Thomson Scattering and Collisional Ionization in the X-rays Grating Spectra of the Recurrent Nova U Scorpii

1.4 MB·English

by  M. Orio

Mon.Not.R.Astron.Soc.000,1–13(2002) Printed4January2013 (MNLATEXstylefilev2.2)
Thomson Scattering and Collisional Ionization in the
X-rays Grating Spectra of the Recurrent Nova U Scorpii
3 M. Orio,1,2 E. Behar,3 J. Gallagher,2 A. Bianchini,4 E.Chiosi,1, G.J.M. Luna,5
1
T. Nelson,6 T. Rauch,7 B.E. Schaefer8 and B. Tofflemire2
0
2 1INAF–Osservatorio di Padova, vicolo dell’ Osservatorio 5, I-35122 Padova, Italy
2 Department of Astronomy, Universityof Wisconsin, 475 N. Charter Str., Madison WI 53704
n
3Department of Physics, Technion, Haifa, Israel
a
4 Astronomy Department, Padova University,vicolo dell’ Osservatorio 3, I-35122 Padova, Italy
J
5 Institutode Astronoma y Fisica del Espacio (IAFE/Conicet), CC67, Suc. 28 C1428ZAA CABA, Argentina
3 6 School of Physicsand Astronomy, University of Minnesota, 116 Church St SE, Minneapolis, MN 55455
7 Institute for Astronomy & Astrophysics, KeplerCenterfor Astro & Particle Physics,
]
Eberhard Karls University,Sand 1, 72076 Tu¨bingen, Germany
R
8 Physicsand Astronomy, Louisiana State University, Baton Rouge, LA 70803
S
.
h
p Received:
-
o
r
t ABSTRACT
s We present a Chandra observation of the recurrent nova U Scorpii, done with the
a
HRC-Sdetector andthe LETGgratingonday18 after the observedvisualmaximum
[
of 2010, and compare it with XMM-Newton observations obtained in days 23 and 35

2 after maximum. The total absorbed flux was in the range 2.2-2.6 ×10−11 erg cm−2
v s−1, corresponding to unabsorbed luminosity 7-8.5 × 1036×(d/12 kpc)2 for N(H)=2-
6 2.7×1021 cm−2.Onday18,70%ofthesoftX-trayfluxwasinacontinuumtypicalof
2
a veryhot white dwarf (WD) atmosphere,which accountedfor about 80% of the flux
0
on days 23 and 35. In addition all spectra display very broad emission lines, due to
5
higher ionization stages at later times. With Chandra we observed apparent P Cygni
.
2 profiles.Wefindthatthesepeculiarprofilesarenotduetoblueshiftedabsorptionand
1
redshiftedemissioninphotoionizedejecta,liketheopticalPCygofnovae,buttheyare
2
ratherasuperpositionofWDatmosphericabsorptionfeaturesreflectedbythealready
1
discoveredThomsonscatteringcorona,andemissionlines due tocollisionalionization
:
v in condensations in the ejecta. On days 23 and 35 the absorption components were
i no longermeasurable,havinglostthe initiallargeblue shift thatdisplacedthem from
X
the core of the broad emission lines. We interpret this as indication that mass loss
r
a ceased between day 18 and day 23. On day 35, the emission lines spectrum became
verycomplex,withseveraldifferentcomponents.Modelatmospheresindicatethatthe
WD atmospheric temperature was about 730,000 K on day 18 and reached 900,000
K–one million K on day 35. This peak temperature is consistent with a WD mass of
at least 1.3 M⊙.
Key words: stars: novae, cataclysmic variables – stars: winds, outflows — stars:
individual(UScorpii)–whitedwarfs–X-rays:binaries–X-rays:individual(UScorpii)
1 INTRODUCTION 2005 and references therein). The models predict that RN
occureither with moderately high accretion rate m˙ on WD
The recurrent nova (RN) U Scorpii is one of a handful of ofabout1M⊙,oratveryhighm˙ (M>10−8 M⊙ year−1)on
such objects known in the Galaxy, although this class may
WD with M> 1.2 M⊙. The velocity of the ejecta increases,
be much more numerous than we have been able to assess
and the duration of the outburst decreases, with the WD
so far. “Recurrent” means that the thermonuclear flash is
mass. About half of all RNe are symbiotic novae, that is
repeated on a human timescale. The outburst supposedly
systemswith agiantsecondary andahot WD,withorbital
recurs also in classical novae, however the time scales can
periodsofhundredofdays(asopposedtomostclassicalno-
beaslongas severaltimes105 yearson WDat thelow end
vae,whichhaveorbitalperiods ofhoursandmain sequence
of the nova mass distribution (0.6-0.7 M⊙), accreting mat- companions).
ter at a low rate of 6 10−11 M⊙ year−1 (see Yaron et al.
2 M. Orio et al.
the event occurred, a strategy was in place to observe the
1 outburst at all wavelengths.
