The Statistical Properties of Galaxies Containing ULXs

0.17 MB·English

by  A. Ptak

Accepted by ApJ
PreprinttypesetusingLATEXstyleemulateapjv.11/26/03
THE STATISTICAL PROPERTIES OF GALAXIES CONTAINING ULXS
A. Ptak1, E. Colbert1,2
Accepted by ApJ
ABSTRACT
We present a statistical analysis of the properties of galaxies containing ultraluminous X-ray objects
(ULXs). Our primary goal is to establish the fraction of galaxies containing a ULX as a function of
ULX luminosity. Our sample is based on ROSAT HRI observations of galaxies. We find that 12%
ofgalaxiescontainatleastoneULXwithL >1039 ergs−1and 1%ofgalaxiescontainatlea∼stone
X
ULX withL >1040 erg s−1. These ULXfrequencies arelowerl∼imits since ROSAT HRI observation
4 X
would miss absorbed ULXs (i.e., with N &1021 cm−2) and those within 10” of the nucleus (due
0 H
∼
0 to the positional error circle of the ROSAT HRI). The Hubble type distribution of galaxies with a
2 ULX differs significantly from the distribution of types for nearby RC3 galaxies, but does not differ
significantly from the galaxy type distribution of galaxies observed by the HRI in general. We find
n
no increase in the mean FIR luminosity or FIR / K band luminosity ratio for galaxies with a ULX
a
relative to galaxies observed by the HRI in general, however this result is also most likely biased by
J
thesoftbandpassoftheHRIandtherelativelylownumberofhighSFRgalaxiesobservedbytheHRI
4
with enough sensitivity to detect a ULX.
2
Subject headings:

1
v
5 1. INTRODUCTION is determined, most often based on Tully (1988).
2 ROSAT HRI observations showed that ultraluminous Colbert & Ptak (2002) chose a recessional velocity cut-
5 X-ray objects (ULXs, a.k.a. IXOs), which were pre- off of 5000km s−1. In this paper we assume the cosmol-
1 viously detected with Einstein (see Fabbiano 1989), ogyH0=75kms−1Mpc−1 andq0=0.5(forconsistency
0 withTully1988),andthereforethisvelocitycorresponds
are compact X-ray sources, are quite common in
4 to 66.7 Mpc. This resulted in a galaxy sample contain-
the local Universe, and are typically offset from the
0 ing 9452 galaxies. We next determined which of these
optical galactic nucleus (Colbert & Mushotzky 1999;
/ galaxieshad“coverage”by computing the overlapofthe
h Roberts & Warwick 2000). Here we define ULXs as X-
RC3 D ellipse and the HRI FOV for each observation
p raysources in galaxiesthat are point-like,extra-nuclear, 25
- and have X-ray luminosities in excess of 1039 ergs s−1. that was used in Colbert & Ptak (2002). We considered
o a galaxy as having coverage if 90% of the galaxy was
Thereisconsiderabledebateastonatureofthesesources
r in the FOV. We found that 76≥6 galaxies had coverage,
t (see van der Marel 2003; Miller & Colbert 2003, for re-
s or 8% of the RC3 subset.
views), with popular explanations including anisotropic
a ∼
In order to properly assess the frequency of ULXs, we
: emissionfromstellar-massX-raybinariesandintermedi-
v need to assess how many galaxies were observed by the
atemassblackholes. Aconstraintonanyexplanationfor
i HRIwithsufficientsensitivitytodetectaULX.Wecom-
X thesesourceswouldbethefrequencyatwhichtheyoccur,
puted the flux limit at the position of each RC3 galaxy
i.e.,whatfractionofgalaxiesharboraULXasafunction
r with coverageby computing the backgroundlevel in the
a of ULX luminosity. For example, this frequency would
point source region size. The point source region used
haveimplicationsforthetypicalrelativisticboostingfac-
by XAssist (Ptak & Griffiths 2003) is a circle with a ra-
torforjetmodelsofULXs(Koerding, Falcke, & Markoff
dius given by 4[exp(θ/9)+1] arcseconds, where θ is the
2002). In this paper we examine the frequency of ULX
off-axisangleofthesourceinarcminutes. Thetotalnum-
occurrenceasafunctionofULXluminositybasedonthe
ber of background counts, b, in this area was then de-
HRI ULX catalog published in Colbert & Ptak (2002).
