The Ultraviolet Spectral Energy Distributions of Quiescent Black Holes and Neutron Stars

0.59 MB·English

by  R. I. Hynes

The Ultraviolet Spectral Energy Distributions of Quiescent Black
Holes and Neutron Stars
2 R.I. Hynes
1 Department of Physics and Astronomy, Louisiana State University, Baton Rouge, Louisiana 70803
0
[email protected]
2

n
a E. L. Robinson
J
Department of Astronomy, The University of Texas at Austin, 1 University Station C1400, Austin, Texas
7
78712
2
[email protected]

]
E
H
.
h
ABSTRACT
p
-
We present HST/ACS ultraviolet photometry of three quiescent black hole X-ray transients:
o
r X-ray Nova Muscae 1991 (GU Mus), GRO J0422+32 (V518 Per), and X-ray Nova Vel 1993
t
s (MM Vel), and one neutron star system, Aql X-1. These are the first quiescent UV detections
a of these objects. All are detected at a much higher level than expected from their companion
[
stars alone and are significant detections of the accretion flow. Three of the four UV excesses

1 can be characterizedby a black body of temperature 5000−13,000K, hotter than expected for
v the quiescent outer disk. A good fit could not be found for MM Vel. The source of the black-
0
body-likeemissionismostlikelyaheatedregionofthe innerdisk. Contrarytoinitialindications
8
from spectroscopy there does not appear to be a systematic difference in the UV luminosity or
6
5 spectral shape between black holes and neutron star systems. However combining our new data
. with earlier spectroscopy and published X-ray luminosities there is a significantdifference in the
1
0 X-ray to UV flux ratios with the neutron stars exhibiting LX/LUV about 10× higher than the
2 black hole systems. This is consistent with earlier comparisons based on estimating non-stellar
1 optical light, but since both bandpasses we use are expected to be dominated by accretion light
:
v we present a cleaner comparison. This suggests that the difference in X-ray luminosities cannot
i simplyreflectdifferences inquiescentaccretionratesandsothe UV/X-rayratioisamorerobust
X
discriminatorbetweenthe blackholeandneutronstarpopulationsthanthe comparisonofX-ray
r
a luminosities alone.
Subject headings: accretion, accretion discs—binaries: close—stars: individual: GU Mus—stars: indi-
vidual: V518 Per—stars: individual: MM Vel—stars: individual: AqlX-1
1. Introduction star, typically a dwarf or sub-giant of solar mass
or less accretes onto the compact object through
The discovery that many transient X-ray
an accretion disk. The combination of large dy-
sources contain stellar mass black holes has pro-
namicrangeofbehaviorobservedintheseobjects,
vided many opportunities to study the astro-
and accessible timescales on which to study their
physics of how black holes accrete matter. These
variations through this range are unique to the
black hole X-ray transients (BHXRTs) and their
stellar mass objects but may also illuminate the
neutronstarcounterparts(NSXRTs)arelow-mass
snapshots of a particular state seen in supermas-
X-ray binaries (LMXBs) in which a companion
sive black holes.
1
One of the most interesting aspects of this CenX-4(McClintock & Remillard2000)andhave
study is to observe quiescent LMXBs in which confirmed this excess, as have more recent COS
the outer parts of the accretion disk are cool observations of A0620–00 (Froning et al. 2011).
and dim and the mass transfer onto the com- Intriguinglyinthetwoblackholesystemstheflux
pact object proceeds at an extremely low rate; droppedoffsteeplyintheUV,whereasinCenX-4
in XTE J1118+480 McClintock et al. (2003) in- it actually rose (in νFν) leading to the suggestion
ferred an accretion rate of only 10−8.5 of the Ed- thatthe shapeofthe UVspectrumis adiagnostic
dington limit. Accretion in this regime is ex- of the presence of a black hole or neutron star.
pected to proceed very differently as the inner Clearly such a conclusion based on a sample of
discisunstabletoevaporationintoaverticallyex- three UV spectra is unsatisfactory; but spectro-
tended, radiatively inefficient accretion flow. Be- scopic observations of other (fainter) objects be-
causeoftheradiativeinefficiency,itispossiblefor come prohibitively expensive, even with the Cos-
gas heated by viscosity to either deposit the heat mic Origins Spectrograph. To expand the sam-
on the neutron star surface or carry it through ple, we present here UV photometry performed
a black hole event horizon before it can radiate with HST/ACS of three more black holes, X-ray
away. In the neutron star case the energy will Nova Mus 1991 = GU Mus, GRO J0422+32 =
eventually escape whereas in the black hole case V518 Per, and X-ray Nova Vel 1993 = MM Vel,
it will not, leading to the naive expectation, ap- and one more neutron star, Aql X-1. To-
parentlysupportedbyobservations,thatquiescent gether with the three spectroscopic observations
neutron stars should be systematically brighter discussed earlier, and the UV photometry of
thanblack holes(Narayan, Garcia, & McClintock V404 Cyg reported separately by Hynes et al.