InthispaperwepresentgratingX-rayspectra,theonly
U Scorpii
Swift/XRT means to probe the effective gravity, chemical composition
[0.3-10keV] andeffectivetemperatureoftheunderlyingwhitedwarfonce
0
the photosphere shrinks back to the WD radius (e.g. Nel-
sonetal.2008andreferencestherein).Oneobservationwas
proposedbyus,17 daysafterthevisualmaximumwith the
-1 Chandra LETG grating. We adopt here the epoch of the
observed visual maximum 2010-01-28.1 (JD 2,455,225,605)
as the initial time (although Schaefer, 2010, extrapolated
thelightcurveconcludingthatthepeakhadoccurredafew
-2 hours earlier). Two XMM-Newton observations were done
on2010-02-19at15:41:09UT,andon2010-03-14at14:33:32
UR,22and35daysaftermaximum,respectively(notethat
for the different XMM-Newton detectors the beginning of
-3
theobservationsvariesslightly,andfordetailsseealsoOrio
et al. 2011, Nesset al., 2011a and 2012).
InadditiontopresentingandanalysingtheearlyChan-
dra spectrum, in this paper we compare it with the XMM-
-4
0 20 40 60 NewtonRGSgratingspectra.Theproposersofoneofthese
days after 2010-01-29
observations, Ness et al. (2012) have analysed the X-rays
and UV data and examine thetime variability in these two
Figure1.SwiftXRTcountrateversustimemeasuredsincesince
bandsindetail.TheyfindthatapartialeclipseoftheX-ray
therecordedoutburston2010-01-29, inthe0.2-10keVrange.
continuumcanbeexplainedonlyifthehotcentralsourceis
observedafterhavingbeenThomsonscattered,aconclusion
that we reached independently (Orio et al. 2011), and with
U Sco has an orbital period of 1.23 days and belongs whichwefullyagree.Theseauthorsremarkaboutthelackof
to the class of RNe with a near-main sequence or moder- absorption lines, expected and usually observed in novaein
ately evolved secondary (Schaefer 1990, Schaefer & Ring- outburst in X-rays, and attribute it to possible “smearing”
wald1995).Thesecondaryspectrumisnotdetectedinopti- byThomsonscattering.Theyalsodescribeandexaminethe
cal spectra at quiescence, butat a late post-outburst phase spectra, proposing lines identifications and attributing the
Anupama & Dewangan (2000) measured the spectral fea- emissionlinestoresonantscattering.Inthiswork,wefound
turesofaK2subgiant.Inthepreviousoutburststheejecta that the nova development with time, including the XMM-
were extremely depleted in hydrogen (Barlow et al. 1981), Newtonobservations,isfundamentaltofullyunderstandthe
so the secondary must have lost a significant fraction of its physics of the X-ray grating spectra. Without the intent to
hydrogenrich envelope.At quiescence,HeII linesare dom- replicate Ness et al.’s (2012) work, we analyse and attempt
inant in optical spectra and the hydrogen lines are absent to model with detailed physics all the grating X-ray spec-
orweak,afact that hasbeen interpretedasaccretion of He tra, thereby reaching different conclusions than the above
rich material from a H-depleted secondary (e.g. Duerbeck authors on the origin of the emission lines observed with
et al. 1993). Although Maxwell et al. (2011) cast doubtson XMM-Newton and on the reason for the lack of absorption
thehighHeabundance,thepossibilityofveryenhancedhe- features at late stages.
lium in U Sco is very interesting as it may result in a type In Section 2 we describe the observations, the data ex-
Ia SN (SN Ia) without a hydrogen rich envelope, which is tractionmethod,andwhentheobservationsoccurredinthe
consistent with the upper limits on hydrogen in the SNeIa context of the long term X-ray evolution. In Section 3 we
spectra. show the short term light curve, which we find relevant to
TheshortandluminousoutburstofUScowasobserved understand the physics of the spectra. In Section 4 we dis-
10 times (1863, 1906, 1917, 1936, 1945, 1969, 1979, 1987, cuss the apparent P Cyg profiles observed in the Chandra
1999, 2010), with peak magnitude V≃7.5, decay times by2 spectrum. After having analysed the origin and meaning of
and 3 optical magnitudes, t2=1.2 and t3=2.6 days respec- thesefeatures,weperformedatmosphericfittingforallspec-
tively, and a return to quiescence within 67-68 days. These tra,andevaluatedtemperatureevolutionandpeaktemper-
parametersindicateaverymassiveWD,suchthatsufficient ature(Section4).InSection5wediscussthephysicalmech-
pressureforathermonuclearrunawayisbuiltupatthebase anisms producingtheemission lines and howwe attempted
oftheenvelopeevenwithasmallaccretedenvelope(e.g.few a global fit;finally conclusions are presented in Section 6.
610−7 M⊙, see Starrfield et al. 1988). We adopt 12±2 kpc
as distance to the nova, measured thanks to its eclipse in
theoptical (Schaefer et al. 2010).
2 THE OBSERVATIONS, AND THEIR
Since U Sco has shown outbursts on an almost regular
POSITION IN THE LONG TERM LIGHT
timescaleof10±2years(assumingacoupleoferuptionswere
CURVE
missedbecausetheyoccurredwhenthenovawasbehindthe
sun),Schaefer(2010)wasabletopredictthelasteruptionof TheSwiftsatellitemonitoredUScoalmostdailyonalldates
Januaryof2010withanuncertaintyofonlyoneyear.When when it was technically possible. We extracted the Swift
X-ray spectra of U Scorpii 3
threespectraarelabelled(foradditionallinesidentification,
see Table 1).
