XMM-Newton EPIC & OM observations of Her X-1 over the 35 day beat period

0.37 MB·English

by  Silvia Zane

Mon.Not.R.Astron.Soc.000,000–000 (0000) Printed2February2008 (MNLATEXstylefilev1.4)
XMM-Newton
EPIC & OM observations of Her X-1 over the
35 day beat period
Silvia Zane1, Gavin Ramsay1, Mario A. Jimenez-Garate2, Jan Willem den Herder3,
Charles J. Hailey4
1Mullard Space Science Laboratory, University College London, Holmbury St. Mary, Dorking, Surrey, RH56NT, UK
2MITCenterfor Space Research, 77 Massachusetts Avenue, Cambridge, MA 02139, USA
4 3SRON, the National Institute for Space Research, Sorbonnelaan 2, 3584 CA Utrecht, The Netherlands
0 4Columbia Astrophysics Laboratory, Columbia University,New York, NY 10027, USA
0
2

n
Accepted 23Jan2004:
a
J

3
ABSTRACT
2

1 We present the results of a series of XMM-Newton EPIC and OM observations
v of Her X-1, spread over a wide range of the 35 day precession period. We confirm
8 that the spin modulation of the neutron star is weak or absent in the low state - in
9 marked contrast to the main or short-on states. During the states of higher intensity,
4 weobserveasubstructureinthebroadsoftX-raymodulationbelow∼1keV,revealing
1 thepresenceofseparatepeakswhichreflectthe structureseenathigherenergies.The
0 strongfluorescenceemissionlineat∼6.4keVisdetectedinallobservations(apartfrom
4
one taken in the middle of eclipse), with higher line energy, width and normalisation
0
during the main-on state. In addition, we report the detection of a second line near
/
h 7keVin10ofthe 15observationstakenduringthe low-intensitystatesofthe system.
p This feature is rather weak and not significantly detected during the main-on state,
-
when the strong continuum emission dominates the X-ray spectrum. Spin resolved
o
r spectroscopyjustafterthe risetothemain-onstateshowsthatthevariationoftheFe
t Kαat6.4keViscorrelatedwiththesoftX-rayemission.Thisconfirmsourpastfinding
s
a based on the XMM-Newton observations made further into the main-on state, and
: indicates the common origin for the thermal component and the Fe Kα line detected
v
atthese phases.We alsofind that the normalisationof the 6.4keVline during the low
i
X state is correlated with the binary orbital phase, having a broad maximum centered
r nearφorbit ∼0.5.Wediscusstheseobservationsinthecontextofpreviousobservations,
a investigatethe originofthe softandhardX-raysandconsiderthe emissionsiteofthe
6.4keV and 7keV emission lines.
Key words: accretion, accretion disks – X-rays: binaries – individual: Her X-1 –
stars: neutron
1 INTRODUCTION asthe“beat”periodthroughoutthispaper,andthephasing
on this period is marked as Φ35.
The beat cycle begins with a sharp rise in X-ray in-
Her X-1 is one of the best studied X-ray binaries in the tensity, which reach the maximum flux, Fmax, in less than
sky; this is borne out by the fact that since 2000 there has ∼0.5dayandthenlasts∼10day(“main-on”state).Thisis
been more than 30 papers published in refereed journals followedbyaperiodoflowX-rayemission(“low-state”)and
with “Her X-1” in the title. We list very briefly the most later again by a “short-on” phase, during which the X-ray
salientobservationalcharacteristicswhicharerelevanttothe flux reaches ∼ 1/20Fmax and ∼ 1/3Fmax, respectively. A
presentstudy.HerX-1isabinarysystem,whichconsistsof secondlow-statefollowsuntiltheonsetofthenextmain-on.
aneutronstarandan A/Fsecondary star;ithasanorbital Thespin modulation of theneutronstar is most prominent
period of 1.7 day and the spin period of the neutron star during the main-on state, becoming much less so (if seen
is ∼1.24 sec. It is one of only a few systems which shows a at all) duringthelow-state and increasing again duringthe
regularvariationinX-rays:the35dayvariationisreferredto short-on (see e.g. Deeter et al. 1998).
(cid:13)c 0000RAS
2
The observed X-ray spectrum changes markedly over al. (2002, hereafter R02) reported the first evidence for a
thebeat cycle.However,it hasbeen recognized that theX- substantial change in the phase difference along the beat
rayspectratakenatdifferentbeatphasescanbefittedusing cycle. We also found that the values of the phase shift ob-
a constant emission model and adding one or more absorp- served by XMM-Newton during the main-on and short-on
tioncomponentswhichvaryincolumndensityoverthebeat differfromthosemeasuredinthepastatsimilarbeatphases.
period,reachingamaximumvalueduringthelow-state(e.g. However,atthetimeofthatpaperonlythreeXMM-Newton
Miharaetal.1991,Leahy2001).Thisstrengthenedtheidea observations, at three different beat phases, had been per-
thatthe35daycycleisduetotheprecessionofanaccretion formed.
disk, that periodically obscures the neutron star beam. To date, Her X-1 has been observed by XMM-Newton
A rather broad Fe Kα emission line at ∼6.5 keV has on15separateoccasionsgivinggoodcoverageoverthebeat
beenobservedbyGinga(seeLeahy,2001andpastreferences period. The first three observations have been discussed in
therein);lineenergyandwidthhavebeenfoundtovaryover detail by R02 and Jimenez-Garate et al. (2002). Here we
both the beat and the spin period (e.g. Choi et al. 1994). report the analysis of the remaining datasets, focusing on
UsingASCAdatatakenduringthemain-onstate,Endo,Na- results obtained using the broad energy band EPIC detec-
gase & Mihara (2000) were able to resolve the feature into tors and puttingour past findingsin abroader contest. We
two discrete emission lines, at ∼6.4 keV and 6.7 keV. How- also present the simultaneous UV observations made using
ever, the second feature has never been detected in other the Optical Monitor (OM). Observations are described in
observations. With the advent of imaging X-ray satellites § 2. UV data and timing analysis are presented in § 3 and
with large effective area over a broad energy band and rel- §4respectively,whilethebehaviouroftheFefeature(s)over
atively good spectral resolution at energies near 6 keV, it the beat, the spin and the ornital period is reported in § 5.
is now possible to resolve the Fe lines with better signal to Summary and discussion follow in § 6.
noise than what was possible using ASCA.Moreover, while
ASCAhadalowerlimitof∼0.5keV,theenergyresponseof
XMM-Newton extendsdown to ∼0.2 keV and this allows a
2 OBSERVATIONS
moreaccuratequantification oftheabsorption columnden-
sity. 2.1 Instrumental setup and data reduction
Anothermanifestation ofthe35 daycycle istheevolu-
XMM-Newtonhasthreebroad-bandimagingdetectorswith
tion oftheHerX-1pulseprofileat energies above∼1keV.