1997; Garcia et al. 2001; Hameury et al. 2003). (2009)thismorethandoublesthesampleavailable
The observational claim has lead to great contro- for comparison.
versy, however, and objections have been raised
to this simple picture. For example, it is neces- 2. Observations
sary to compare objects which are believed to be
accreting at the same rate from their companion Ultraviolet photometry used the Advanced
stars. Menou et al. (1999) argued that this was Camera for Surveys (ACS; Gonzaga et al. 2005)
the case if Eddington-scaledX-ray luminosities of on the Hubble Space Telescope (HST). The obser-
black holes and neutron stars of the same orbital vations are summarized in Table 1. For each tar-
periodwerecompared. Questionswerealsoraised get2or3satelliteorbitswereused,withallfilters
aboutwhethersignificantamountsofaccretionen- observed in each orbit in the sequence F330W–
ergy may be carried by the bulk motion of a jet F250W–F220W–F330W. This provides some av-
(Fender, Gallo, & Jonker 2003). Finally, there is eraging over variability and ensures each short
oneNSXRT,1H1905+000,whichappearstodefy wavelength observation is bracketed by nearby
the trend (Jonker et al. 2007) being fainter than F330W ones.
comparable black hole systems. Observations be- Reduction used standard ACS techniques.
yond just a comparison of X-ray luminosities are Individual images were pre-processed to the
needed. flat-fielded stage using the automatic pipeline,
One promising avenue to improve on this sit- CALACS; we found no need to refine this re-
uation is the ultraviolet. Unlike the optical duction. All of the images, or an orbit-by-orbit
emission which appears dominated by the com- subset, were then combined offline into a geomet-
panion star, UV observations with HST/FOS rically corrected master image using Multidrizzle
of a quiescent black hole revealed a signifi- (Koekemoer et al. 2002) with standard settings.
cant blue excess attributed to accretion light We found excellent registration between individ-
(McClintock et al. 1995). Subsequent more sen- ual images.
sitive spectroscopic observations have been per- PhotometrywasperformedusingtheIDL/AstroLib
formed with STIS of two black holes, A0620–00 aperture photometry routine aper. As our tar-
(McClintock & Remillard2000)andXTEJ1118+480 gets are faint we used a relatively small aperture
(McClintock et al. 2003) and one neutron star, to performphotometry, ofradius0.125”,with sky
2
defined by a 2.5” annulus. Other stars in the 3. Astrometry
field were typically too faint to determine per-
Severalofourtargetsareknowntohavenearby
field aperturecorrections,so we use the tabulated
contaminating stars in the optical or IR which
values from Sirianni et al. (2005). The aperture
have been barely resolved. HST images provide
collects about 75% of the total light, and almost
an ideal opportunity to obtain more precise rel-
all of the sharp core of the point spread function.
ative astrometry of the contaminating stars if
In each case, the position of the target was mea-
they can be identified. We detected the stars
sured in the F330W bandpass and then fixed for
near MM Vel (Filippenko et al. 1999) and Aql X-
the otherfilters,afterverifyingthatbrighterstars
1 (Chevalier et al. 1999). We were unable to de-
were consistently positioned in all the filters.
tect the star north-east of V518 Per reported by
A significant source of systematic uncertainty
Reynolds, Callanan, & Filippenko (2007).
in ACS photometry is charge transfer inefficiency
The F330W image of MM Vel is shown in
(CTI) due to radiation damage to the CCDs
Fig. 1. Star identifications are taken from
(Riess & Mack 2004; Pavlovskyet al. 2005). This
Filippenko et al. (1999). The close blend of
is worstat low light levels, so maximized for faint
stars seen by Filippenko et al. (1999) is well re-
sources in the UV. CTI is also least well cali-
solved and no new stars are detected. The
brated for these cases, and the correction pre-
separation of MM Vel from star A is ∆α =
scriptionsprovidedbyPavlovsky et al.(2005)for-
−0.048s, ∆δ = +1.37”, corresponding to a sep-
mally diverge for negligible sky background or
aration 1.47”, roughly consistent with the esti-
source counts. For the low background case (0.5–
1.5e−/pixel) at source brightnesses up to a few mate ofFilippenko et al.(1999)that MMVelwas
1.6arcsec north-north-west of star A.
hundred electrons, Riess & Mack (2004) found
losses of around 0.035mag at maximum distance WeshowtheF330WimageofAqlX-1inFig.1.
fromthereadoutamplifierfromactualdata. Ifwe StaridentificationsaretakenfromChevalier et al.
rescale this to the times of our observations and (1999). Aql X-1 is star (e), muchbrighterrelative
positionsofthesourceswefindlossesof4–7%. We to star (a) than it is in the I band. Star (b) is
use these latter estimates to correctour measured weakly detected, but stars (c) and (d) are unde-
sourcebrightnesses(orupperlimits)andassignan tected in this bandpass. We find no new stars in
additional error estimate of 2% for most cases to theimmediatevicinityofAqlX-1. Theseparation
account for the uncertainty in applying CTI cor- of Aql X-1 from star (a) is ∆α = −0.031s, ∆δ =
rections. Weincreasethiserrorestimateto3%for +0.07”,corresponding to a separation0.48”, con-
V518 Per as this was the last observation made, sistent with that found by Chevalier et al.(1999).