WeseeinFig. 2thatforat least threelines(theNeVI
Heγ linewithrestwavelength23.77˚A,theNVIIhydrogen
αline,withrestwavelength24.78˚A,andtheresonanceline
oftheNVItripletat28.78˚A)theemissionportionappears
red-shiftedandtheabsorptionwingblue-shifted.Thisisin-
triguingly similar to the P Cyg profiles observed in optical
spectra.
We compared the Chandra spectrum with archival
XMM-Newton grating spectra obtained on 2010-02-19 at
15:41:09 UT and on 2010-03-3 at 14:33:32 UT, (5 and 17
dayslater,respectivelydays23and35).TheXMM-Newton
spectrawereextractedwithXMM-SASversion11.0.2.There
was variability during both XMM-Newton observations, as
wediscussinthenextSection,butinFebruary(day23after
theoutburst)wemeasuredaveragecountrates0.881±0.004
cts s−1 and 0.824±0.004 cts s−1 with the RGS1 and RGS2
Figure 2. The U-Sco spectrum measured with the Chandra gratings respectively, while in March (day 35) the average
LETGgratingon2010-02-14,day18postopticalmaximuminthe count rates were 1.037±0.005 cts s−1 and 0.951±0.004 cts
Vband(+1and-1ordershavebeensummed).Theredsolidline s−1.Theunabsorbed fluxseems was rising moderately, and
shows the best fit with two components: an atmospheric model
wemeasured2.5±0.2×10−11and2.6±0.2×10−11inthesec-
with a 740,000 K Teff and a collisionally ionized optically thin
ond and third observation, respectively. All XMM-Newton
plasma with tenfold He and N abundance abundances with re-
specttosolar,andplasmatemperaturekT=93eV(seeTable2). instrumentswereobservingthenovasimultaneously,includ-
ing the Optical Monitor (OM), and the broad band X-ray
imagers EPIC-pn and MOS1 and MOS2. The RGS spectra
data and present the XRT X-ray light curve in Fig. 1. Al- are shown in Figures 3 to 5. Figures 4 and 5 present the
thoughthispaper’sfocusisonthelongerobservationsdone count rate spectrum (like Fig. 2 for Chandra). The fit with
with Chandra and the comparison with the RGS of XMM- atmosphericandplasmaemissionmodels,alsoshowninthe
Newtondata,theSwift-XRTlongtermlightcurveputsthe Figures, are discussed in detail in Sections 4 and 5.
data in the context of the outburst evolution. The arrows
indicate the times at which the grating observations were
done. The X-ray flux was still on the rise during both the
3 THE SHORT-TERM LIGHT CURVE: A
2010-02 observations (days18 and 23), while themaximum
SIMULTANEOUS X-RAY AND OPTICAL
X-ray flux was recorded around the date of the third and
ECLIPSE
final X-ray grating observation in March, on day 35. The
X-rayspectrum remained remarkablysoft duringthewhole At optical wavelengths, the eclipse of the accretion disk by
post-outburst phase, with more than 95% of thecountsbe- theKdwarfcompanionwasdiscoveredandmeasuredatqui-
low 0.7 keV.A sharp decay followed theMarch observation escencebySchaefer&Ringwald(1995),anditwasobserved
untilday67, when thenovawas almost back toX-raymin- again in the outburst by Schaefer et al. (2011), indicating
imum. Further information on the UV and X-ray evolution thattheaccretiondiskhadalreadybeenreestablishedbythe
observed with Swift can be found in Ness et al. (2012). third week after the optical maximum. The orbital period
The first grating observation of U Sco was proposed of U Sco is 1.23054695 days (Schaefer & Ringwald 1995).
by us and was done with the LETG/HRC-S instrument of For the ephemeris we assume in this paper the result of
theChandraobservatoryon2010-02-14, at11:38:22UT.As the linear fit for the whole 1999-2010 inter-eruption inter-
stated above,thiswas day18 afteroptical maximumin the val,withaHJDtimeoftheopticalminima2451234.5387 +
V band (see Schaefer et al. 2010). The exposures lasted for N×1.23054695. Wecautionthatthemanyobservedeclipses
23 ks (6.38 hours). The observed count rate spectrum is duringthelast foureruptionsshowsignificant deviationsof
showninFig.2,whileFig.3showsthefluxedspectrum,com- theminimumfromthisephemeris,withtheoffsetschanging
pared with the spectra measured with the XMM-Newton throughout the eclipse. This means that the center of opti-
RGSgratings on days23 and 35. Fig. 2also shows a model callightcurvechangesthroughouttheeruptions,andisnot
fit, discussed in detail in Sections 4 and 5. In order to ex- exactly at thesame position duringthequiescent periods.
tracttheLETGspectrum,weappliedthemostupdatedcal- WedidnotdetectsignificantvariabilityintheChandra
ibration to the LETG pipeline processed level 2 event and zeroorderlight curve,butall theobservation occurred out-
phaChandradatafiles.Responsematricesandancillary re- sidetheeclipseobservedinoptical.Theexposurestartedat
sponse(effectivearea)fileswerecreatedusingthetoolsmk- orbital phase0.4463 and covered almost a quarter(22%) of
grmf and mkgarf, part of the CIAO v4.2 package. Wemea- theperiod.
suredasummed±1orderscountrateof0.477±0.008ctss−1 Both XMM-Newton observations covered about 60%
in the 0.2-0.8 keVenergy range in which fluxis measurable of the orbital period and thus overlapped with the optical
above the background (corresponding to approximately 15- eclipse time. The day 23 observation started at phase 0.84,
60˚Ainwavelengthrange).Wemeasuredanunabsorbedflux and theone of day 35 started at phase 0.55.
