervalsof∆T = 200s.
dances(andgraincomposition).Wesetalternativelytheabundance Thisisagoodcompromiseforallthesourcesbecauseitallowsto
ofOxygen orIrontozero. Thespectrumwasthenfittedwiththe haveatleast20countsineachtimebinandatleast20bins.InTa-
EPIC continuum best fitting model (keeping all parameters, but ble4wereportthefractionalvariabilitymeasurementsforallthe
normalizations, fixed) plus an absorption edge, that accounts for observations.Whenthestatisticsarenotsufficientweprovideonly
theobservedabsorptionfeature.ForNGC5408X-1,weremoved the3σupperlimit.
theAPECcomponentwhichmaysignificantlyaffectthemeasure- UnlikeSuttonetal.(2013)whoselectedonlythehighestqual-
ment. The parameters of the edge are then used to compute the ityobservationsforeachsourceoftheirsample,westudythevari-
abundance (assuming for the solar Oxygen metallicity the value ability of all the observations. We find that the 0.3-10 keV band
8 FabioPintore,Luca Zampieri,AnnaWolter,TomasoBelloni
Table4.FractionalvariabilityoftheULXsanalysedinthiswork.
Source Obs.ID 0.3−10keV 0.3−2.0keV 2.0−10keV
countss−1 Fvaar countss−1 Fvaar countss−1 Fvaar
0093640901 0.490±0.010 613 0.248±0.009 619 0.252±0.009 2±19
0206890101 1.090±0.010 7±2 0.548±0.009 613 0.550±0.009 8±3
IC342X-1 0206890201 0.561±0.007 6±2 0.294±0.005 2±8 0.276±0.005 8±3
0206890401 1.430±0.020 21±2 0.67±0.01 12±3 0.77±0.02 29±2
0110900101 0.146±0.007 640 0.132±0.006 639 0.066±0.007 695
0152020101 0.408±0.004 29±1 0.416±0.004 25±1 0.137±0.003 43±3
0304851101 0.148±0.006 631 0.121±0.006 640 0.070±0.005 668
NGC253X-1 0304850901 0.180±0.006 12±5 0.137±0.005 623 0.077±0.005 647
0304851001 0.199±0.006 618 0.152±0.005 623 0.082±0.005 645
0304851201 0.205±0.005 618 0.157±0.004 620 0.077±0.003 641
0304851301 0.209±0.009 627 0.157±0.008 629 0.085±0.007 652
0112521001 2.150±0.020 1±3 1.55±0.02 4±2 0.62±0.01 612
0112521101 2.500±0.020 65 1.78±0.02 66 0.72±0.01 610
0200980101 1.709±0.006 2±1 1.185±0.005 65 0.537±0.004 2±3
HoIXX-1 0657801601 5.000±0.100 1±21 2.6±0.1 621 1.48±0.07 621
0657801801 3.680±0.050 1±7 2.56±0.04 612 1.16±0.03 23±3
0657802001 2.370±0.040 69 1.68±0.03 4±3 0.74±0.02 615
0657802201 4.770±0.030 66 3.25±0.03 67 1.56±0.02 610
0142770101 0.667±0.008 5±2 0.557±0.007 5±2 0.121±0.003 624
0142770301 0.920±0.020 613 0.79±0.02 614 0.146±0.008 637
0150650301 1.090±0.020 612 0.95±0.02 614 0.147±0.008 634
NGC5204X-1 0405690101 1.370±0.020 68 1.20±0.02 69 0.176±0.007 626
0405690201 1.118±0.007 4±1 0.975±0.006 5±1 0.154±0.003 614
0405690501 0.839±0.007 67 0.714±0.006 67 0.135±0.003 619
0112290501 1.440±0.030 17±2 1.37±0.02 16±2 0.094±0.007 646
0112290601 1.410±0.020 5±2 1.34±0.02 5±2 0.091±0.006 638
0112291201 − − − − − −
0302900101 1.070±0.004 8.8±0.5 0.995±0.004 8.1±0.5 0.094±0.002 624
NGC5408X-1 0500750101 1.000±0.006 12.9±0.7 0.924±0.006 11.3±0.8 0.102±0.002 625
0653380201 1.200±0.004 7.0±0.5 1.110±0.004 6.9±0.5 0.111±0.002 620
0653380301 1.182±0.004 6.8±0.4 1.091±0.004 6.8±0.5 0.109±0.001 618
0653380401 1.123±0.004 7.7±0.5 1.038±0.004 6.9±0.5 0.105±0.002 621
0653380501 1.084±0.004 9.0±0.5 0.997±0.004 7.9±0.5 0.107±0.002 619
0112520601 3.370±0.030 3±1 2.97±0.03 4±1 0.40±0.01 614
0112520701 3.100±0.100 621 2.6±0.1 625 0.43±0.02 622
HoIIX-1 0112520901 0.880±0.020 2±6 0.81±0.02 2±7 0.090±0.006 640
0200470101 3.390±0.010 2.6±0.9 2.98±0.01 3.8±0.8 0.423±0.005 613
0561580401 1.330±0.010 21.4±0.7 1.201±0.009 21.5±0.8 0.143±0.003 618
aCalculatedfromthebackgroundsubtractedEPIC-pnlightcurves,with200stimebins.Valueswithouterrorbarsindicate3σupperlimits.