hence subject to the largestdetector degradation,
and was the faintest source. In F220W observa- 4. AdoptedParametersandArchivalMea-
tions that did not detect a source, the brightness surements
is beyond the regime calibrated by Riess & Mack
4.1. GU Mus = X-ray Nova Muscae 1991
(2004) and we increase the CTI correction error
estimate to 5%. The uncertainties in the CTI For the reddening and distance to GU Mus, we
corrections are always much smaller than the sta- follow Hynes (2005) and do not repeat the argu-
tistical uncertainties in the measured fluxes and ments made there; see that work for the original
so do not significantly affect our conclusions. references. ThesewereareddeningofE(B−V)=
WetabulateourphotometricresultsinTable1. 0.30 ± 0.06 and distance d = 5.89 ± 0.26kpc.
Since we are measuring counts across a broad The spectral type of GU Mus has been estimated
bandpass rather than monochromatic fluxes, we as K3–K5V (Orosz et al. 1996) and K3–K4V
tabulatethedataintwoways: thecorrectednum- (Casares et al.1997). AK4Vclassificationimplies
ber of electrons per second and the average flux aneffectivetemperaturearound4550Kinterpolat-
per unit wavelength. Both have been corrected ing onthe tables inCox (2000). Using the system
for CTI, and to a nominally infinite aperture. parameters from Gelino, Harrison, & McNamara
(2001), including a 10.38hr orbital period, we ex-
pect a surface gravity of logg =4.32.
3
QuiescentphotometrywascompiledbyGelino, Harrisonf,o&r MMMcNaVmelarias that by Filippenko et al. (1999)
(2001) from both their own observations and who find a spectral type of K7–M0V, although
those of Remillard, McClintock, & Bailyn (1992), not rejecting K6V strongly. Other authors pre-
Orosz et al.(1996),King, Harrison, & McNamara ferred earlier spectral types (Shahbaz et al. 1996;
(1996),anddella Valle, Masetti, & Bianchini(1998). della Valle et al. 1997). We will adopt K7V as
Theseresultsarequite varied,withV magnitudes a compromise, implying an effective temperature
ranging from 20.35 to 20.83. This variability is of 4180K. Using the system parameters of Gelino
consistentwiththelargeamplitudequiescentvari- (2004) together with the 6.86hr orbital period of
ations seen by Hynes et al. (2003) and indicates Shahbaz et al. (1996) implies a surface gravity of
that non-simultaneous optical photometry cannot logg =4.55.
reliably be combined with our UV data. Published quiescent photometry of this object
is much sparser than for GU Mus and V518 Per,
4.2. V518 Per = GRO J0422+32
limitedtoRbandmeasurementsofShahbaz et al.
As for GU Mus, we follow Hynes (2005) (1996) and Filippenko et al. (1999), and V band
and adopt E(B − V) = 0.35 ± 0.10 and d = by Hynes et al. (2003).
2.49±0.30kpc. Opinion on the spectral type of
4.4. Aql X-1
V518Perisdivided. Earlyestimatesfavoredearly
M spectral types: M0–M4V (Casares et al. 1995) The interstellar reddening to Aql X-1 has been
or M1–M4V (Harlaftis et al. 1999). Webb et al. estimated at E(B − V) = 0.5 ± 0.1. Although
(2000) favored slightly later types, M4–M5V, Chevalier et al. (1999) estimate a distance of d=
but Gelino & Harrison (2003) argued that the 2.5kpc, Rutledge et al. (2001) argue that this is
spectral energy distribution was best fit by an too low and instead obtain a distance of 4.0–
M1V star. Gelino & Harrison (2003) assumed 6.5kpc by requiring that the companion must fill
negligible disk contamination even in the opti- its Roche lobe, and using the peak luminosity of
cal, however, whereas the other authors cited photosphericradiusexpansionX-raybursts. They
found significant amounts (above 50%), suggest- adopt a preferred distance of 5kpc and we fol-
ing that a later spectral type with blue con- low this. The companion star was estimated by
tamination may indeed be a better description. Chevalier et al. (1999) to be a K7V star. With
In fact, Reynolds, Callanan, & Filippenko (2007) an18.95hrperiod,1.4M⊙ neutronstar,andmass
find substantial flickering even in the near-IR, so ratio q = 0.33 (Welsh, Robinson, & Young 2000)
a spectral type determined from the SED should we expect logg = 3.93. Plausible uncertainties in
be viewed with considerable caution. We there- q have a very small effect on this.
fore consider both M1V and M5V classifications,
Like MM Vel, Aql X-1 is very crowded in
adopting effective temperatures of 3680K and
ground-based observations making optical pho-
3170K respectively. Using the system param-
tometry rare and somewhat uncertain. The pri-
eters of Gelino & Harrison (2003), including a
mary source is Chevalier et al. (1999).