termined using the median background counts per pixel
While HRI observations suffer from significant biases
(also computed as a part of the XAssist processing of
against the detection of ULXs (as discussed below), we
each field). The 3σ detection threshold count rate is
note that the relatively large field of view of the HRI
resulted in a large number of galaxies having been ob- given by 3√b/t, where t is the exposure time, and the
served, and of course large area surveys are crucial for limiting flux isthen 3√bf v(θ)/t wherev(θ) givesthe
conv
establishing the bright end of a luminosity distribution. vignetting correction and f is the on-axis count rate
conv
to flux conversion factor. f was computed using the
conv
2. METHODOLOGY FTOOLPIMMS, assuminga power-lawslope of1.7 and
the Galactic column density (Dickey & Lockman 1990,
In addition to Colbert & Ptak (2002), we also usingtheFTOOL“nh”). Wecomputedthe limitingflux
made use of a subset of the RC3 galaxy cata- inboththe0.5-2.0keVbandpassF andthe
lim,0.5−2.0 keV
log (de Vaucouleurs et al. 1991) where the distance 2-10keVbandpassF . WeusedF
lim,2−10 keV lim,2−10 keV
to compute the limiting luminosity since the 2-10 keV
1 The Johns Hopkins University, Homewood Campus, Balti-
bandpasswasusedin Colbert & Ptak (2002), for consis-
more,MD21218,[email protected]
2 The Catholic University of America, 620 Michigan Ave NE, tency with current Chandra and XMM-Newton surveys.
Washington, DC20064, [email protected]
2 Ptak & Colbert
Any ULX could then have been detected in a galaxy
when L is 1039erg s−1, while more gener-
lim,2−10 keV
≤
ally L is >1039erg s−1.
lim,2−10 keV
Finally, it is necessary to determine how many ULXs
are likely to be background sources. For this pur-
pose we used the logN-logS given in Hasinger (1998),
which we note is consistent with a simple power-law
logN(> S) 100(S )−1.4 sources deg−2, where S
−14 −14
is the 0.5-2.0∼keV flux in units of 10−14 ergs cm−2 s−1.
The number of expected background sources is then
n = 100(F lim,0.5−2.0 keV)−1.4πr2, where r is the ma-
xrb 10−14
joraxisradiusofeachgalaxyindegrees. Colbert & Ptak
(2002)usedr =D (theRC3majoraxisdiameter)how-
25
ever here we also consider r = 0.5D . Also note that
25
when L is 1039erg s−1, the relevant num-
lim,2−10 keV
≤
berofbackgroundsourcesisthenumberofsourcewhose
apparentluminosityexceeds1039ergs−1,andsinceweare
assuming a simple power-law logN-logS this simply en-
tails decreasing the total number of background sources
expected by the factor (L /1039erg s−1)1.4.
lim,2−10 keV Fig. 1.— The frequency of ULX occurrence in galaxies as a
function of ULX luminosity. The frequency of ULXs is plotted
3. RESULTS using effective galactic radii of D25 (blue diamonds with dashed
errorbars)and0.5D25 (redcrosseswithsoliderrorbars). Wealso
3.1. ULX Frequency plotthefrequencyobservedafterlimitingthesampletoonlyspirals
(with Hubble type >= 0and radius = 0.5D25; black filledcircles
InTable1weshowthenumberofgalaxies(N )
Llim≤LX withdotted errorbars).
with HRI coverage and a limiting 2-10 keV luminosity
less than or equal to L ranging from 1039erg s−1to
X
1041erg s−1(i.e., the number of galaxies where a ULX
tive radius for the galaxies does not significantly im-
as bright as L or brighter could have been detected).