fractionalisbetweenafewpercentand∼30%,withmarginalevi- elsof short-termvariability.Intheother twocasesthevariability
denceforhighervaluesintheharderband.However,thevariability is less than 10%. Therefore, we do not find significant evidence
isnotclearlycorrelatedwithanyspecificspectralregime,varying forthiseffect,indicatingthatinsoftultraluminousstatethewind
almostrandomlyacrossdifferentobservationsandfromsourceto openinganglemightassumealargerrangeofvalues.
source. Suttonetal. (2013) found a possible correlation between
Fvar and the spectral shape, showing that a higher variability is FinallywementionthatinthelastobservationofIC342X-1
seenduringthesoftultraluminousstate,i.e.whenthesoftcompo- thevariabilityisdefinitelylargerthantheaverageF (∼ 10%),
var
nent isdominant. They suggested that such short termvariability and stronger at higher energies (∼10% at 0.3−2.0 keV against
isrelatedtoturbulencesat theedgeof theoutflow ejectedbythe ∼ 30%at2.0−10.0keV;Table4).Thisisprobablycausedbya
discandwouldthenbelargerwhenourlineofsightintersectsthe significantdropinfluxoccurringduringtheobservation.Asimilar
edgeofthewind.Thiswouldimplythathighlevelsofvariability behaviourwasobservedalsointhelastobservation(#5)ofHolm-
canbeseenonlyinfavourableepochswhenthewindopeningan- bergIIX-1,duringwhichthesourceswitchesfromahightoalow
glecrossesourlineofsight.AccordingtoSuttonetal.(2013),the flux level with a decrement of a factor of 2 (Kajavaetal. 2012).
occurrence of this condition is more likely in the soft ultralumi- Ontheotherhand,NGC253X-1showsalsohigherthanaverage
nousstate,whenthewindismoreextendedanditsopeningangle F butinthiscasethevariabilityisintrinsicand,notably,notre-
var
smaller. However, among the three sources of our sample with a latedtoanysignificantfluxchange.Thisfindingmakesthissource
strongersoftcomponent,onlyNGC5408X-1showsveryhighlev- differentfromtheotherULXsofoursample(seenextsection).
UltraLuminousX-raySources:a deeper insightintotheirspectralevolution 9
3.5 Lookingatsinglesources 0.6
∆T=200 s
The spectral analysis reported above shows that the spectra of 0.5
ULXs can be well described in terms of a disc plus comptonisa-
tionmodelinwhichtheComptonisingcomponentisusuallyopti- 0.4
ceanltlyretghiiocnksaonfdthceooτl.−ThkeTsourpcleasnoec(ctuhpicykparnedfevreenrytiathlliycktwstoatdei)f.feICr- Fvar 0.3
cor
342X-1andNGC253X-1populatetypicallytheverythickstate, 0.2
while Holmberg II X-1, Holmberg IX X-1, NGC 5204 X-1 and
0.1
NGC 5408 X-1 stay predominantly in the thick state. Only NGC
1313X-1andX-2appeartocrossconvincinglybothregions,while 0
1 10
HolmbergIXX-1andNGC5204X-1maydoitonlymarginally. Energy (keV)
As mentioned above, NGC 5408 X-1 and Holmberg II X-1
showprominentsoftcomponentsintheirspectra.InNGC5408X- Figure 3. NGC 253 X-1: RMS fractional variability spectrum evaluated
1 thesignificance of the soft component isslightlydependent on onbackgroundsubtractedEPIC-pnlightcurvesofobservation0152020101,
the total luminosity. At variance with the general trend, the soft sampledwithtimebinsof200s.Afitwithaconstantvalue(Fvar∼28%)
component contributes almost 40-50% of the total flux (L ∼ isconsistentwiththespectrum.