2.2±0.2 erg cm−2 s−1. In Fig. 3, the strongest lines of the In Fig. 6 we present the XMM-Newton EPIC pn light
4 M. Orio et al.
Figure3.Measuredfluxversuswavelengthinthethreegratingspectraatdays18,23and35afterthediscovery,TheChandraspectrum
isdisplayedinblack,theaveragedRGS1-RGS2spectrumof2010-02-14isplottedininredandthe2010-03-02RGS1-RGS2spectrumis
inblue.
Figure 4. The XMM-Newton-RGS-1 (left) and RGS-2 (right) spectra of U Sco on day 23 after the optical maximum. The blue line
showsthefitwithaWDatmospherewithabsorptionlinesblue-shiftedby2000kms−1,andonlythefirstBVAPECcomponentinTable
2(above) oraddingthesecondBVAPECcomponents fortheRGS-2hardportionofthespectrum(smallpanelbelow).
curves and compare them with the optical light curves of not regular) period of about 3 hours, extensively discussed
Schaeferet al. (2010) measured in thesame twoweeksdur- byNessetal.(2012).Theseauthorsalsoextractedthespec-
ingwhichthetwoX-raydatasetsweretaken.Alllightcurves trum in and out of the oscillation, showing no variation of
arefolded overtheopticallymeasured orbitalphaseandre- the emission lines. The authors attribute the phenomenon
peated twice. For theX-ray light curve,we plotted a quan- to large clumps, of absorbing material at a distance about
tity that is 2.5 times the logarithm of the X-ray flux nor- 3.5 R⊙ from the WD, intersecting the line of sight in the
malized to its minimum vale, for a direct comparison with direction of the accretion stream, during the process of re-
optical magnitudes. The EPIC-pn exposures are 63185 and establishmentofthedisk.Wedonotwishtofurtherdiscuss
62318 seconds long respectively. theseoscillations,althoughwesuggestthatthissemi-regular
On day 23 in the X-ray light curve the most evident period(evenifitisassociatedwithejectionofparcelsofma-
oscillation is a pronounced one with a semi-regular (albeit
X-ray spectra of U Scorpii 5
Figure 5.TheXMM-Newton-RGS-1(left)andRGS-2(right)spectraofUScoonday35after theopticalmaximum.Thefitisshown
inred,theparametersaregiveninTable2.
Table1.Rest,measuredwavelength,andfluxesinphotonscm−2 ks−1 fortheemissionlinesidentifiedand
measuredinthethreegratingspectra.Seethetextforanevaluationoftheerrorsinthefluxestimate.
Chandra02-14 XMM02-19 XMM03-05
line λ0 (˚A) λm (˚A) flux λm (˚A) flux λm (˚A) flux
CVILyα 33.7342 33.78±0.02 0.39 33.78±0.05 0.43 33.75±0.05
NVIf 29.5346 29.50±0.02 0.93 29.50±0.03 1.23 29.60±0.05 0.35
NVIi 29.0819 29.16±0.02 0.91 29.20±0.02 1.65 29.22±0.05 0.42
NVIr 28.7800 28.93±0.02 2.20 28.82±0.05 2.10 28.80±0.05 0.31
CVILyγ 26.990 26.80±0.02
NVIHeβ 24.90 63.8 24.95±0.01 61.3
NVIILyα 24.78 24.94±0.03 1.71 24.92±0.03 5.64 24.80±0.02 2.51
NVIHeγ 23.771 23.81±0.05 0.43 23.81±0.05 0.60 23.82±0.02 2.70
OVIIf 22.097 22.108±0.050 22.13±0.05 0.30
OVIIi 21.801 21.807±0.020 60.24 21.80±0.02 0.49
OVIIr 21.602 21.603±0.050 0.35 21.61±0.02 0.39
OVIIILyα 18.969 19.05±0.03 0.74 19.00±0.02 3.21
OVIIHeβ 18.627 18.57±0.05 18.55±0.02
FeXVII 16.78 16.78±0.05 16.80±0.05
OVIILyβ 16.00 15.95±0.03 16.00±0.02 0.44
FeXVII 15.262 15.30±0.05 15.25±0.05 >0.10
FeXVII 15.01 15.15±0.02 15.03±0.05 0.12
FeXVIII 14.207 14.22±0.03 0.13
FeXIX 14.667 14.70±0.05 14.67±0.03
NeIXf 13.6984
NeIXi 13.5503 13.57±0.05 0.12
NeIXr 13.4474 13.45±0.02 13.48±0.05 0.15
NeXα 12.1321 12.14±0.05 0.15
NeIXHeβr 11.55 11.60±0.10
MgXIHer 9.09 9.20±0.03 9.20±0.03
terial), may represent the disk rotation period, which may O VII triplet). Extracting the spectrum over the short du-
beslightly variable while the disk is still being formed. ration of the eclipse we find a variation of this line only at
the2 σ level, so wesuggest that thisis not statistically sig-
A clear orbital modulation in theday 35 spectrumwas
nificant. In any case, the main purpose of Fig. 6 is to show
noted by us and by Ness et al. (2012). These authors also
thattheorbitalmodulationseemsalreadyclearlypresentin
stress that almost all emission lines observed in the spec-
theXMM-Newton EPIC-pnlight curveof day 23, although
tra are not eclipsed, while the continuum is, and we agree
it is not as pronounced as on day 35. On day 35, the aver-
with their analysis. However, we cast some doubts regard-
agecountrate duringorbitalphases 0to0.12 and0.88 to1
ing the dimming in eclipse of only one specific line, the O is 0.75 and 0.84 cts s−1 with RGS1 and RGS2 respectively,
VIII Lyα (the above authors remark this is the only line while we measured 1.10 and 1.05 cts s−1 with RGS1 and