moderate spectral resolution: two EPIC MOS detectors
Afewotherpulsarsshowpulseshapechangeswhicharecor-
(Turner et al. 2001) and one EPIC PN detector (Stru¨der
related withhigh-lowstatesofX-rayintensity,buttheyare
et al. 2001). During the observations of Her X-1, PN and
typicallytransientssuchasGX1+4(Dotanietal.1984)and
MOS1 were configurated in timing mode, while MOS2 was
EXO 2030+375 (Parmar, White & Stella 1989) or flaring
configured in either small window or full frame mode. All
sourcesasLMC X-4(Levineetal. 1991). HerX-1isunique
threeEPICcameraswereusedinconjunctionwithmedium
inhavingarepetitiveevolutionarypatterninpulsechanges
filters. The Optical Monitor (Mason et al. 2001) was con-
tied to a regular high-low intensitycycle, and is therefore a
figured in imaging mode, by using the UVW1 and UVW2
subject of continued interest. Particularly successful in ex-
filters(effectivewavelength 2910 and 2120 ˚A respectively).
plaining the basic features observed by Ginga and RXTE
The details of the observations are summarized in Ta-
is the accretion column model by Scott et al. (2000). This
ble 1; as previously mentioned, the results from the first
modelinvokesthesuccessiveobscuration ofacompact pen-
three observations have been already reported in detail by
cil beam from the neutron star, two fan beams with size
R02 and Jimenez-Garate et al. (2002). The raw event files
∼2Rns and a more extended scattered component.
were first examined and corrected for any time anomalies
At lower energies (below ∼ 0.7 keV), the light curve
(see R02), and then reduced using the XMM-Newton Sci-
does not show evidence for a similar variety of emission
enceAnalysisSoftwarev5.3.3.Theeventsfromtimingmode
components. Past detections of pulsations in the soft X-
rays of Her X-1 obtained using EXOSAT (O¨gelmen & datawereextractedinawaysimilarasinR02.Fordatasets
takeninfull-frameandsmallwindowmode,eventswereex-
Tru¨mper 1988), ROSAT (Mavromatakis 1993), and Bep-
tracted from an aperture of ∼ 40′′ centered on the source
poSax (Oosterbroek et al. 1997, 2000, 2001) only show a
and a background, source-free region; the background sub-
broad and quasi-sinusoidal modulation, which is shifted in
tracted spectra were obtained by scaling appropriately for
phase with respect to the maxima observed at higher ener-
the size of the two regions. Finally, the event arrival times
gies. The phase shift, as measured during the main-on and
were corrected to the solar system barycenter and to the
short-onstate,differsfrom180◦.Undertheassumptionthat
center of the binary system, using the ephemeris of Still et
soft X-rays are due to the reprocessing of the pulsar beam
al. (2001).
by the inner edge of the disk, this is usually interpreted as
evidence for a tilt angle in the disk (see e.g. Oosterbroek
at al. 1997, 2000). As predicted by precessing disk mod-
2.2 Determining the beat phase
els (Gerend & Boynton, 1976), the hard/soft shift in phase
should varyalong thebeat cycle.Theideal way totest this The beat period of Her X-1 is known to vary by up to a
wouldbetotrackthepulsedifferenceovertheentire35day few days (Baykal et al. 1993). In order to determine how
period. However, past measurements were restricted to the the epochs of the XMM-Newton observations relate to the
main-on andshort-on states, duetothedifficultyof detect- precession cycle, we extracted the RXTE ASM (2-10) keV
ing a spin modulation during phases in which the system quick-looklight curvefrom theMITweb site andwe deter-
is highly absorbed. Using XMM-Newton data, Ramsay et mined the 35 day phase by computing the time difference
(cid:13)c 0000RAS,MNRAS000,000–000
XMM-Newton observations of Her X-1 over the 35 day beat period 3
666
Observation XMM- Exp MOS Orbital 35d 555
Date Newton (ksec) Mean Phase Phase
Orbit Ct/s 444
2001Jan26 207 10 97.5 0.20–0.26 0.17
2001Mar04 226 11 3.6 0.47–0.56 0.26 M Ct/s 333
2001Mar17 232 11 11.3 0.52–0.60 0.60 AS 222
2002Feb25 405 10.3 2.5 0.55-0.62 0.46
2002Feb27 406 11.7 3.2 0.71-0.79 0.51 111
2002Mar07 410 8.8 5.2 0.40-0.46 0.74
2002Mar09 411 7.3 3.4 0.57-0.62 0.79 000
2002Mar15 414 7.3 2.2 0.09-0.14 0.96
---111
2002Mar17 415 7.3 130 0.26-0.31 0.02
2002Mar27 420 4.4 3.4 0.17-0.20 0.31 110000..00
2002Sep20 509 5.9 2.9 0.56-0.60 0.35
2002Sep22 510 7.3 1.8 0.72-0.77 0.41
2002Oct04 516 10.3 2.1 0.78-0.85 0.77
2003Mar26 603 15.6 2.0 0.81-0.88 0.72 1100..00
2003Mar28 604 6.1 0.5 0.98-0.03 0.75 S Ct/s
O
M
Table1.SummaryoftheXMM-NewtonobservationsofHerX-1.
11..00
FortheorbitalphasingweusetheephemerisofStilletal.(2001);
for the 35 d phasing we define phase 0.0 at the main high state
turn-on (see text for a discussion). The count rate refers to the
rate in either the MOS1 or MOS2 detector depending on the
00..11
sourceintensity.Analysisofthefirstthreedatasetshasbeenpre- 00..00 00..55 11..00 11..55 22..00
35 day phase
sentedindetailsinR02andJimenez-Garateetal.(2002).