X
pact the estimate of ULX frequency once background
We also show the number of galaxies with at least one
sources are taken into account. The frequency of ULXs
ULX(N ), andthenumberofgalaxieswithatleast
wULX
observed by the HRI evidently is 12% and 1% of
one ULX whose luminosity exceeds the limiting lumi-
galaxies harboring a ULX with L ∼>= 1039erg∼s−1and
nosity (NLULX≥LX). This latter quantity, NLULX≥LX, L >= 1040erg s−1, respectivelXy. We also plot the
is of interest since it represents the number of galaxies X
galaxyfractionsforspiralsonly(i.e.,HubbletypeinRC3
with a ULX at least as bright as the list L , and we
X
0). The Hubble type distribution ofgalaxieswith and
then computed the resultant number of galaxies after
≥
withoutaULXisdiscussedbelow,howeverwenote that
subtracting the expected number of background sources
thefractionofgalaxiesharboringaULXisslightlyhigher
(N , i.e., the number of galaxies where the
net,LULX≥LX in spirals.
1-σ lower limit on the number of ULXs is greater than
0.;thelowerlimitwascalculatedusingthe mehologyde-
3.2. Hubble Type Distribution
scribed in Kraft, Burrows, & Nousek (1991), with the
background being the expected number of background In Figure 2 we show the Hubble type distribution of
sources). Finally we list the values N , N , the RC3 subsample (discussed in 2 ), the Hubble type
wULX LULX≥LX §
and N using both r = D and r = 0.5D . distributionofgalaxiesthathadHRIcoverage(hereafter
net,LULX≥LX 25 25
Note thatinthe latter casethe totalnumber ofULXs in HRI galaxies)andthe HRI galaxiesthatcontainatleast
the Colbert & Ptak (2002) sample is reducedfrom 87 to one ULX. The ULX galaxy type distribution is shown
49. both before and after background subtraction, and here
InTable1showsthatalargefractionofULXsdetected we only consider r <0.5D ULXs. There is a slight ex-
25
at radii larger 0.5D are consistent with background cess of elliptical ULX galaxies relative to the RC3 sam-
25
sources,i.e.,the netnumber ofgalaxieswith ULXs is of- ple however the larger excess in the HRI galaxy sam-
ten larger when 0.5D is chosen as the limiting radius. ple suggests that this bias due in part to a bias toward
25
Also,insomecases,thenumberofULXsdetectedwithin early type galaxies in the HRI observing patterns rela-
the0.5D radiusisabovethebackgroundlevelwhenr = tive to RC3 galaxies in general. The most conspicuous
25
0.5D but is not above the background level when r = feature of the ULX galaxy distribution is the excess of
25
D (i.e., since inthe latter casethe expectednumber of T=5-7 galaxies, corresponding to Sc and Scd galaxies.
25
background sources is a factor of four larger). In Figure However, in general the ULX galaxy type distributions
1 we plot the fraction of galaxies harboring a ULX as a are not significantly different statistically from the HRI
function of ULX luminosity (N /N ) RC3 galaxy type distributions, with the χ2 tests giving
net,LULX≥LX Llim≤LX
for both r = D and r = 0.5D . Here the errors were χ2/dof 15/8, which can only be rejected at the 95%
25 25
∼
computed according to the methodology described in level.