5.09hr orbital period, implies a surface gravity
of logg =4.66.
5. Spectral Energy Distributions
For optical/IR photometry we use data from
Zhao et al.(1994),Casares et al.(1995),Garcia et al. 5.1. Fitting Methodology
(1996),Callanan et al.(1996),Chevalier & Ilovaisky
We begin by fixing the spectral type and red-
(1996),Gelino & Harrison(2003),andReynolds, Callanan, & Filippenko
dening to the values estimated in Section 4. We
(2007). As for GU Mus, substantialvariationsoc-
thenadoptasuitablespectrumfromHauschildt, Allard, & Baron
cur from epoch to epoch.
(1999) for the companion star, redden it us-
ing the extinction law of Fitzpatrick (1999),
4.3. MM Vel = X-ray Nova Velorum 1993
and evaluate synthetic photometry of the red-
ForMMVel,wefollowHynes(2005)andadopt dened companion in the bandpasses of interest
E(B−V) = 0.20±0.05 and d = 3.82±0.27kpc. using the synphotsyntheticphotometry package
The most persuasive spectral type determination (Laidler et al. 2005). One of the primary advan-
4
tages of this approachis to correctly allow for red withonedegreeoffreedom,sothefitisnotasgood
leak from the companion star’s red light into the as might be expected. The inferred blackbody is
UV bandpasses. In a similar way, we calculate small compared to the accretion disk, and a re-
models for the accretion light over the range of gion of radius 6% of the maximal accretion disk
parameters of interest, redden them, and perform radius could readily account for this. It could be
synthetic photometry. This gives us tabulated interpreted as either the accretion stream-impact
fluxes for the two components, allowing us to fit point, or a hot inner annulus of the disk. The
the composite model to the UV photometric data effect of uncertainty in the companion star is in-
in its native form with the normalization of each deed small. Reducing its contribution to zero, or
component as a free parameter. doubling it, changes the derived temperature by
We adopt blackbody models for the disk emis- less than 1000K.
sion. A single-temperature model provides the For comparison to XTE J1118+480, we also
simplestcharacterizationofthedata. Forcompar- performed an analogous fit with a multi-color
ison with the results of McClintock et al. (2003) black body disk. The best fit was obtained with
on XTE J1118+480,we also fit models where the Tin =21500±5700K, with the observed flux cor-
UVemissioncomesfromtheinneredgeofamulti- respodingtoRin ≃3×109cm. Thisishotterthan
colorblackbodydisk. Notethatthismodelexplic- found by McClintock et al. (2003), but the inner
itlyassumesthattheaccretionrateisindependent disk radius is actually inferred to be about the
ofradius(steady-state),so this may notbe anac- same.
curate description of the behavior of a quiescent Since the inferred area is so low, we can also
disk. In the case of GU Mus, we also compare consider the possibility that the light comes from
a model of a slab of recombining hydrogen im- a larger but optically thinner region of the disk.
plemented in synspec, described by an effective We should be cautious as we have introduced an
temperature, a column density, and an area. It additionalfreeparameterandarenowfittingthree
can be seen from Figs. 2–5 that although we have photometricbandswithathree-parametermodel.
consideredthecontributionofthecompanionstar, With this caveat, we find a statistically better
both directly and via its red-leak, in every case fit with the optically thin model than with the
the companion contributions in the UV are much blackbody, χ2 =3.34withnodegreesoffreedom.
smaller than those from the accretion disk and so The best fit is found for a column density of ∼
have very small impact on our fitting results. 1020cm−2 and temperature T =16,100±2000K.
Because the temperature is higher, while the op-
5.2. GU Mus = X-ray Nova Muscae 1991
tical depth remains relatively high, the required
Since the spectral type of the secondary star surface area is actually a little smaller than for
in GU Mus is relatively well constrained we fix the blackbody fit, correspondingtoa regionofra-
its spectral type. Published optical photometry dius about 4% of the maximal disk radius. Thus
spans rather a large range, but this is consistent whether the UV is modeled by an optically thick
with observed short term variability (Hynes et al. or thin component, we come to the same conclu-
2003). Wefixthebrightnessofthecompanionstar sion: the emission originates from a hot region
spectrum to lie at the lower envelope of the opti- with a rather small projected cross-sectional area
cal points. The brightness adopted is not critical, compared to that of the disk.
as this results in only a 5% contribution to the A potential advantage of the optically thin
F330W filter, and negligible impact in other UV modelcanbeseeninFig.2. Theblackbodymodel
passbands. predictsasubstantialdiskcontributioninB,much
TheresidualphotometricSEDvisuallyappears higherthanobservedby severalgroups. Theopti-
well described by a single-temperature black- callythinmodelpredictsamuchlowerBfluxthan
body with temperature T = 13000 ± 1400K theblackbodymodel,soismucheasiertoreconcile
and area about 0.3% of the projected area of with the archival optical photometry. Since the
a tidally truncated accretion disk (∼ 0.9RRoche; optical photometry were not simultaneous with
Whitehurst & King 1991) at 5.89kpc. This fit is the UV data, however,an alternative explanation
shown in Fig. 2. Formally the fit yields χ2 =8.25 isthattheUVobservationsoccurredwhentheac-
5
cretion light was brighter than seen in the optical tain a similar temperature, Tin = 5800±2900K,
photometry. Thisisaquiteplausibleexplanation, with a rather larger inner radius than GU Mus,
since the source is known to exhibit large ampli- Rin ≃13×109cm.