that seems to vary,even there is clearly no variation of the
6 M. Orio et al.
RGS2 respectively during the rest of the observation. We theOI 1s–2p transition at ∼23.476 ˚A,which doesoccurs in
find that the variation in count rate in and out of eclipse, theISMalong theline of sight toU Sco. Thefact that this
outsidethe≃3hoursoscillations,isbyatleast25%evenon absorptionfeaturewashardlymeasurableinNLMC2009,in
day 23. one LETG spectrum that appears similar to this one (Orio
Becausetheeclipsetakesoutonlyabout50%oftheflux et al. 2011, and 2012 article in preparation), confirms the
in the X-ray continuum, on day 35 a large portion of this identificationbecauseNLMC2009isinadirectionofmuch
continuum must originate in a much more extendedregion. lowerinterstellar absorption (N(H)≃4×1020 cm−2).Thus,
We assume that this is a Thomson scattering corona, re- we realize that at least this particular P Cyg-like profile
processes a fraction of the WD radiation in an achromatic is in fact only a sort of “pseudo P Cyg”, composed of an
way (Ness et al. 2011a, 2012, Orio et al. 2011). The incli- absorption and an emission feature that do not have origin
nation angle in U Sco is 82.7o±2.9o (Thoroughgood et al. in the same medium.
2001),soitisquiteconceivablethatwedidnotobservemost We examine now the P Cyg profiles of the other two
of the WD flux directly. The accretion disk was placed by strong lines, shown in Fig. 7, to understand whether they
Schaefer et al. (2011) at 3.4 R⊙ distance from the WD in may be due to the “real”, classic P Cyg phenomenon. As
a light curve model for day 35. Because of the nearly edge- Fig. 7 shows in velocity space, the N VII emission appears
on inclination, the short-duration and deep X-ray eclipse blue-shifted by 2200±200 km s−1 in absorption, and red-
does not appear to be due to the disk. We assume that shifted by the same amount in emission. The N VI res-
the source of the continuum is eclipsed by the secondary, onance line with rest wavelength 28.78 ˚A is instead blue-
exactly like for the eclipse of the disk by the secondary ob- shifted by2700±300 km s−1,in absorption, andred-shifted
served at quiescence at optical wavelength. Following the by 1500±300 km s−1 in emission, so the two components
reasoning of Schaefer et al. (2011, see their Fig. 13 for the do not cross around zero, but rather towards the red. In
optical eclipse) we conclude that most of the continuum is optical spectra of novae, crossing in the red is observed in
observedasreflectedatadistanceequalorlargerthanthea spectra taken shortly before the P Cyg profile disappears.
3.4R⊙ distancefromtheWDatwhichSchaeferetal.place Typically,thePCygprofileisnolongerobservedinthefol-
theouter rim of the disk. lowing weeks because the blue-shifted absorption becomes
We intend to show that the Chandra spectrum with negligible compared to the emission. Fig. 8 shows thesame
its absorption features is best explained assuming that this line profile in velocity space in the RGS spectra of days 23
scattering corona was already present, and at comparable and 35.
distancefromtheWD,onday18,althoughnoX-rayobser- P Cyg profiles in X-ray spectra in general are rarely
vations at eclipse minimum were done. observed, and the absorption wing, if measurable, is very
shallow. This is because in X-rays we are likely to observe
outflows and nebulae around compact objects, while in op-
tical we observe them around stars with at least hundred
4 FITTING THE SPECTRUM, STEP BY STEP: of times larger radii. This includes the optical (as opposed
THE WD toX-rays)spectraofclassical andrecurrentnovae,whereP
Cyg profiles at optical wavelengths are observed when the
4.1 The absorption lines observed with Chandra
photospherearoundtheWDhasnotrecededtoWDdimen-
Two puzzles are immediately apparent when we examine sions yet, but is still of the order of at least a solar radius.