Kraft, Burrows,& Nousek (1991), with the background We also investigated the possible connection between
taken as the number of galaxies at each luminosity that star-formation rate (SFR) and ULX frequency by tabu-
are consistent with all ULXs being background sources lating the far-infrared luminosities (a well-known SFR
(i.e., N N ). The choice of effec- indicator; Kennicutt 1998) of the galaxy samples dis-
LULX≥LX − net,LULX≥LX
HRI ULX Statistics 3
Table1. GalaxyNumberCounts
r=D25 r=0.5D25
LX NLlim≤LX NwULX NLULX≥LX Nnet,LULX≥LX NwULX NLULX≥LX Nnet,LULX≥LX
(1) (2) (3) (4) (5) (6) (7) (8)
1 228 45 45 25(8.7%) 34 34 28(12.3%)
2 316 49 38 23(7.3%) 37 27 23(7.3%)
5 446 53 23 13(2.9%) 38 14 14(3.1%)
10 540 54 11 7(1.3%) 38 6 6(1.1%)
50 736 54 1 1(0.1%) 38 0 0(0.0%)
100 762 54 0 0(0.0%) 38 0 0(0.0%)
Note. — (1) minimum ULX 2-10 keV luminosity in 1039erg s−1. (2) NLlim≤LX = number of galaxies with a limitingluminosity Llim less
thanorequaltoLX. (3)and(6)NwULX =numberofgalaxieswithLlim ≤LX andaULX,forr=D25 andr0.5D25, respectively. (4)and(7)
NLULX≥LX =numberofgalaxieswithaLlim≤LX andaULXwithluminosityLULX ≥LX. (5)and(8)Nnet,LULX≥LX =numberofgalaxies
with a Llim ≤LX and a net numberof ULXs with luminosityLULX ≥ LX > 0 after subtractingthe expectednumberof backgroundsources,
withthepercentageofgalaxies(i.e.,Nnet,LULX≥LX/NLlim≤LX)giveninparentheses.
Fig. 2.— (left) The Hubble type distribution of nearby galaxies in the RC3 catalog (black solid line), the galaxies with HRI coverage
(reddashed line), the galaxies with aULX(blue dot-dashed line)and the galaxies withaULX after excluding galaxies with atleast one
“net” ULX (i.e., after subtracting the expected number of background sources; blue dotted line). (right) The Hubble type histograms
forgalaxies with HRIcoverage (red, soliderrorbars) and with a“net” ULX (blue, dotted error bars)shown with errorscomputed using
Kraft,Burrows,&Nousek (1991). The galaxy types T correspond to: ellipticals have T =-5 to T=-4, lenticulars have T=-3 to T=-1,
spiralshaveT=0(S0)toT=9,withoddnumbersfrom1to9representingSa,b,c,dandevennumbersrepresentingSab,bc,cd,dm. Irregulars
haveT=10-11andweassigned16togalaxieswithT>11. BothplotswerecreatedusingonlyULXswithinr=0.5D25.
cussed above. These distributions are shown in Fig- sity of 6.7 10−21erg cm−2 s−1Hz−1 (Zombeck 1990)3.
×
ure 3, where we used NED IRAS fluxes densities at 60 The K band luminosity is a proxy for stellar mass since
and 100 µm to compute F = 1.26 10−23[2.58 the variation in mass-to-light ratio for K band is only a
FIR
10−12f +10−12f ]. The peak o×f the distribu×- factor of 2 for various galaxy types (Bell & de Jong
60µm 100µm
∼
tions with a ULX is slightly skewed toward higher FIR 2001). These distributions are shown in Figure 4 and
(i.e.,nearlogL =43.5)howeverthemeansofthedis- again we do not see any significant differences between
FIR
tributions do not differ significantly (all are in the range galaxieswithULXs andgalaxiesobservedbythe HRI in
log L = 42.7-43.0). The HRI galaxy FIR luminos- general.
FIR
ity distribution differs from the “net” ULX galaxy FIR
luminosity distribution at less than the 90% level based 4. SUMMARY&CONCLUSIONS
on the χ2 test. Restricting the sample to spiral galax-
We have assessed the frequency of ULX occurrence
ies (T >= 0) shifted the means by only logL 0.1,
FIR
∼ in galaxies as a function of ULX luminosity, including
reflecting the predominance of spirals in these samples.
properassessmentofthecontaminationfrombackground
Similarly we plot the distribution of L /L where
FIR K
sources. Wehavefoundthatassuminganeffectiveradius
L is the K band luminosity computed using the
K
ofD tendstoresultinasignificantcontaminationfrom
2mass Extended Source catalog K band magnitude (see 25
Jarrett et al. 2003) and assuming a zero-point flux den-
3 seealso
http://www.ipac.caltech.edu/2mass/releases/allsky/doc/sec6 4a.html
4 Ptak & Colbert
Fig. 3.—TheFIRluminositydistributionofnearbygalaxieswithIRASdetections at60µmand100µm(blacksolidline),thegalaxies
withHRI coverage (red dashed line), and the galaxies witha ULX before (blue dot-dashed line) and after (blue dotted line) background
subtraction. The left panel shows the results without selecting on galaxy type and the right panel shows the results when selecting only
spiralgalaxies(Hubbletype≥0).