tude optical flaring (Hynes et al. 2003). As dis-
cussedinSection6.1andshowninFig.7,GUMus 5.4. MM Vel = X-ray Nova Velorum 1993
alsohas the highestinferredUV luminosity ofthe
While a UV excess is strongly detected in
sample here, consistent with this being an unusu-
MM Vel, we encountered difficulties in properly
ally high state.
fitting it. Neither the black body nor optically
Anadditionalargumentagainstattributingthe
thin recombination model could reproduce the
UV emission of quiescent BHXRTs to optically
combination of a rising spectrum from F330W
thin emission (in general, rather than in the spe-
to F250W and the non-detection in F220W. We
cificcaseofGUMus)isthatthismodelpredictsa
show in Fig. 4 a representative 12,000K model
large Balmer jump in emission. While our photo-
butdonotfeelitappropriatetoquoteaformalfit
metricobservationscannotdiscriminatethis,spec-
in this case. This model over-predictsthe F220W
troscopic observations of other sources have been
flux,butsincethediscrepancyislessthan3σ this
performed and reveal no such Balmer continuum
may be a statistical fluctuation. We note that
emission. In both A0620–00 (McClintock et al.
the source size implied by a 12000K black body
1995) and XTE J1118+480 (McClintock et al.
model is smaller than in the other systems in this
2003) simultaneous data either through or on ei-
paper, just 0.06% ofthe estimated projectedarea
ther side of the Balmer jump shows no significant
of a tidally truncated disk, which may be another
discontinuity, supporting an optically thick inter-
indication that there are differences in the source
pretation at least in these cases. In view of this
of the UV emission in this system. Using a multi-
evidence,webelievethattheopticallythick(black
color black body disk model we derive rather an
body)modelisthemostcredibleandapplythisto
inner temperature of T = 17800K, with a lower
the remaining sources. It is certainly possible, of limit (1−σ) of 9100K, and the upper limit un-
course, that the optical depth varies from source
constrained. Forthebestfit,theinnerdiskradius
to source, but without strictly simultaneous cov- derived is Rin ≃1.1×109cm.
erageacrosstheBalmerjumpwecannotconstrain
It is possible that the F250W flux is high due
this. to the contribution from the Mgii line which
is known to be strong in quiescent LMXBs
5.3. V518 Per = GRO J0422+32
(McClintock et al. 1995; McClintock & Remillard
WeconsiderbothM1VandM5Vspectraltypes 2000; McClintock et al. 2003). Another explana-
for V518 Per (Fig. 3). Allowing that the strength tion for the discrepancy is intrinsic source vari-
(and possibly temperature) of the hot component ability. To test this, we examined individual
could vary, both possibilities are acceptable, al- F330W observations (which by design bracketed
thoughM5Vappearsto provideabetter fitinthe the F250W and F220W observations). We found
optical region of the SED. Both models suggest the F330W flux showed a standard deviation of
a large or dominant accretion contribution in the only 15% of the mean with no systematic trend,
optical, and a measurable one in the infrared, in small enough to be a constant flux within uncer-
agreement with observations. tainties. This does not rule out variability as an
BothspectraltypesproducenegligibleUVlight explanation,eitherforalowF220Wfluxorahigh
(evenallowingforredleaks)andconsequently,the F250W one, but provides no evidence for such an
hot component derived is largely independent of explanation.
the assumedcompanion spectraltype. For an M1
companion we find T = 5200±2200K, whereas 5.5. Aql X-1
for M5 we find T =5100±2000K. In both cases,
Aql X-1 has the least well-constrained optical
the hot component has an area 0.3% of the pro-
SEDofthe sourcesconsidered,butasinthe other
jected area of a tidally truncated disk at 2.49kpc
cases,thishasrelativelylittleimpactontheinter-
(as was also found in GU Mus). If we instead
pretation of the UV data as all plausible compan-
use a multi-color black body disk model, we ob-
6
ion star spectra make negligible UV contribution. (2001)forA0620–00andV518Per,fromCampana et al.