disc
(3 − 4) · 1039 erg s−1) when the source is in a low luminos-
ity state and becomes less significant (∼ 25% of the total flux)
2.4
when the luminosity increases (L ∼ 6−7·1039 erg s−1).
disc
Wealsofoundevidenceforadiscluminosity-temperaturerelation
2.2
of thetype L ∝ T1.8±0.8. Short termvariabilityisgenerally
disc disc
presentatlowenergiesbutessentiallyunconstrainedathighener-
2
gies because of the low statistics2. The presence of a strong soft V)
e
cmoamypboeneenxtpalanidnesdignwiifithcainnttshheosrct-etneramriovainriawbhiliictyhitnheNGwCind54c0o8mXpo-1- T (kdisc 1.8
nent becomes very extended, its opening angle narrower and its k 1.6
edgeintersectsourlineofsight,consistentwithwhatproposedby
Middletonetal.(2011)bandSuttonetal.(2013). 1.4
InHolmbergIIX-1,wefoundadiscluminosity-temperature
relation of the type L ∝ T3.6±1.4. While the correlation of 1.2
disc disc 0.45 0.5 0.55 0.6 0.65 0.7 0.75
Holmberg II X-1 may be reminiscent of a standard disc relation,
p
we do not interpret it in this way because in the model adopted
here the corona is optically thick and the temperature of the soft Figure 4. NGC253 X-1: inner disc temperature versus p-index for the
componentreferseithertotheoutervisiblepartofthediscortothe diskpbbmodelfit.
wind.
bymaterialthatoriginatesfromtheimpactoftheaccretionstream
3.5.1 NGC253X-1 ontothediscitselforbyalow-density,equatorialwind(compare
withSuttonetal.2013).Thismaterialmayberesponsiblealsofor
ThebehaviourofNGC253X-1appearsslightlydifferentfromthe theshortdecrementsofthecountrateobservedbyBarnard(2010)
other ULXs of our sample. It shows a significantly curved spec- (Figure2inhispaper),thatmayappearsimilartothedipsobserved
trumanditsshorttermvariabilitymayreach∼20-30%.Asshown insomeXRBs(e.g.White&Swank1982;D´ıazTrigoetal.2006).
in Table 4, the RMS fractional variability in the highest quality In this scenario the source would be seen at rather large inclina-
observation (which is also the more luminous) is ∼ 30% in the tions.
0.3−10keVenergyband.Thecountingstatisticsofobservations
#2 and 4 allows us to study the RMS spectrum of the source,
Table5.Best fitting spectral parameters ofNGC253X-1obtained with
that was calculated by selecting background subtracted EPIC-pn
theabsorbedslimdiscmodel(diskpbbinXspec).Errorbarsareat90%for
lightcurvesintheenergyranges0.3-0.5, 0.5-0.7, 0.7-1.0,1.0-1.3, eachparameterofinterest.
1.3-1.6, 1.6-2.1,2.1-4.0, 4.0-10keV (Figure3).Whenfittedwith
a constant the spectrum of the fractional variability is consistent
withavalueof ∼ 28%. Thesefindingssuggest thattheobserved NGC253X-1
emission may come from asingle component. However, we note No. NHa pb kTdiscc LX[0.3-10keV]d χ2/dof
that,althoughnotstatisticallysignificant,inobservation#2there (1021cm2) (keV) (1038ergs−1)
mbaanydbaensdom∼e4h0in%toifnvathrieab2i.l0it-y1:0FkveaVribsa∼nd25(s%eeinaltshoeS0.u3t-to2n.0ekteaVl. 12 10..496+−+−0000....00001144 0.05.858+−+−00..0005..00010011 12.5.2241+−+−0000....0000220044 82.37+−+−0011..68 815167..2811//89192
2013).Thisfact maybeconsistent withtheexistenceof anaddi- 34 06.4+−00.02..11 00..675023+−+−0000....00001111 11..2452+−+−0000....00001212 71.00+−+−0011..76 759.08/89/375
tional component affecting the high energy part of the spectrum. 5 1.2+−00..22 0.581+−00..000033 1.98+−00..0022 13+−29 88.97/85