Figures2to5.Firstofall,theChandraspectrumshowsdeep However, this typically occurs weeks or months before the
absorption features of nitrogen, while the XMM-Newton supersoftX-rayphase,andweexpectitonlyinthefirstfew
ones do not display any features in absorption. Although days, or even in the first hours, in a very fast nova like U
model atmospheres show that the absorption becomes less Sco.PCygprofilesatopticalwavelengthsareduemainlyto
pronounced at high temperatures (Rauch et al. 2010), the resonancetransitions.Theblue-shiftedabsorptionisobserv-
models predict deep absorption features of nitrogen, al- ableonlywherethematerial movesin ourdirection,that is
ways detected in the previous X-ray gratings’ observations along the line of sight of the photo-ionizing central object.
of other novae (e.g. Rauch et al. 2010, Ness et al. 2011b). The optical P Cyg absorption features occur, like theemis-
The other surprising element, as we already noted, are the sion,inthe“wind”ornebulaaroundaphoto-ionizingcentral
apparent P Cyg-like profiles in the Chandra spectrum. P- object, but they are observed in a cylinder with the radius
Cyg profiles are typical of the optical spectrum of novae in of the central object, and only as long as the radius of the
X-rays, but they have never been observed in their X-ray cylinder is not negligible compared to the rest of the shell,
gratingspectra,except(possibly)inthefirstXMM-Newton from which we observe theemission lines.
spectrum of another fast recurrent nova, LMC 2009 (Orio Luminous supersoft X-raycontinuum - observed in the
2012). Chandra spectrum as well as in the XMM-Newton ones -
Emissionlinesattributedtotheejectawereobservedin in novae arises when the central object has shrunk and re-
theX-raysspectraofseveralothernovae(Drakeetal.2003, gainedcompactdimensions,witharadiusoftheorderofat
Nessetal.2003, Nelsonet al.2008, Rauchetal. 2010). Are most 109 cm, or else the X-ray luminosity would be largely
theabsorptionwingsofthese“PCygprofiles”inthisUSco super-Eddington. Shortly after the outburst, velocities up
spectrum produced in the nova shell, instead of being due to 5000 km s−1 were measured by Yamanaka et al. (2010),
to the WD? At least one profile cannot be a P Cyg in the and on day 35 (the day of the last XMM-Newton observa-
conventional sense, and it is the one of the N VI He γ line. tion), Mason et al. (2012) estimated an expansion velocity
The absorption in this case is not the same line, but it is of 4000 km s−1 from thefullwidth at half maximum of the
mainly of another overlapping line at zero velocity, due to opticallines.Assumingthisvaluefortheexpansionvelocity,
X-ray spectra of U Scorpii 7
0 0
Magnitude difference 0.5 Magnitude difference 0.5
1 1
1.5 1.5
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Orbital phase Orbital phase
1.5 1.5
1 1
2.5*Log(EPIC-pn count rate/9.21) 0. 50 2.5*Log(EPIC-pn count rate/11.0) 0. 50
-0.5 -0.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Orbital phase Orbital phase
Figure6.TheaverageopticalVmagnitudelightcurveofUScoduringdays21-26(lefttoppanel)anddays32-41aftervisualmaximum
(right top panel; Schaefer et al. 2011), compared with the observed X-ray lightcurve measured with the EPIC-pn instrument aboard
XMM-Newton on days 23 (left bottom panel) and on day 35 (right bottom panel). The logarithm of the count rate normalized to the
minimum value of the measurements and multiplied by 2.5 is plotted as a function of phase, to match the magnitude scale. The light
curvewasbinnedintimewith50sbinsbeforeconversiontophase,onday23(right)and35(left)aftertheoutburst.
the nebular ejecta around U Sco at the time of the Chan- et al. 2008). Since the emission lines were strong in a very
dra observation (day 18) had reached a distance from the different range, therewas no overlap with WD lines and no
WD representing almost 300,000 times the WD radius of a apparent P-Cyg profile. We conclude that the P Cyg pro-
very massive WD (1.3 M⊙). The region in which emission files in the Chandra spectrum were not “real P Cyg”, but
occurred was overwhelmingly large in comparison with the theywereratherduetosuperimposedabsorptionandemis-
cylinder around the line of sight towards the WD. We con- sion arising in different regions and different mechanisms.
clude that the absorption features observed with Chandra Theabsorptionfeatureswereproducedintheoutmostlayer
cannothavebeenproducedintheejecta.Byday18,anyab- of the WD atmosphere, which was still loosing mass, and
sorption features oftheejecta would already behiddenand reflected at a distance of a few solar radii by the Thom-
embeddedin thebroad emission lines. There isno evidence son scattering corona of the WD. The corona’s “reflection”
ofnebularX-raysabsorptionlinesintheejecta,butinother allowed the lines to be detectable, despite the high inclina-
novae absorption lines were always observed, and they are tionangleoftheWDinUSco.Sincethiscoronamusthave
usually attributed to the WD atmosphere. There is only a already been present on day 23, we assume that it existed
recent case in which the blue-shifted lines were attributed alreadyonday18(evenifwecouldnotmeasurealightcurve
insteadtocollisional ionization inathinoutershell,thatof to proveit).