We can obtain a good fit to the UV data with a (2004)forCenX-4,fromNarayan, Garcia, & McClintock
blackbodyofT =9300±700Kandabout7.5%of (1997) for Aql X-1, from Sutaria et al. (2002) for
the areaofa tidally truncateddisk. The available GUMus,fromHameury et al.(2003)forMMVel,
data and model fit are shown in Fig. 5. With the fromMcClintock et al.(2003)forXTEJ1118+480
alternative multi-color black body fit, we obtain and from Hynes et al. (2009) for V404 Cyg. It
Tin =12400±1500K and Rin ≃7×109cm. should be noted that only XTE J1118+480, and
V404 Cyg have simultaneous UV and X-ray mea-
6. Discussion surements. As anticipated, the neutron star sys-
tems appear systematically above those contain-
6.1. Comparison between UV and X-ray
ing black holes (ranks 1 and 2 out of 8 ranked by
luminosities
LX/LUV). All neutron stars have LX > LUV and
We compile the SEDs in νFν form in Fig. 6. allblackholeshaveLX <LUV. Noteinparticular
thatGUMusissecurelydetectedatamuchhigher
McClintock et al.(2003)suggestedthattheshape
UVluminositythaneitheroftheNSsystems,even
of the UV spectrum could discriminate between
though it is much fainter at X-ray energies.
black hole and neutron star systems. That does
not seem to be borne out by our larger sample This suggests that black hole systems are less
whereitinsteadappearsthattheshapeoftheUV efficient at producing X-rays from accreted mate-
spectrum reflects random variance in the location rialthan neutronstars (assuming that the UV lu-
of the peak of the spectrum from system to sys- minosityisameasureoftheaccretionratefeeding
tem. Note in particular that both GU Mus and the compact objects). This is not a new idea, of
Aql X-1 can be well fitted by black body mod- course, and similar claims have been made based
els, but that it is the black hole system, not the on a comparison of X-ray luminosities alone or in
neutronstar,whichhasthehighertemperatureof combination with estimates of the non-stellar op-
the pair. As can be seen from Fig. 6, however, tical light (Narayan, Garcia, & McClintock 1997;
our sample does seem to bear out the trend that Campana & Stella2000;Garcia et al.2001;Hameury et al.
the X-ray to UV ratio is higher in NSXRTs than 2003; Narayan& McClintock 2008). However,
in their black hole counterparts. Both Cen X-4 comparison of the UV flux with the X-ray flux
(McClintock & Remillard2000)andAqlX-1show gives us more confidence that differences reflect
SEDs rising from UV to X-ray, whereas the black efficiencyofproductionofX-rays,ratherthandif-
hole systems all decline. This result in itself is ferences in mass accretion rate. Furthermore, in
notnew,andbearsoutthe earlierrealizationthat the case of Aql X-1, the UV properties are very
black holes have optical/UV non-stellar luminosi- similar to those ofthe black hole sample, suggest-
tiesexceedingtheirX-rayluminosities,whereasin ing that the difference between the two is indeed
neutron stars this is reversed (Campana & Stella driven by X-ray differences rather than UV ones.
2000; see also Narayan& McClintock 2008). The
advantage we have, however, is that by working 6.2. The UV light source
with the almost pure UV accretion light, we are
Where we are able to obtain good fits to
notatthemercyofuncertaintiesinthestellarcon-
the SEDs, they are characterized by tempera-
tribution; we can directly compare UV accretion
tures higher than expected in quiescent accre-
light with X-ray emission.
tion disks, which should be . 3000K (Menou
We can perform the comparison more rig- 2002), and have much smaller emitting areas
orously by examining the relationship between than the accretion disk. Previously published
X-ray and (dereddened) UV luminosities. In spectroscopic observations show similar trends.
Fig. 7 we compile these data for the four sources A0620–00 was fitted by McClintock et al. (1995)
in our sample, and for archival STIS spectra with a 9,000K black body with area 1/80th of
of A0620–00, Cen X-4, and XTE J1118+480 that of the accretion disk. XTE J1118+480 was
(McClintock & Remillard 2000; McClintock et al. described by McClintock et al. (2003) as being
2003),andACSphotometryofV404Cyg(Hynes et al. fitted with a multicolor black body disk model
2009). X-ray data are taken from Garcia et al. with inner edge temperature 13,000K. Cen X-
7
4 exhibits a UV spectrum that rises monotoni- transfer rates from their donor stars in excess of
cally in νFν, indicating a hot source of emission 10× the plausible average rate to explain the UV
(McClintock & Remillard 2000). The most likely lightsourceasastream-impactpointatoroutside
nature of a localized, hot region of UV emission the circularization radius in the disk, and so this
would seem to be the accretion stream impact explanation seems untenable. An alternative ex-
point, but a hot inner edge to the accretion disk planationisthattheUVoriginatesmuchdeeperin
near a transition to an evaporated flow might be the gravitational potential well, in the inner disk
another possibility. just outside the transition radius, as suggested
If the UV emission originates at the stream- by Campana & Stella (2000) and specifically con-
impact point we would expect the UV luminos- sidered for XTE J1118+480 by McClintock et al.