Ifweinterpretitasanabsorptioncomponent,itmaybeproduced 6 1.0+−00..11 0.620+−00..000033 1.66+−00..0011 13.0+−11 174.94/163
7 0.9+−00..22 0.635+−00..0011 1.59+−00..0022 13.0+−22 44.38/47
2 Different values ofthefractional RMS ofNGC5408X-1were found aColumndensity;bexponentoftheradialdependenceofthedisctempera-
byCaballero-Garciaetal.2013,probablybecauseofthedifferentchoiceof ture;cdisctemperature;dunabsorbedtotalX-rayluminosityinthe0.3-10
thetimesampling. keVrange;
10 FabioPintore,Luca Zampieri,AnnaWolter,TomasoBelloni
NGC 253 X-1 differs from the other sources of our sample instrumentshavedifferentresponsematrices,becauseofthehigher
alsobecauseitsX-rayspectrumcanbewelldescribedbya mod- throughput we used only EPIC-pn data and selected four energy
ified/slimdiscmodel.Indeed,thissourcewasclassifiedasbroad- bands:0.3-0.7,0.7-2.0,2.0-4.0and4.0-10keV,definedforsimplic-
eneddiscbySuttonetal.(2013).Themodified/slimdiscmodelin- ityas1,2,3and4,respectively.Thischoiceprovidesanadequate
corporatestheeffectsthatareexpectedtosetinatorslightlyabove samplingofthespectralregionsinwhichthesoftexcess,spectral
theEddington limitand wasmodelledthrough adiskpbb compo- pivoting(seeKajava&Poutanen2009;Pintore&Zampieri2012)
nent3. Although thefitsaregenerally statisticallyacceptable (Ta- and/orthecomptonizingcomponentareusuallycontributing.Inad-
ble5),wefoundoneobservations(#4)inwhichthecolumndensity dition,itguaranteesasimilarcountingstatisticsineverybandand
pegged at 0 indicating that the absorption was anomalously low. itallowsustodiscriminateinaclearerwaythedifferencesinthe
However,fixingthecolumndensityattheaveragevalueoftheother spectralpropertiesofoursources(seebelow).Ineachenergyrange
observations,thediskpbbmodelprovidesstillagooddescriptionof ofinterest,weaddedthecountsofeverygoodchannelofthespec-
thespectrum(χ2/dof =80.61/76,kT ∼1.76andp∼0.6). trum and subtracted the total counts of the corresponding back-
disc
ThevariationoftheinnerdisctemperaturekT withtheparame- ground channels, properly scaled for the extraction areas. Count
disc
terpofthediskpbbmodelisshowninFigure4.Thereisatendency rates are evaluated dividing the counts by the net EPIC-pn GTI
forthep-indextodecreasewithincreasingdisctemperature,sothat per observation andare thenrescaled toadistance of 1Mpc (er-
thehigheristhetemperaturethemoreadvectiondominatedisthe rorsonthedistancesarenottakenintoaccount).WeaddtheNGC
disc.Atlowluminosity(observation#1,3),thespectrumcanalso 1313 X-1 and X-2 data (Pintore&Zampieri 2012) to the sample
bewelldescribedintermsofacanonicaldiskbb+powerlaw (with forcomparison,findingasimilarspectralevolution.
kTdisc ∼ 0.3keVandΓ ∼ 2.2,χ2/dof = 92.80/98) ordiskbb Figure5-topshowsacolour-colour diagraminwhichthey-
spectral model (with kT = 1.1 keV, χ2/dof = 92.29/94; axisisdefinedastheratiobetweenthecountsintheenergybands4
disc
see also Kajava&Poutanen 2009). Hence this source, which is and(2+3)andthex-axisastheratiobetweenthebands1and(2+3).
also the faintest of our sample, may be in a state similar to the ThesourcesliealongasequencestartingfromIC342X-1andend-
soft/steeppowerlawstateofGalacticXRBsandmayswitchtothe ingwithNGC5408X-1.ApartfromIC342X-1andlettingaside
ULXregimeonlyathighluminosity. thedetailsoftheevolutionofsinglesources,wenotethatthescat-