Figure 1. Top Panel: a typical RXTE/ASM (2-10) keV light
curve of Her X-1 over the 35 day beat period. Bottom Panel:
between each observation and the start of the previous rise the mean count rate of Her X-1 registered by the EPIC MOS
tomaximum.Wethentooktheperiodtobethetimetillthe detectors. Thelowestcountratecorrespondstoaneclipseofthe
next rise to maximum. We estimate the uncertainty on the neutronstarbythesecondary.
phaseofthebeatperiodtobe∼0.01-0.02cycles.Theresult-
ingbeatphaseisreportedinTable1,alongwiththedatesof
ison averagesmall (less than10% asestimated byBoroson
the observations and the mean X-ray brightness. Note that
etal.2001).However,itismodulatedoverthebinaryorbital
thelast observation hasbeen takenduringaneclipse of the
period.TheUVcountratedetectedusingthetwoOMfilters
primarybythesecondarystar:thiscorrespondstothepoint
isshowninFigure2,asafunctionoftheorbitalphase.Note
with thelowest X-ray count rate in Figure 1.
thatobservationswerenotalwaysperformedinbothfilters,
The phase coverage of the XMM-Newton observations
and in the last two exposures no UV measurements were
overthe35dayperiodisshowninFigure1.Forcomparison,
made. The illuminated secondary star is expected to domi-
wealsoshowtheASMlightcurverecordedduringatypical
nate around φ ∼0.5. As we can see, near this orbital
cycle. We note that, since the actual period of the X-ray binary
phase the UV flux measured by OM reaches a broad max-
modulation variesaround its nominal valueof 35 d,thede-
imum, similar to that found using IUE in the 1750-1850˚A
termination of the state of the system (either low, main on
band (e.g. Vrtilek & Cheng 1996). However, there is a flux
or short on) based on a direct comparison with the ASM
decrease at φ ∼0.5, possibly due to the fact that part
lightcurveinthetoppanelmustbetakenonlyasindicative. binary
of the secondary star is shadowed by the accretion disk or
Aswecansee,thebeatcycleisreasonablywellsampled,
to a warp in thedisk itself.
with most of the exposures taken outside the main-on and
Therisefromtheeclipse(φ ∼0.1-0.3)issteeperin
short-on phases. Three observations have been performed binary
UVW1 (longer wavelength) than in UVW2. Since, at these
during states of high intensity: in addition to the main-on
orbital phases, the accretion disk mainly contributes to the
and short-on datasets discussed by R02 (Φ35 = 0.17 and
UV flux, this is consistent with a scenario in which, after
0.60) we have a further exposure at Φ35 = 0.02. We note
theeclipse,theregionsofthediskfarthestfromtheneutron
that the ASMlightcurve
star come into view first.
3 THE UV DATA
4 TIMING ANALYSIS
The UV emission probably originates from a blend of at
4.1 Search for modulation close to the spin period
least two different components: the illuminated face of the
secondarystarandtheaccretion disk.Theirrelativecontri- The spin period of Her X-1 has been found to vary from
bution varies during the orbital motion, and in general is ∼1.23772 s to 1.23782 s, with alternating phases of spin-up
difficult todisentangle. Unlikethe X-rays,theUV emission andspin-down(Parmaretal.1999,Oosterbroekatal.2001).
does not show an obvious correlation with the beat period. Furthermore, it can vary significantly on relatively short
Thisisinagreementwiththefactthatthediskcontribution time scales. The values of the spin period measured using
(cid:13)c 0000RAS,MNRAS000,000–000
4
330000 666
555
225500 UVW1
444
s
220000 M Ct/ 333
W1 ct/s 115500 AS 121212
UV 000
110000 ---111
11000000
0.3-0.7keV
5500
110000
660000 hi^2
c
1100
5500 UVW2
11
4400
W2 ct/s 3300 11000000 2-10keV
V
U 110000
2200 hi^2
c
1100
1100
11
00
00..00 00..55 11..00 11..55 22..00 00..00 00..55 11..00 11..55 22..00
Orbital Phase
35 d phase
Figure 2.TopPanel:The UVW1(top panel)and UVW2(bot- Figure 3.Toppanel:the meanASMX-raylightcurveover the
tom panel) data folded on the orbital period. data have been 35daycycle. Middlepanel:theχ2obtained byfitting aconstant
phasedbyusingtheephemerisofStilletal.(2001). value to the 0.3-0.7 keV folded light curve. Bottom panel: same
butfor2-10keVfoldedlightcurve.Infoldingthedatawefolded
the data on the spinperioddetermined usingaDiscrete Fourier
Transform (main-on or short-on observations) or, when unavail-
thefirstthreeXMM-Newton observationswerereportedby able,themorerecent measureof1.237777 s.The χ2iscomputed
R02:usingthePNdetector,wefound1.237774s,1.237753s
byusing25phasebins.
and 1.237751 s, respectively. The first measure confirmed
the slow-down trend monitored by BeppoSax (Oosterbroek
etal.2000)andChandra(Burwitz,privatecommunication); Usingthismethodwefindthat,outsidethemain-onand
although we observed a slight decrease in thetwo following short-on states there is evidence for significant modulation
epochs. above2keVinthetwodatasetstakenatΦ35=0.26and0.31.
We performed a search for pulsations using a Discrete In the first case (which is one of those already reported by
FourierTransform forallthenewdatasets, findingastrong R02)theχ2 increasesabovethe90percentconfidencelevel
evidence of pulsations only in one case (revolution 415). if we reduce thenumberof phase binsfrom 25 to 20.
This is the only new exposure taken outside the low state, Defining the amplitude as (max-min/mean) of a fitted
precisely during the main-on at Φ35 = 0.02. The corre- sinusoid, we find an amplitude of 11 and 13 percent for
sponding spin period, as obtained from theEPIC PN data, Φ35=0.26 and 0.31, respectively. Previous observations of
is 1.237777(1) s, where the number in parenthesis repre- Her X-1 during the low state include Coburn et al. (2000),
sents the error on the last digit as determined by using whofound aspinmodulation of∼13percent(assuming the
the Cash (1979) statistic. The spin period is just slightly samedefinitionasours)inthe3–18keVrangeusingRXTE
increased with respect to the value measured by XMM- data in an anomalous low state, while Mihara et al. (1991)
Newton in 2001. determinedanupperlimitforthepulsedfractionof2.4per-
All the remaining datasets showed amplitude spectra cent in the1.2-37 keV energy range usingGinga data.
withonlyweakornosignificantpeaksclosetotheexpected
spinperiod.Inordertoinvestigatefurtherthepossiblepres-
4.2 Spin-resolved light curves
enceofamodulation,wefoldedallthedatasetsonthemost
recent measured period, i.e. 1.237777 s, by using 25 phase InFigure4,weshowthespinprofilesof HerX-1asderived
bins. We then fitted a constant value to the resulting light for those observations in which we detected a significant
curveanddeterminedtheχ2.Therefore,weexpectapoorer modulation in the 0.3-0.7 keV or in the 2-10 keV energy
fit in datasets with a stronger spin modulation. Since the band. Due to the uncertainty in the spin period at each
profile of the folded light curve varies considerably in the epoch, the spin phases are not on the same absolute scale.