ity to come fromenergy liberated in falling to the (2003). Fitting our photometry with multi-color
stream-impactpoint. Fortheluminosityofthehot black body disk models yields comparable results
sourcewe integratethe unreddenedblackbody fit to XTE J1118+480, with high temperatures and
tothe UVphotometry. Wecanestimatethe max- radii of order 1–10×109cm, comparable to ex-
imum energy we might expect to release as half pectedtransitionradiiof1000-10,000Rsch. Inthis
the binding energy of material at the circulariza- case a lower mass accretion rate is needed to sup-
tion radius in the accretion disk. If the disk ex- ply the energy than if the UV emission originates
tends beyond the circularization radius, then the at the stream-impact point. If some of the ac-
stream-impact point will be higher in the poten- cretion stream overflows the disk, it will impact
tial well and less energy will be liberated. For closer to the compact object and could provide
GU Mus, we require a very high accretion rate of an alternative mechanism for increasedheating of
about 4×10−9M⊙yr−1 to explain the UV lumi- the inner disk although in this case we might ex-
nosity if this is sustained. This can be compared pect to see evidence for stream-overflowin phase-
to the estimated mass required to power the 1991 resolved spectroscopy of quiescent LMXBs. We
outburst of 10−8M⊙. At this accretion rate, the notethatanotherexplanation,thattheinnerdisk
massrequiredbytheoutburstcouldbesuppliedin is simply heated by the X-ray source,cannot ade-
∼3years,whereaswehaveonlyseenoneoutburst quately explain the UV emission when we see UV
of GU Mus, suggesting a recurrence time at least luminositiescomparabletoorexceedingtheX-ray
ten times larger than this. If the energy source luminosity.
of the UV emission is to be gravitational energy
liberatedatthehot-spot,wethenrequirethatthe 7. Conclusions
accretionrate, andUV luminosity, be highly vari-
We have found UV excesses in four quiescent
able and that the GU Mus observations were well
LMXB systems, three black hole systems and one
above the average level as suggested earlier.
containing a neutron star. In every case, the
Performing similar calculations for the other
UV detection is secure and greatly exceeds that
sources yields mass transfer rates from the donor
expected by the companion star. The spectral
star of 3 × 10−10M⊙yr−1 for V518 Per, 1 × shapes are heterogeneous, and black body mod-
10−10M⊙yr−1 forMMVel,and3×10−9M⊙yr−1 els require a variety of temperatures and emitting
forAqlX-1. ThevaluesforV518PerandMMVel
areas. In some cases a good fit could alterna-
are substantially below that of GU Mus, and do
tively be obtained with an optically thin recom-
notimposestrongconstraints,butthatforAqlX-
bination spectrum, although this would be incon-
1 is comparably high. In this case, we have ob-
sistentwiththelimitedspectroscopicobservations
served multiple outbursts, and so have a much
of other sources across the Balmer jump suggest-
moresecureestimateofthetime-averageaccretion
ing it is not present in emission in those cases. In
rate. Rutledge et al.(2001)haveestimatedthisat
general, the temperatures are higher (sometimes
2.5×10−10M⊙yr−1 (assuminga5kpcdistanceas much higher) than expected for quiescent disks,
we have used). This again is a discrepancy of an
and the UV light sources are also much smaller
order of magnitude with the observed UV bright-
than the whole disk. The mostlikely originof the
ness.
stream-impact point can probably be discounted
BothGUMus andAqlX-1 wouldrequiremass as the requisite accretion rates for both GU Mus
8
and Aql X-1 are at least an order of magnitude Fender R. P., Gallo E., Jonker P. G., 2003, MN-
above estimates, or limits on, the time-averaged RAS, 343, L99
accretion rate. It is therefore more likely that the
Filippenko A. V., Leonard D. C., Matheson T.,
UV emission originates from a hot inner regionof
Li W., Moran E. C., Riess A. G., 1999, PASP,
the disc.
111, 969
This work includes observations with the Fitzpatrick E. L., 1999, PASP, 111, 63
NASA/ESA Hubble Space Telescope, obtained at
Froning, C. S., Cantrell, A. G., Maccarone, T. J.,
STScI, which is operated by AURA Inc. under
et al. 2011, ApJ, 743, 26
NASA contract No. NAS5-26555. Support for
HST proposal GO10253 was provided by NASA
Garcia M. R., Callanan P. J., McClintock J. E.,
through a grant from STScI. R.I.H. also acknowl-
Zhao P., 1996,ApJ, 460, 932
edgessupportbytheNationalScienceFoundation
under Grant No. AST-0908789. This work has Garcia M. R., McClintock J. E., Narayan R.,
made use ofthe NASA AstrophysicsData System Callanan P., Barret D., Murray S. S., 2001,
Abstract Service. ApJ, 553, L47
Facilities: HST (ACS).