terisnotverylarge.Thismaysuggestthatdifferencesinducedby
thelikelydiverseBHmassesandinclinationofthesourcesarenot
4 COLOURS sostrongandthattheBHmassesthemselvesarenotverydifferent.
Atleasttwogroupsofobservationscanbeidentifiedonthecolour-
InthepreviousSectionwehaveshownthatalmostallthespectra colourplot:group1with1/(2+3)∼0.1−0.3andgroup2with
oftheULXsofoursamplecanbedescribedbyacombinationofa 1/(2+3)∼0.4−1.1.Ingroup1,therearethelessluminousand
softcomponentandacomptonisationmodel.Thespectralparame-
most absorbed sources: NGC1313 X-1andX-2, HoIXX-1and
tersspandifferentrangesbut,ifinterpretedatfacevalue,indicate
NGC253X-1(althoughoneofitsobservationsisalsoconsistent
similarphysicalconditions,suchasanopticallythickcoldcorona
with belonging to group 2). The observations of NGC 1313 X-2
andacooldisc.However,thelowestcountingstatisticsspectrado
areclearlysplitintwosubgroupswhicharereminiscentofthevery
notprovidestrongconstraintsonthespectralfitsbecauseofthede-
thickandthickstatesidentifiedinpreviousworks(Feng&Kaaret
generacyofsomespectralparameters,especiallyifthetemperature
2009; Pintore&Zampieri 2012). A similar trend can also be ob-
ofthediscandthatofthesoftphotoninputarefreetovaryinde-
servedforNGC1313X-1andHolmbergIXX-1.
pendently. Infact,thissituationisrathercommonformostofthe
Group2containsthemostluminousandlessabsorbedsources
availableXMM-NewtonobservationsofULXsandalsoforsimpler − NGC 5204 X-1, Ho II X-1 and NGC 5408 X-1 − where a
models, as the diskbb+powerlaw for example. Hence we tried to
strongsoftcomponentisseen.Usingadifferentspectralselection
supportthefindingsofthespectralanalysisusingacomplementary
andanalysis,Suttonetal.(2013)classifiedULXsinthreespectral
approachbasedonthehardnessratios.Thistechniqueisverypow-
states/groupsof sources that theycalledbroadened disc,hard ul-
erfulforlowcountingstatisticsdataandappearsthenparticularly
traluminousandsoftultraluminous.Wesuggestthatthebroadened
suitableformanyobservationsofULXs.Inthissectionwewillre-
discandhardultraluminoussourcescanbeidentifiedwithgroup1
consideralltheobservationsusingthisapproach,thatinthefuture
andcannotbeeasilydistinguishedastheirX-raycoloursaresim-
maybeadoptedalsoforanalysinglargersamplesoflowercounting
ilar.Ontheotherhandthesoftultraluminousstatecanbeassoci-
statisticsdata.
atedtogroup2.Therefore,thespectralgroupsfoundviaadetailed
The method of the hardness ratios or colour diagrams has
spectralanalysiscanbeidentifiedequallywelladoptingamodel-
been successfully adopted in the past to study the behaviour of
independent colour-basedapproach, makinguseofcolour-colour
XRBs and, more in general, of X-ray sources (Maccacaroetal.
diagramsincrediblypowerfulalsoforpoorqualitydata.Ingroup
1988). Extensive monitoring of some Galactic BH binaries with
2, NGC 5204 X-1 shows variations inthe 4/(2+3) colour up toa
Rossi-XTEledtothediscoveryofacommonevolutionarypathin factorof3while1/(2+3)staysnearlyconstant(∼ 0.5),indicating
thehardness-intensitydiagram,thehysteresiscycle,thatalltheBH
either that the high energy component issignificantly variable or
binary systems accreting at sub-Eddington rates appear to follow
thatthesoftbandsvarytogether.Ontheotherhand,the1/(2+3)
(see for example Belloni 2010). This cycle describes how the X-
colour of NGC 5408 X-1 changes by almost a factor of 2, while
rayspectrumchangeswiththesourceintensity.
4/(2+3)remainsnearlyconstant,meaningthatvariabilityismostly
Ahardness ratiocanbedefinedasB /B ,inwhichB and
i j i inthesoftcomponentorthattheharderenergybandsvarytogether.
B arethetotalcountsintwogivenenergybands.Sincedifferent
j Theonlysourcethatappearsclearlyseparatedfromtheothers
inthecolour-colourdiagramisIC342X-1.Weknowthat,interms
3 ThismodeldependsontheinnerdisctemperaturekTdiscandaparam- ofspectralparameters,IC342X-1showssimilaritieswiththevery
eterpthatdescribestheradialprofileofthedisctemperatureasTdisc ∝ thickstateofNGC1313X-2.Ontheotherhand, IC342X-1has
r−p,wherep=0.75forastandarddiscandp=0.5foraslimdisc. thelargestintrinsicNH ofthewholesampleandhenceitsposition