soft and hard energy band, we split the eventsinto two en- Therefore,onlytherelativephasingbetweenthelightcurves
ergy ranges: 0.3-0.7 keV and 2-10 keV. The resulting χ2 is indifferentenergybandstaken during the same observation
shown in Figure 3, as a function of the35 day phase. For 1 is physically meaningful (see § 4.3).
degree of freedom (the value of the period), the difference The change of the spin profile of Her X-1 in different
to thefit is ∆χ2=2.71 at the90 percent confidencelevel. energybandshasbeenwelldocumentedintheliterature.In
(cid:13)c 0000RAS,MNRAS000,000–000
XMM-Newton observations of Her X-1 over the 35 day beat period 5
φ =0.02 φ =0.17 φ =0.26 φ =0.31 φ35=0.60
35 35 35 35
330000 225500 77..00 2222 3300
2288
2211
225500 220000
66..55 2266
220000 2200
s 115500 2244
Ct/
115500 66..00 1199 2222
110000
110000 2200
1188
5500
5500 55..55 1188
1177
00..00 00..55 11..00 11..55 22..00 00..00 00..55 11..00 11..55 22..00 00..00 00..55 11..00 11..55 22..00 00..00 00..55 11..00 11..55 22..00 00..00 00..55 11..00 11..55 22..00
Spin Phase Spin Phase Spin Phase Spin Phase Spin Phase
φ =0.02 φ =0.17 φ =0.26 φ =0.31 φ =0.60
35 35 35 35 35
220000 11..2200 1144
225500 118800 33..5500
116600 11..1100 1133
33..4400
220000 114400
s
Ct/ 112200 33..3300 11..0000 1122
115500 110000 33..2200
00..9900 1111
8800
33..1100
110000
6600
33..0000 00..8800 1100
00..00 00..55 11..00 11..55 22..00 00..00 00..55 11..00 11..55 22..00 00..00 00..55 11..00 11..55 22..00 00..00 00..55 11..00 11..55 22..00 00..00 00..55 11..00 11..55 22..00
Spin Phase Spin Phase Spin Phase Spin Phase Spin Phase
Figure4.Upperpanel:theEPICPNspinprofileinthe2-10keVbandasafunctionofthe35daycycle.Lowerpanel:asabovebutfor
the0.3-0.7keVband. Inallpanels,because oftheuncertaintyinthespinperiodineachepoch, thespinphases arenotonanabsolute
scale.ThetwoverticallinesatΦ35=0.17markthesecondatypicalpeakintheinterpulse(hardband)andthesmall“notch”likefeature
(softband)discussedinthetext.
ordertocomparethespinprofilesobtainedbyusingXMM- sinusoidal. There is evidence of a double peaked structure
Newton with thepreviously published data, we take as ref- atΦ35=0.26:R02showthehardX-raymodulationresolved
erence, in the range above ∼ 2 keV, the work of Deeter et intofinerenergybandsandfindthatatthisbeatphasemod-
al.(1998)andScottetal.(2000).Theseauthorsreportedthe ulation at the hardest energies is broad and single-peaked.
variation of the hard X-rays light curve obtained by Ginga Note that the profile obtained at Φ35=0.31 is less certain,
and RXTE overa wide range of beat phases. sinceit hasbeenobtainedbyfolding dataoverthemostre-
centmeasuredperiodandnotusingtheperiodmeasuredat
Aswecansee,theXMM-Newtonlight curvesobtained
thetime of theobservation. However, we haveexplored the
shortlyafterthemain-onturnon(Φ35=0.02)inthe2-10keV
effect of folding the data on slightly different spin periods,
energybandcloselyresemblemostofthemain-onstatespre-
centeredaround1.237777s:theresultingfoldedlightcurves
viously reported. Theonly exception is that thesmall peak
are very similar. The profile detected during the short-on
observed during the inter-pulse (at φspin=0.9 in Figure 4) (Φ35=0.60) is similar to previous observations at a similar
appears slightly more sharp than previously reported. At
beat phase, although the high signal to noise ratio of the
thenextmain-onbeatperiod,Φ35=0.17,wedetectanatyp-
XMM-Newtondatarevealedamorecomplexstructuredur-
ical profile, with a second large peak in the interpulse on
ing therise of themain peak.
the leading side of the main pulse (the feature occurs at
φspin=0.05inFigure4).Thepeakbecomesmoreprominent
asthehighstateprogresses,andhasbeenobservedonlyoc-
casionally in Her X-1: it is present in the Ginga data taken
ThehighthroughputofXMM-Newtonallows ustoob-
during the main-on state “D” by Deeter et al. (1998), at
tainalightcurvewithveryhighsignaltonoiseratioalsoin
the two beat phases Φ35=0.073 and 0.162. However, it was
thesoft X-rays(0.3−1 keV).Consistent with pastmeasure-
notpresentintheotheroftheirmain-onobservationsnorin
ments, the spin profiles of the soft X-ray band in Figure 4
those discussed by Scott et al. (2000). Deeter et al. (1998)
showabasicallysinusoidalshape.However,inbothmain-on
noted that such atypical profile is similar to a previous one
observationsthebroadpeakiscutbyadipwhichhasnotbe
recorded by using Kvant (Sunyaev et al. 1988) and suggest
seen before. Further,data taken at Φ35=0.17. show a small
it may be dueto a reduction in theaccretion rate.
“notch” like feature preceding the maximum (at φspin=0.4
Moving later in beat phase, the detected amplitude of inFigure4).Duringtheshort-on(Φ35=0.60)themodulation
the hard X-ray modulation decreases sharply at Φ35=0.26 is broader, with evidence for a fairly complex substructure
and Φ35=0.31: here the modulation is broad and quasi- duringtheleading edge of the peak.
(cid:13)c 0000RAS,MNRAS000,000–000
6
666
555 1)
11..00 φ35=0.26 φ35=0.60 M Ct/s 234234234
AS 111
000
00..55 ---111
66..6655
66..6600 2)
CorrelationCorrelation 00..00 KeV 66666666........4455445505050505
Cross Cross 6688..33005500 3)
660000
WV
--00..55 φ =0.17 E e 440000
35 220000
φ =0.02 00
35 440000

F--11i..0011g 88 u00re 5. The c22r77 o00ss correPPlahhaatssieeo SSnhhiisff00 tt ((fddoeeggrrreeeetssh))e soft (990 00.3-0.7 keV) a11n88 d00 Sigma eV 123123000000000000 4)
hard(2-10keV)lightcurvesatfourdifferent35dayphases. 00
00..001100
00..000088 5)
m 00..000066
4.3 The phase shift between the soft and hard Nor 00..000044
00..000022
light curves
00..000000
77..3300
R02 found that the phase shift between the soft and hard 77..2200 6)
X-ray light curves, folded on the spin period, varied over V 77..1100
e
the beat period. To determine how the dataset taken at K 77..0000
Φ35=0.02 compared with these previous observations we 66..9900
cross-correlated the hard and soft light curves (see R02 for 550000
440000 7)
details). This is the only new observation that can be used
W V 330000
for such purpose, since the others do not show a significant Ee 220000
modulation in both soft and hard X-rays. As we can see, 110000
the two observations taken during the main-on have a very --4400 vv v v v vv v v v
similar correlation, with data taken at Φ35=0.02 showing a 44..00••1100--44 8)
m 33..00••1100
strong anti-correlation and a phase shift between the max- or 22..00••1100--44 v v v v
ima of the soft and hard light curves of ∼ 130◦. This dif- N 11..00••1100--44 v v
ference in phase is slightly smaller than what was observed v v
atΦ35=0.17,andconsistentwiththegeneraltrendoverthe 00..00 00..55 11..00 11..55 22..00
35day phase
beat period.