Fig. 4.—TheFIR/KbandluminositydistributionsfornearbygalaxieswithIRASdetectionsat60µmand100µmand2massextended
sourcecatalogKbanddetections (blacksolidline). OthercurvesareasinFigure3. Theleftpanel showstheresultswithoutselectingon
galaxytypeandtherightpanel showstheresultswhenselectingonlyspiralgalaxies(Hubbletype≥0).
backgroundsources. Wefoundfrequenciesof 12%and result is based in part on Chandra data and uses lumi-
1% for ULX luminosities of at least 1039e∼rg s−1and nosities in the 0.3-10.0 keV bandpass as opposed to our
1∼040erg s−1. The Hubble type distribution of galaxies choice of the 2-10 keV band. However,as can be seen in
containingatleastoneULXissignificantlydifferentfrom Figure 2b, even when considering all ULXs in our sam-
those that do not contain a ULX in general, however it ple (i.e., with L 1 1039erg s−1), the frequency of
X
≥ ×
is not significantly different from galaxies observed by early-typegalaxiescontainingULXsisonlymarginallyin
the HRI. This implies that any bias in the Hubble type excess of zero (with ULXs occurring in globular clusters
distribution of galaxies observed by HRI with ULXs is asinthecaseofNGC1399contributingtotheearly-type
mostly due to the observing patterns of the HRI. Re- galaxy ULX frequency).
cently Irwin et al. (2004) reported on an analysis of The main caveat with an analysis of the HRI ob-
ULXsinearly-typegalaxiesandfoundthat“very”lumi- servations is the softness of its bandpass (i.e., 0.1-2.2
nous ULXs (with L > 2 1039erg s−1) are consistent keV). Objects absorbed by column densities in excess of
X
with the expected number×of background sources. This 1021 cm−2 are therefore difficult with the HRI. Since
∼
HRI ULX Statistics 5
columns of this order are typically observedthrough the sensitivity to detect a ULX, and 4 harbor a ULX (af-
disks of galaxies,it is likely that ULXs are being missed ter “background subtraction” and using a galaxy ra-
that are close to the nucleus of the galaxies and/or are dius of 0.5D ). This amounts to a larger ULX fre-
25
observed in edge-on disks. Chandra and XMM-Newton quency of 17+12%, although still statistically consistent
−8
observations will therefore be much more adept at de- with the full sample result. Recent Chandra surveys of
tecting ULXs in these types of environments. This is the point source luminosity functions in galaxies sug-
particularly true for Chandra whose much smaller PSF gest that, particularly at high star-formation rates, the
relative to the HRI and XMM-Newton allows nuclear point source luminosity function normalization is cor-
ULX to be discerned from AGN at smaller offsets (i.e., related with the SFR (Grimm et al. 2003, Colbert et
1” compared with 10”). Sipior (2003) finds that 8/41 al. 2004), and that the luminosity function is flatter
20% of nearby galaxies in a relatively unbiased Chan- in spirals (Colbert et al. 2004) or (similarly) in higher
∼
dra survey harbor a ULX. This suggests that our HRI SFR galaxies (Kilgard et al. 2002). Having both a flat-
survey is missing of order 40% of ULXs (albeit with ter luminosity function slope and larger normalization
∼
largeuncertaintyduetothesizeoftheChandrasample). at higher SFR strongly implies that higher SFR galax-
Hornschemeier et al.(2003)reportsthatthefrequencyof ies (i.e., with SFR greater than several to ten M yr−1
⊙
ULXs with LX >= 2 1039erg s−1is 8% locally and basedonColbertetal. 2004)wouldtendtoharbormulti-
× ∼
36% at z 0.1 (Hornschemeier et al. 2003) (suggest- pleULXs. Thisisalsosuggestedanecdotallybythelarge
∼ ∼
ing that the frequency of ULXs may evolve). We are numbers of ULXs found in merging galaxies such as the
similarly finding 8% (see Table 1) which implies that Antennae (Fabbiano, Zezas, & Murray 2001). Accord-
∼
any biases in our survey are not particularly large for inglythedistributionofL andL /L forgalaxies
FIR FIR K
LX >=2 1039erg s−1ULXs, within the errors. withat least oneULXmaynotbesignificantlyimpacted
×
The lack of any obvious connection between FIR lu- by an increase in the total number of ULXs in galaxies
minosity (and hence SFR) or the FIR/K band luminos- with large SFR. However the resolution of these issues
ity ratio distribution and the presence of a ULX sug- will ultimately require more detailed statistical analysis
gests that if such a connection exists, then the biases basedon Chandra and XMM-Newton wide-areasurveys.