Gelino D. M., 2004, Rev. Mexicana Astron. As-
trofis., 20, 214
REFERENCES
Gelino D. M., Harrison T. E., McNamara B. J.,
Beekman G., Shahbaz T., Naylor T., Charles
2001,AJ, 122, 971
P. A., Wagner R. M., Martini P., 1997, MN-
RAS, 290, 303
GelinoD.M.,HarrisonT.E.,2003,ApJ,599,1254
Callanan P. J., Garcia M. R., McClintock J. E.,
Gonzaga, S., et al. 2005, ACS Instrument Hand-
ZhaoP.,RemillardR.A.,HaberlF.,1996,ApJ,
book, Version 6.0, (Baltimore: STScI)
461, 351
Gray R. O., Corbally C. J., 1994, AJ, 107, 742
Campana, S., & Stella, L. 2000, ApJ, 541, 849
Hameury J.-M., Barret D., Lasota J.-P., McClin-
Campana S., Israel G. L., Stella L., Gastaldello
tock J. E., Menou K., Motch C., Olive J.-F.,
F., Mereghetti S., 2004, ApJ, 601, 474
Webb N., 2003, A&A, 399, 631
Casares J., Martin A. C., Charles P. A., Martin
Harlaftis E., Collier S., Horne K., Filippenko
E.L.,ReboloR.,HarlaftisE.T.,Castro-Tirado
A. V., 1999,A&A, 341, 491
A. J., 1995, MNRAS, 276, L35
Hauschildt P. H., Allard F., BaronE., 1999,ApJ,
Casares J., Martin E. L., Charles P. A., Molaro
512, 377
P., Rebolo R., 1997, New A, 1, 299
Hynes R. I., 2005, ApJ, 623, 1026
ChevalierC.,IlovaiskyS.A.,1996,A&A,312,105
Hynes R. I., Charles P. A., Casares J., Haswell
Chevalier C., Ilovaisky S. A., Leisy P., Patat F.,
C. A., Zurita C., Shahbaz T., 2003, MNRAS,
1999,A&A, 347, L51
340, 447
Cox A. N. (Ed.), 2000, Allen’s Astrophysical
Hynes, R. I., Bradley, C. K., Rupen, M., Gallo,
Quantities, 4th Edn., Springer E., Fender, R. P., Casares,J., Zurita, C., 2009,
MNRAS, 399, 2239
dellaValle M.,BenettiS.,CappellaroE.,Wheeler
C., 1997, A&A, 318, 179 Jonker P. G., Steeghs D., Chakrabarty D., Juett
A. M., 2007,ApJ, 665, L147
della Valle M., Masetti N., Bianchini A., 1998,
A&A, 329, 606 KingN.L.,HarrisonT.E.,McNamaraB.J.,1996,
AJ, 111, 1675
9
Koekemoer,A.M.,Fruchter,A.S.,Hook,R.N.,& Shahbaz T., van der Hooft F., Charles P. A.,
Hack,W.2002,inProc.The2002HSTCalibra- CasaresJ.,vanParadijsJ.,1996,MNRAS,282,
tion Workshop: Hubble after the Installation L47
of the ACS and the NICMOS Cooling System,
Sirianni, M., et al. 2005, PASP, 117, 1049
Eds.S.Arribas,A.Koekemoer&B.Whitmore,
337, Baltimore: STScI
Sutaria F. K., et al., 2002, A&A, 391, 993
Laidler, V., et al., 2005, Synphot User’s Guide,
WebbN.A.,NaylorT.,IoannouZ.,CharlesP.A.,
Version 5.0, Baltimore: STScI
Shahbaz T., 2000, MNRAS, 317, 528
McClintock J. E., Horne K., Remillard R. A.,
WelshW.F.,RobinsonE.L.,YoungP.,2000,AJ,
1995,ApJ, 442, 358
120, 943
McClintockJ.E.,RemillardR.A.,2000,ApJ,531,
Whitehurst R., King A., 1991, MNRAS, 249, 25
956
Zhao P., Callanan P., Garcia M., McClintock J.,
McClintock J. E., Narayan R., Garcia M. R.,
1994,IAU Circ., 6072
Orosz J. A., Remillard R. A., Murray S. S.,
2003,ApJ, 593, 435
MenouK., EsinA. A., NarayanR., GarciaM. R.,
LasotaJ.-P.,McClintockJ.E.,1999,ApJ,520,
276
Menou,K.2002,ASPConf.Ser.261: ThePhysics
of Cataclysmic Variables and Related Objects,
387
NarayanR.,GarciaM.R.,McClintockJ.E.,1997,
ApJ, 478, L79
Narayan, R., & McClintock, J. E. 2008, New A
Rev., 51, 733
Orosz J. A., Bailyn C. D., McClintock J. E.,
Remillard R. A., 1996, ApJ, 468, 380
Pavlovsky, C., Koekemoer, A. & Mack, J., Eds.,
HST Data Handbook for ACS v4.0, Baltimore:
STScI
Remillard R. A., McClintock J. E., Bailyn C. D.,
1992,ApJ, 399, L145
ReynoldsM. T.,CallananP.J.,Filippenko A. V.,
2007,MNRAS, 374, 657
Riess, A., Mack, J., Time Dependence of ACS
WFC CTE Corrections for Photometry and
Future Predictions, Instrument Science Report
ACS 2004-006,Baltimore: STScI
Rutledge R. E., Bildsten L., Brown E. F., Pavlov
G. G., Zavlin V. E., 2001, ApJ, 559, 1054
This2-columnpreprintwaspreparedwiththeAASLATEX
macrosv5.2.
10