Figure 6. Variation of the Kα and Fe XXVI line parameters
alongthebeatcycle,asmeasuredwithEPICPN.Fromthetop:
1)mean ASMlightcurve; 2)-5) central energy, equivalent width
5 SPECTRAL ANALYSIS: THE VARIATION (EW),widthandnormalizationoftheprominentKαFeemission
OF FE COMPLEX line; 6)-8) central energy, EW and normalization of the Fe line
at∼7keV.Forthosedatasets wherethesecondfeatureisunde-
5.1 Line variation over the beat cycle tected, we report upper limits to the EW and normalization, at
the90percentconfidence level(“v”symbols).
The integrated X-ray spectra of Her X-1 over the 35 day
cycle have been modeled by many authors (e.g. Mihara et
al.1991,Oosterbroekatal. 1997,Coburnetal.2000,Oost- orescent line varies over the beat period, increasing during
erbroek at al. 2000, Choi et al. 1994, Leahy 2001). In ac- the main on. However, only three epochs were reported in
cordance with these studies, themain on, short on and low that paper.
states all require a two component continuum with a ba- Inordertofitthespectra,werestrictedtheenergyrange
sic emission spectrum and an absorbed component whose to 4-10 keV and fitted an absorbed power law with one or
strengthismaximumduringthelowstates.Thenaturalin- more Gaussians. The power law and Gaussian parameters
terpretationisthatthe35dayvariationisduetotheoccul- were then fixed and the energy range 0.3-4 keV included.
tationoftheneutronstarbeambytheinterveningaccretion The absorption component was allowed to vary and we in-
disk.ModelfittingoftheXMM-Newtonspectraobtainedus- cluded a blackbody component and aneutral partial cover-
ing the first three datasets has been reported by R02 and ing model if required. This allows us to fix the absorption
confirm this picture. Here, we used the same model for fit- components at an appropriate level, which is crucial since,
ting the newest datasets: since the overall fits do not show for large values of the column density, a characteristic ab-
new characteristics we do not discuss further the details of sorption feature appears at ∼ 7.1 keV. This feature may
theoverall spectral shape. affecttheidentification oftheemission Felines, therefore it
Instead,wefocusonthebehaviouroftheX-rayspectra has to bemodeled carefully.
near the Fe line component(s), detected in the past around Consistent with R02, we found that a prominent fluo-
∼6.6 keV. As shown by R02, the central energy of the flu- rescentlineisdetectedatallbeatphases(apartfromobser-
(cid:13)c 0000RAS,MNRAS000,000–000
XMM-Newton observations of Her X-1 over the 35 day beat period 7
vation taken during at mid-eclipse), with a central energy
that varies between ∼6.4-6.6 keV and is highest during the
main-on. However, the most striking result is the detection
ofasecond component,anFeXXVIlineat∼7keV,which
is only present outside the main-on. This line was not re-
portedbyR02,butacloseranalysisofthetwoobservations
taken at Φ35 =0.26 and 0.60 revealed its detection in both
datasets. Thelineparametersare reported,asafunction of
thebeat phase, in Figure 6.
We find that the equivalent width (EW) of the Kα Fe
lineshowsavariationthroughthebeatcycle,withnotrend
being visible. This is likely to be due to a varying contin-
uum, since the Kα Fe line normalisation is highest during
themain-on statewhileit remainslowand almost constant
6 7 8 9
at other 35 day phases. Similarly, the line width is greatest
during the main-on state. The energy of the second com-
ponent remains approximately constant at ∼7.0 keV, ex-
ceptduringtheshort-onstatewhereitissignificantlylower
(∼6.85 keV). The feature is strongest (larger EW) during
thestatesoflowX-rayintensity,althoughitisundetectable
in some of the exposures in this phase range. On the other
hand, its normalisation is strongest during the short-on. In
some of theobservations, we do not find evidence for a sig-
nificant Fe XXVI line. In order to determine upper limits
to EW and line normalisation, we added a Gaussian com-
ponent to the best fit spectral model, by fixing the central
energyat7keVandthewidthat0.1keV.Theupperlimits
obtainedforthedatasetstakenduringthemain-onarecon-
sistentwiththelinebeingundetectedbecauseofthestrong
continuum emission.
InordertoshowthesignificanceofboththeFefeatures,
5 6 7 8 9 10
we show in Figure 7 the X-ray spectrum between 5-9 keV
for Φ35=0.02 and Φ35=0.79. The solid line is the best fit
spectralmodelwith thenormalisations ofthetwoGaussian Figure 7. The spectral region around 6.6keV from Φ35=0.02
componentssettozero.Wefindnoevidenceforthepresence (top) and Φ35=0.79 (bottom). The y-axis is the normalised
counts, i.e. the number of counts per energy bin per second, as
of a Compton downscattered shoulder (expected near E0−
E0/(mec2+E0) whereE0 is thecentroid oftheline) of the mfiteaafstuerretdhbeynoErPmIaClisPaNti.onInobfoththeopnaeneolratwsoolGidaluinsseiasnhocwomstphoenbenetsst
kind recently observed in HETGS Chandra spectra of the
havebeensettozero.
high mass X-ray binary GX 301-2 (Watanabe et al. 2003).
followthesoftX-raylightcurveandareanti-correlatedwith
thehard, non-thermalflux emitted above 2 keV.