discussed above are particularly acute in star-forming
galaxies. Thismaynotbesurprisinggiventhathigh-SFR
galaxies tend to contain significant amounts of molecu- This work was supported by NASA grant NAG 5-
lar clouds and dust which would obscure soft X-rays. 11670. We made use of the High Energy Archive and
In addition the relatively small size of our sample may ResearchCenter at NASA/GSFC and the NASA Extra-
also be impacting this result. For example, only 23 galacticDatabase. We thanksthe anonymousrefereefor
galaxies have logL >43.5 erg s−1and enough HRI useful comments.
FIR
REFERENCES
Bell,E.&deJong,R.2001, ApJ,550,212 Kilgard,R.etal.2002, ApJ,573,138
Colbert, E., Heckman, T., Ptak, A., Strickland, D. & Weaver, K. Koerding,E.,Falcke,H.,&Markoff,S.2002, A&A,382,13
2004, ApJ,inpress,astro-ph/0305476 Kraft,R.,Burrows,D.,&Nousek,J.1991,ApJ,374,344
Colbert,E.&Ptak,A.2002,ApJS,143,25 Miller,M.C.&Colbert,E.2003,InternationalJournalofModern
Colbert,E.&Mushtozky, R.1999,ApJ,519,89 PhysicsD,submitted,astro-ph/0308402
deVaucouleurs, G.,deVaucouleurs,A.,Corwin,H.G.,Jr.,Buta, Ptak,A.&Griffiths,R.2003,AstronomicalDataAnalysisSoftware
R. J., Paturel, G., & Fouqu, P. 1991, Third Reference Catalog and Systems XII ASP Conference Series, H. E. Payne, R. I.
ofBrightGalaxies(New York:Springer)(RC3) Jedrzejewski,andR.N.Hook,eds.,295,465
Dickey,J.&Lockman,F.1990,Ann.Rev.Ast.Astr.28,215 Roberts,T.&Warwick,R.2000,MNRAS,304,52
Fabbiano,G.1989,ARA&A,27,87 Sipior,M.2003,Ph.D.Thesis,PennsylvaniaStateUniversity
Fabbiano,G.,Zezas,A.&Murray,S.2001,ApJ,554,1035 Tully, R. B. 1988, Nearby Galaxies Catalog (Cambridge:
Gehrels,N.1986,ApJ,303,336 CambridgeUniv.Press)
Grimm,H.,Gilfanov,M.,Sunyaev, R.2003, MNRAS,339,793 van der Marel, R. 2003, Carnegie Observatories Astrophysics
Hasinger,G.1998, AN,319,37 Series, Vol. 1: Coevolution of Black Holes and Galaxies, ed.
Hornschemeier,A.etal.2003,ApJ,inpress,astro-ph/0308408 L. C. Ho (Cambridge: Cambridge Univ. Press), in press,
Irwin, J., Bregman, J., & Athey, A. 2004, ApJ, in press, astro-ph/0302101
astro-ph/0312393) Zombeck,M.V.1990,HandbookofAstronomyandAstrophysics,
Jarrett, T., Chester, T., Cutri, R., Schneider, S. E.,& Huchra, J. SecondEdition(Cambridge,UK:CambridgeUniversityPress)
2003, AJ,125,525
Kennicutt, R.1998,ARA&A,36,189