V2491Cyg(Pintoetal.2012),buttheabsorptionlinesofU
An intrinsic blue-shift that lasted while the WD was
Sco are so broad compared to this model, that we can rule
still loosing material in a wind, explains why the features
out this origin.
werestillobservableonday18,eveniftheemissionfeatures
are so broadened in U Sco that they should hide the atmo-
We concluded that the velocity of the absorption fea-
sphericabsorption.OncethewindfromtheWDceased,the
turesalsoforUScowasduetoadifferentphenomenon,one
absorption lines velocity became close to zero, so they were
causinganintrinsicvelocitythathadanalmostabruptend,
embedded in the center of the emission lines and were not
so that the features were still significantly blue shifted on
measurable any longer.
day 18, but not any more a few days later. This process
must be mass loss from the nova. Absorption lines in X- Whatever absorption features may have actually been
ray grating spectra of other novae in the early phases were producedintheejectabyphoto-ionization,iftherewassuch
significantly blue-shifted, with comparable blue shift to the aphenomenon,wereembeddedintheemissionfeatureslong
apparent“absorption wing”ofthe“PCyg”intheChandra beforeday18,duetothesmalldimensionsoftheabsorption
spectrum of U Sco (e.g. Ness et al. 2003 and Rauch et al. regioncomparedtothefastejecta.Itisalsoconceivablethat
2010forV4743Sgr,Nessetal.2011bforV2491Cyg).InRS the Thomson scattering corona changed in geometry and
Ophmanyofthesameabsorptionlinesasintheformertwo the WD absorption lines may have become more smeared
novae were observed without measurable blue-shift (Nelson out between day 18 and 23, and perhaps better “hidden”.
8 M. Orio et al.
0.2 Although the Chandra spectrum absorption features
were eroded by the blue wing of the partially overlap-
ping emission lines, we note that they were still remark-
ably close in depth and broadening to those observed in
0.15
other novae without X-ray eclipses, such as V4743 Sgr
med counts (WReauacdhopettedal,th2e01a0tm) oasnpdheRricS mOopdhels(Ndeelsvoenlopeetdabl.y2c0o0a8u)-.
m 0.1 thor Thomas Rauch, publicly available on the web site
ETG +-1 su hvettrpy:/h/oatstwroh.iutnei-dtuweabrifns,geen.g.d.eR/∼SrOaupchh (Nanedlsondeevteloapl.ed200f8o)r
L and V4743 Sgr (Rauch et al. 2010). We fitted the spectra
0.05
with XSPEC, v12.6, assuming that the absorption features
wereblue-shifted by2000 km s−1 (as in Rauchet al.2010).
To date, Rauch’s models represent the best approximation
0 to the atmosphere of a hydrogen burning white dwarf, al-
-10000 -5000 0 5000 10000
though residual outflows that produce blue-shifted absorp-
velocity in km/s
0.18 tion are not accounted for in these static models. Models
including elements with atomic number up to the one of
0.16 oxygen are available for a grid of values of effective grav-
ity, with a log(g) grid from 5 to 9 with 1 step increments.
0.14
The most recent grid of models, including all elements up
mmed counts 0 0.1.12 twhooittmhcaotgnenmtienpsuieuurmamt,uowrfeastshacebaUolcvuSelca5ot0es0dp,0ef0co0trratKh.eSainhndoctetiseasbstuovwiteha5bit0lee0,df0ow0r0atrKhfse,,
G +-1 su 0.08 tchomeWe lDarmgeulystsbuepveer-rEydcdominpgtaocnt.wanitdhtlhoeg(lugm)<in9o,sairteywpuobullidshbeed-
LET 0.06 only for log(g)=9.
0.04 Model 201 in Rauch’s web page was developed specif-
ically with similar abundances to the ejecta of U Sco mea-
0.02
sured using the optical spectra in the previous outbursts:
0 depletedhydrogen([He]=0.489oraheliummassfractionof
-10000 -5000 0 5000 10000
70%, and almost 50 times enhanced nitrogen with respect
velocity in km/s
tothesolarvalue([N]=1.668).Wefittedmodelsofthisgrid
Figure 7. The P Cyg profiles of the Ne VII Hydrogen α line
with XSPEC, after blue-shifting the absorption features by
(above),andNVIresonanceline(below),intheChandraLETG
spectrum (±1 orders summed) plotted in velocity space. The fit 2000kms−1.Wedonotobtainthebestfitwithmodel201,
withanatmosphericmodel(seeTable2)withabsorptionfeatures butratherwithmodel003,with[He]=0.382and[N]=1.803.
blue-shiftedby2000kms−1isshownbythedottedline.Thetwo Thisisnot surprising,becausetheabundancesintheburn-
weaker lines of the triplet appear on the right of the resonance ing layer can be quite different from those of the ejecta,
line of N VI in emission. There is a contribution of N VI He β where mixing with unburnedmaterial has occurred.
lineinemissionat24.889˚A(seeTable1). The final product in our fit, shown in Fig. 2 for the
Chandra LETG, is a composite model of WD atmosphere
andsuperimposedemission lines,butfortheChandraspec-
In any case the transition giving rise to absorption lines in
trumwe started byperforming an experiment,ignoring the
theatmosphere, evenwhen not observableand measurable,
channels containing the strong emission lines (24.8–25.6 ˚A
must haveoccurred until theWD turned off.