5.2 Pulse-resolved spectroscopy
The variation of the Fe Kα line over the neutron star spin
period and its correlation with the thermal or non-thermal
5.3 Variation of Fe complex over the orbital
spectral components was studied by R02. We found no ev-
period
idence for a significant variation in the line parameters at
Φ35=0.26 or 0.60. In contrast, the observation made dur- Further information about the origin of the Fe Kα 6.4keV
ing the main-on (Φ35=0.17) showed a clear correlation be- line can be obtained by observing its variation over the or-
tween the EW of the 6.4keV line and the soft X-ray light bital period. This is shown in Figure 9, for all observations
curve. This supported the idea of a common origin for the taken outside the main-on. The average line normalization
∼6.4 keVfluorescent line and the thermal flux. is 2−4×10−4 counts/cm2/s, with larger peaks at ∼10−3
Thenewdatatakenduringthemain-on(Φ35=0.02)can counts/cm2/s and a dip at φbinary =0.43−0.56. (The line
beusedtotestthisscenariofurther.Wephasedalltheevents normalisationduringthemain-onstateisgreaterbyafactor
onthespinperioddeterminedin§4andextractedspin-phase if>5-10comparedtothelowstate).Thelineisnotdetected
resolved spectra. Details of the model fitting are as in §5.1. during the middle of the eclipse, at which point the upper
ResultsarereportedinFigure8,togetherwith thesoft and limit on theline normalisation is 3.7×10−5 counts/cm2/s.
hard X-ray light curves. Intheprevioussectionweshowedthatduringthemain-
TheshapeoftheEWvariationisbroadlysimilartothe on state the line normalisation phased on the spin period
profileatΦ35=0.17 (R02),although that observationshows is correlated with the soft X-ray light curve suggesting a
amore‘top-hat’profile.ConsistentwithR02,wefindinthe common origin for thethermal soft X-ray emission and the
main-onstatethatboththelinenormalization andtheEW 6.4keV line. Wenowfind that in thelow state theline nor-
(cid:13)c 0000RAS,MNRAS000,000–000
8
0000....0000000011114444
225500 0.3-0.7keV
Ct/s 220000 0000....0000000011112222
n
p
C 0000....0000000011110000
EPI 111105050000 K alpha NormK alpha Norm0000....0000000000008888
Fe Fe 0000....0000000000006666
220000 2-6keV
0000....0000000000004444
pn Ct/s 115500 0000....0000000000002222
C
EPI 110000 v
0000....0000 0000....5555 1111....0000 1111....5555 2222....0000
OOrrbbiittaall PPhhaassee
5500 Figure 9. The Fe 6.4keV Kα line normalisation as a function
00..001100

oftheorbitalphase,forallobservations taken outsidethe main-
00..000088 on. Data have been taken with EPIC PN. Only an upper limit
on has beenobtained fromthe observation taken duringthe eclipse
ati 00..000066
Linemalis (arrow).
or 00..000044
N

00..000022 sidethemain-onstate.Wenowproceed bydiscussingsome
00..000000 physical interpretations for our results.

440000
330000 6.1 The spin-resolved light curves: evidence for
W structure in soft X-rays (0.3−1 keV)
E 220000
The complex, multipeaked structure of the pulse profile of
110000 Her X-1 above ∼ 1 keV and its evolution pattern over the
beat cycle is the subject of several studies. The basic fea-
00
00..00 00..55 11..00 11..55 22..00 tures of the pulse profile are nominally explained by the
Spin Phase
oblique rotator model (Lamb, Pethick & Pines 1973), but
Figure8.TheobservationmadeusingtheEPICPNdetectorat
there are still many unresolved questions regarding the de-
Φ35=0.02.Fromthetoppanel:thelightcurveinthe0.3-0.7keV
band; the 2-6 keV band; the Fe Kα line normalisation and EW tails of the mechanism, such as whether the radiation pat-
asafunctionofthespinphase. tern is best described by a superposition of ‘fan’ and ‘pen-
cil’ beams and whether the obscuring material that modi-
fies the visibility of the direct beams from the neutron star
malisation iscorrelated with theorbitalperiod.Wegoonto contributes to the pulse formation. Particularly successful
explore this furtherin §6.3. in explaining the pattern observed by RXTE and Ginga
is the accretion column model by Scott et al. (2000). This
model invokes the successive obscuration of a direct beam
(that originates from the polar caps of the neutron star),
twofan beamsfocussed intheantipodal direction (possibly
6 DISCUSSION
duetocyclotron backscattering) andamoreextendedscat-
We have reported the results based on EPIC and OM data teredhalo.Asthemain-onprogresses,theinneredgeofthe
for a set of XMM-Newton observations of Her X-1. Con- accretion disk cover first one of the fan beam components,
sistent with past reports, we find that the spin modulation then the direct pencil beam and then the second fan beam
of the neutron star is most prominent during the main-on component.
and short-on state, becoming not significant duringthelow In contrast, the pulsation detected below ∼ 1 keV us-
state. These data sets allow us to follow the change in the ingEXOSAT(O¨gelmen&Tru¨mper1988),ROSAT(Mavro-
relative phasing of thermal and non-thermal emission and matakis 1993), and BeppoSax (Oosterbroek et al. 1997,
the change in the pulse profile over the beat period. The 2000, 2001) is broad and quasi-sinusoidal. This emission is
largecollectiveareaofXMM-Newtonanditsgood response thermal and probably originates in the inner edge of the
downtosoftX-raysalsorevealedacomplexsubstructurein disk, that partially intercepts and reprocesses the hard X-
the light curve in the softest and less well studied energy raysfromtheneutronstar.XMM-Newtondatatakenduring
band. themain-onhaveanexcellentsignaltonoiseratioandallow
Data taken during the main on, at Φ35 =0.02, show a ustoresolveasubstructureinthesoftX-raylightcurve.The
common phase modulation between the fluorescent Fe Kα 0.3-0.7keVpulseprofileobservedatΦ35=0.02clearlyshows
line and the soft emission, confirming the results found by thepresenceoftwoseparatepulses,atφspin ∼0.0−0.1and
R02atΦ35 =0.17.Moreover,fortheveryfirsttimeinCCD φspin ∼ 0.3 −0.4 in Figure 4 (the value of the absolute
nongratingspectra,wereportevidenceforasecondFeline, spinphaseis arbitrary).Similar features arealso evidentin
withcentralenergynear ∼7keV,whichisonlyvisibleout- data taken later in the beat cycle: at Φ35=0.17 the pulse
(cid:13)c 0000RAS,MNRAS000,000–000
XMM-Newton observations of Her X-1 over the 35 day beat period 9
profile shows two maxima (at φspin ∼ 0.7 and φspin ∼ 0.9 ing substantially (R02). In contrast with past observations
in Figure 4) and a small “notch” like feature preceding the (Oosterbroek et al. 2000), R02 did not detect a symmetric
main pulse (at φspin=0.4 in Figure 4). During the short on stateatΦ35 ∼0,0.5.Ifthereprocessing region islocated at
(Φ35=0.60), the soft X-ray modulation is broader, more si- or near the inner region of the disk, such lack of symmetry
nusoidal,andexhibitsacomplexsubstructureintheleading may indicate that the disk is strongly warped (Heemskerk
edge of the pulsefollowed by a faster decay. &vanParadijs1989).Inordertofurthermonitorthephase
Whetherthesefeaturesreflectanintrinsiccomplexityin shiftoverthebeatcycle,wepresentedanewmeasureofthe
the thermal and geometrical properties of the reprocessing pulsedifference,takenduringthemain-on(Φ35 =0.02).We
region or keep memory of the structure of the illuminating find that the trend detected by R02 is confirmed, and the
beam(s)isunclear.However,itisworthnoticingthatduring changeinthetiltangleappearstobesmoothandcontinuous
all three beat phases with bright emission, the features de- overthebeat cycle.
tectedinthesoftlightcurvesqualitativelyresemblethoseof Possible recent changes in the structure of the inner
thecorresponding hard band.Thetwo maxima observed in layers of the disk have been also discussed by Oosterbroek
the softer band during the main-on (Φ35=0.02, 0.17) have et al. (2001), who detected a rapid spin down in Her X-1
a phase separation similar to that of the two main peaks following an anomalous low state. This spin down was not
observed in the 2-10 keV range, i.e. ∆φspin ∼ 0.2 −0.3. accompanied by a significant simultaneous change in the
Similarly, the small “notch” preceeding the main pulse at mass accretion rate (Oosterbroek et al. 2001). Instead, it
Φ35=0.17 may be associated with the atypical pulse ob- wasmoreplausiblyassociated withchangesinthestructure
served at higher energies (both precede the main peak by oftheoutermagnetosphereoftheneutronstarand/orinthe
∆φspin ∼0.25−0.3).Iftheseassociations arecorrect,i.e.if innerdisklayers.Theseobservationsimposeimportantcon-
atleastsomeofthemaincharacteristics oftheneutronstar straints on any theory modeling the accretion torques and
beam are not “washed out’ during the reprocessing of hard phenomenarelatedtotheinneredgeofthediskinHerX-1.
X-rays,thismeansthattheradiativetimescaleintheinner
edgeofthediskismuchsmallerthanthephotontraveltime.
Moreover,thereprocessingregioncannotbetooextendedin 6.3 The Fe Kα line
size. This may be explained by a high inclination angle of
The strong emission line at ∼ 6.4 keV is detected in all
the disk and/or a highly warped inner edge, in which case
XMM-Newtonobservations,withlargerbroadeningandnor-
only afraction oftheinnerregion ofthediskinterceptsthe
malization during the main-on. Past reports of Fe features
neutron star beam. Alternatively, the reprocessing of the
inthespectraofHerX-1includeanASCAexposure(Endo
neutron star beam may be dominated by a shocked, opti-
et al. 2000) and a set of 63 Ginga observations analyzed
cally thick hot spot instead of being distributed over the
by Leahy et al. (2001). The good coverage of the beat pe-
entireinneredge.Asimilarsituation isobservedinSWSex
riod allowed Leahy et al. (2001) to undertake a systematic
stars (see e.g. Dhillon et al. 1997, Groot et al. 2000 and
study of the line variation along the precession cycle. They
theirFigure6),wherethehotspotregionisformedthrough
found a broad feature near ∼ 6.5 keV, with the EW vary-
a shock that occurs along the stream trajectory, but well
ing considerably between main on, short on and low state.
inside theaccretion disk.
Themean valuesfor each stateare 0.48 keV,0.66 keV,and
∼1.37keV,respectively.Similarcorrelationsarenotevident
6.2 The pulse period and the change in the phase in the XMM-Newton data, where instead an additional Fe
shift between soft and hard emission XXVI line, possibly a coronal signature, is detected during
thelow states (see § 6.4).
Given the complexity of the source, pulse-phase spec-
Nevertheless,wefindevidenceforasignificantvariation
troscopy is of paramount importance in disentangling the
ofthelineenergyoverthe35dayperiod:theFelineemission
differentspectralcomponentsobservedinHerX-1.Pastob-
originates in near neutral Fe (Fe XIV or colder) in the low
servations using Einstein (McGray et al. 1982) and Bep-
andshort-onstateobservations,whereasinthemain-onthe
poSax (Oosterbroek at al. 1997, 2000) showed that, during
observedFeKαcentroidenergies(6.65±0.1keVand6.50±
themain-onstate,themaximumofthethermalcomponent
and the power-law components are shifted by ∼ 250◦. The 0.02 keV as measured using PN at Φ35 = 0.02 and 0.17)
correspondtoFeXX-FeXXI(Palmerietal.2003).Theline
departure from a phase shift of 180◦ is usually associated
centroids observed using the EPIC PN deviate by 4σ from
with a disk having a tilt angle, under the assumption that
the 6.40 keV neutral value. This has been already noticed
thesoft X-raysoriginate inthelayersofthediskthatinter-
byR02, whosuggested threepossible explanationsfor both
ceptandreprocesstheneutronstarbeam.Thefirstevidence
theline broadening and the centroid displacement:
forachangeinthepulsedifference,asmeasuredatthreedif-
ferent Φ35 using XMM-Newton, has been reported by R02. (i) anarrayofFeKαfluorescencelinesexistsforavariety
R02 found that during the main-on state (Φ35=0.17) the of charge states of Fe (anything from Fe I-Fe XIII to Fe
soft and the hard X-ray pulse profiles were strongly anti- XXIII);
correlated and that the phase shift between the soft and (ii) Comptonization from a hot corona with a significant
hard maxima was ∼150◦. This was in contrast to previous opticaldepthforanarrowerrangeofchargestatescentered
observationsmadeduringthemain-on.Moreover,thephase aroundFeXX.Similarlinebroadeninghasbeenobservedin
shiftbetweenthesecomponentswassignificantlydifferentin somelowmassX-raybinarieswithASCA(Asaietal.2000),
theXMM-NewtonobservationsperformedatΦ35=0.26and and Her X-1 spectra exhibit one of the larger equivalent
0.60. This led to the suggestion that, during the first three widths within theclass;
exposures, the tilt angle of the accretion disk was chang- (iii) Keplerian motion: if this is responsable for the line
(cid:13)c 0000RAS,MNRAS000,000–000
10
broadening, the Keplerian velocity measured at Φ35 =0.02 6.4 The ∼7 keV Fe line
and0.17 is∼15500 and∼13000 km/sec,respectively.This
givesaradialdistanceof∼2−3×108 cm(foraneutronstar OurobservationshaverevealedanFeXXVIlineat∼7keV,
detected during low-state. The line is not detected during
of 1.4 solar masses), which is close to the magnetospheric
radius for a magnetic field of ∼1012 G. the main-on when the X-ray spectrum is dominated by the
strong continuum emission. This feature has not been de-
tected by Leahy et al. (2001) using Ginga data. However,
In addition to these possibilities, we also notice that
Ginga has a relatively low spectral resolution around these
the region responsible for the Fe Kα line emission is likely
energies (18 percent at 6 keV). ASCA observed the source
to be different for lines observed at different beat phases.
with better spectral resolution (2 percent at 5.9 keV using
TheFelineshavedifferentenergies,flux,andbroadeningin
theSISdetectors)duringanearlymainonstate,andbased
the main-on compared to the low state. While data taken
on these data Endo et al. (2000) reported the first resolu-
duringthemainonclearlysuggestacorrelationbetweenthe
tion of the broad feature into two components, at 6.41 keV
fluorescentFeKαlineandthesoftX-rayemission,thesame
and 6.73 keV. However, this double structure has not been
is not explicitly evident in data taken during thelow state.
detectedinotherobservationstakenduringthemain-onnor
Atsuch phases, instead, theline isa factor >5weaker and
is it present in XMM-Newton data at similar phases.
is clearly modulated with the orbital period.
The EPIC detectors on board XMM-Newton have a
The correlation between the UV and the neutral Fe K
spectral resolution similar to ASCA at 6 keV. However,
pointtoacommonoriginwiththefastriseUVW1emission.
XMM-Newton has a much greater effective area allowing
Still, it is hard to distinguish the contributions from the
lines to be detected with a greater confidence, in particular
companion and the disk to the Fe K line emission. A large
duringthestates oflowintensityof thebinarysystem.The
contributionfrom thecompanion would explain whytheFe
feature at ∼ 7 keV has been observed by XMM-Newton
K line is so strong in thelow state of Her X-1 compared to
for the first time over several 35 day phases. Also, it has
other accretion disk sources: in Her X-1 the companion is
been confirmed by a Chandra HETGS observation of the
muchlarger(2.3M⊙)thanforotherLMXBs.However,due source (the only one made during the low state) taken at
to the fast rise of the Fe K flux with the orbital phase, the
Φ35 = 0.44−0.46 (Jimenez-Garate et al. 2003). The fea-
disk origin scenario is still strong.
ture cannot be produced by fluorescence, and it is more
A possible contribution to the Fe Kα emission may
likely to be a Fe XXVI line originating in widely extended
arisefromrelativelycoldmaterialinanaccretiondiskwind,
photo-ionized plasma. On the other hand, RGS data taken
suchascommonlyobservedincataclysmicvariables(seee.g.
during the low and short-on states also show the presence
Drew1997).Althoughnoconsensushasbeenreachedinthe
of photo-ionized gas (Jimenez-Garate et al. 2002). Grating
case of low-mass X-ray binaries, in the case of HerX-1 evi-
spectraexhibitseveralnarrowrecombinationemissionlines,
denceofoutflowinggas,possiblyawindormaterialablated
themostprominentbeingCVI,NVI,NVII,OVII,OVIII
from HZ Her, has been reported by Anderson et al. (1996)
and Ne IX. The line ratio G = (f +i)/r, as computed for
and Boroson et al. (2001). UV lines observed with theFOS
all the helium-like ion complexes, is G≃4, which indicates
andSTISspectrographsontheHubble Space Telescope per- ⋆
that photoionization is the dominant mechanism. More-
sisteveninthemiddleoftheeclipse,whentheX-rayheated
over, RGS spectra shows two radiative recombination con-
atmosphereofthenormalstarandtheaccretiondiskshould
tinuaofOVIIandNVII,consistentwithalowtemperature
be entirely hidden from view. The velocities inferred from
of the emitting plasma (30000 K <T < 60000 K, Jimenez-
thebroadeningoftheNVlinesarenotconstantoveratime
Garate et al. 2002), which furtherly validate the photoion-
scale of ∼ 1 yr, while a comparison between observations
ization model.Noneofthesefeaturesisdetectedduringthe
taken at φ =0.10, 0.21 and 0.29 show they are stable
binary main-on state (where instead theremight be evidence for a
overtheorbitalphase.Betweenthesephases,thevelocityof
broad O VII i line, and perhaps O VIII Ly alpha). The re-
theneutronstarisexpectedtochangeby≈60km/s.Inthis
combinationX-raylineemissionarenotlikelytooriginatein
respect we note that the Fe Kα line was not been detected
HZ Her, dueto the absence of UV induced photoexcitation
by XMM-Newton during the middle of the eclipse, and the
signatures in the He-like triplets (observed with HETGS,
upperlimit on thelinefluxis∼10orlessof thatmeasured
Jimenez-Garate privatecomm.).
outside the eclipse. This in turn sets an upper limit on a
Instead, as suggested by Jimenez-Garate et al. (2002),
potentialcontributiontotheemission ofthislinebyawind
an extended, photoionized accretion disk atmosphere and
or some sort of circumstellar material. There is no Doppler
corona may be responsable for such features. Jimenez-
signatureofawindintheHETGspectrumoftheFeKline
Garate et al. (2001) computed the spectra of the atmo-
(Jimenez-Garate et al. 2003), and the wind would have to
spheric layers of a Shakura-Sunyaev accretion disk, illumi-
be enclosed by the Roche lobe due to the absence of Fe K
nated by a central X-ray continuum. They found that, un-
in eclipse.
der these conditions, the disk develops both an extended
On the other hand, taken as a body the data reported
corona which is kepthot at theCompton temperature, and
heresuggestacomplexorigin fortheoverallemission ofthe
a more compact, colder, X-ray recombination-emitting at-
Fe Kα line. To our knowledge, a complex of lines which in-
mosphericlayer(seeFigure9inJimenez-Garateetal.2002).
cludeallionizationstatesfromFeXVIIItoXXIVFeKαhas
Theparametersofthecompactatmosphericlayerareconsis-
notbeenobservedinanyastrophysicalsource.Thismaystill
indicate an outflow of relatively cold gas or some complex
dynamics in the disk/magnetosphere interface. Such phe- ⋆ Herei,randf denotetheintercombinationlineblend,theres-
nomena should be time-dependent and may be monitored onance lineand theforbiddenline;forcollisionallyionized gases
in thefutureusing Astro-E2. itisG<1(Liedahl1999, Porquet&Dubau2000).
(cid:13)c 0000RAS,MNRAS000,000–000