for N VII Ly α, and 28.7–30.0 ˚A for the N VI triplet), so
thatthemodelfitwasbetterconstrained bythecontinuum
and bytheabsorption features. For thecomposite fitsplot-
4.2 Atmospheric models: the hottest WD
ted in blue in Figures 2, 4 and 5 we binned the data with
Becauseoftheconclusionsabove,inordertofittheChandra 30 to 50 counts per bin, but for the initial experiment, in
spectrum,weassumedthattheobservedabsorptioncompo- order to retain the high spectral resolution of the LETG,
nents were not produced by photo-ionization of the ejecta we binned the data only by a factor of 4 resulting in wave-
likeintheopticalPCygprofilesofnovae,butthattheyorigi- lengthbinsof0.01˚A.Withalowresultingnumberofcounts
natedinsidetheWDatmosphere,andwerereflectedatlarge ineachbin,weusedtheC-statistic(Cash,1979)ratherthan
distance(>3R⊙)bytheThomsonscattering.Wefittedonly themore commonly used χ2 statistic. Wefound that the C
the continuum of the XMM-Newton spectrum. We assume parameter is smallest for the smallest C/N ratio (typical
thatonday23and25theatmosphericabsorptionlineswere of CNO ashes), and steadily increases as the C/N ratio in-
simply unobservable, hidden by the unusually broad emis- creases.Thedecreaseofthe“C”parameterwithC/Nratiois
sionlinesofthisnovaoncethemassoutflow,andwithitthe about20%frommodel011tomodel003ofRauch’sgrid.All
blue shift, ceased. Of course in the Chandra spectrum the models with solar abundances yield poor fits to the Chan-
observed absorption may have been more pronounced if it dra LETG spectrum. The temperature for the “truncated-
was not super-imposed on emission, so we can only obtain emission” spectrum in the 2σ confidence range (± 2σ) is
lower limits on theabundances and effective gravity. 650,000-750,000 K. The best fit yielded a temperature of
X-ray spectra of U Scorpii 9
0.8 We note that after the previous outburst in 1999, Ka-
habkaetal.(1999),fittedtheBeppoSAXlowresolution,low
0.7 S/N LECS spectrum with a WD atmospheric model. They
derived aluminosity in therange 2.7 to 18 × 1037 ×(d2/12
0.6
kpc) 19-20 days after the optical maximum, and an effec-
tive temperature of 75 eV (870,000 K). The count rate was
0.5
however only 0.06 cts s−1, a value that appears too low by
S cts/s 0.4 a factor of about 2 if translated into both the Chandra or
RG XMM Newton count rates with PIMMS for a blackbody at
0.3 75 eV and N(H)=2.5 ×1021 cm−2. The BeppoSAX obser-
vation lasted for about 14 hours, so the low count rate is
0.2 not explained with orbital variability. We cannot reconcile
it with thelower luminosity ofthepresent WDcomponent.
0.1
Probably the atmospheric model had significant differences
from themost recent anddetailed models ofRauch,butwe
0
-10000 -5000 0 5000 10000 alsocannotruleoutthatthefitwascorrect,andtheThom-
velocity in km/s
0.25 son scattering corona may have been very different, or not
present,in theprevious outburst.
0.2
5 THE ORIGIN OF THE EMISSION LINES:
SHOCKS, CONDENSATIONS AND AN
GS2 counts/s 0.15 In thEeVCOhLaVndIrNaGspeVctIrOuLmE2N0%ToMf tEhDeIflUuxMwas in the emis-
R 0.1
sion features, and in the RGS spectra about 30 %, while
theremainingfluxwasduetothecontinuum.Despitesmall
uncertainties in therelative calibration of theChandra and
0.05
XMM-Newton instruments, as we see in Fig. 3, the spec-
traareremarkablysimilarondays18and23.However,two
0 striking changes happened: theabsorption features seem to
-10000 -5000 0 5000 10000 havedisappearedexceptforOIat23.476˚A,whichweiden-
velocity in km/s
tifiedasinterstellarinorigin,andoxygenemissionlinessud-
Figure8.TheNVIILymannαline(above)andNVIresonance
denlyappeared.TheOVIIIline,whichisnotmeasurablein
line(below)measuredwiththeRGS2gratinginFebruary(dots)
andMarch(crosses),respectively,plottedinthevelocityspace. absorption in the WD atmosphere on day 18 but is clearly
detected in emission with XMM-Newton on day23, was al-
mostsymmetricaroundzerovelocitysinceitappeared,con-
730,000 K, N(H)=2.5 × 1021 cm−2, and an absorbed flux sistentlywithnoerosionbyanearbyblue-shiftedabsorption
1.64×10−11 ergcm−2 s−1,correspondingtoanunabsorbed line. This does not imply that the line did not exist at all
flux4.2×10−10 ergcm−2 s−1.Thebolometricluminosityof intheWDatmosphere,becausetheleveloftheWDcontin-
theWD in this model is 7.2 × 1036×(d/12 kpc)2. uum in that energy range is too low to measure absorption
We fitted also the spectra observed later with XMM- (see Rauch et al., 2010)
Newton,althoughtherearenoabsorptionfeaturestobetter Inemission,thefullwidthathalfmaximumofthisline
constrainthetemperaturerange.Thefitreturnsalmost the isapproximately4000kms−1,consistentlywiththevelocity
same temperature for day 18 and 23. In the spectrum of measured in the optical spectra of Mason et al. (2012). As
day 35 the fit yields instead a higher temperature, in the can be seen comparing Fig. 7 with Fig. 8, the emission fea-
range900,000 to1,100,000 K,andis unvariedif weadd the turesin thethirdspectrum appear symmetrical around the
emission component, discussed in Section 5. The WD tem- rest wavelength of the line, and have an almost triangular
peratureatmaximumX-rayluminosityistightlycorrelated shape,indicatingthatlinesareproducedinanopticallythin
withWDmass.Weknowfromthemodelsthattheeffective expandingmedium.Incontrast, theprofiles oftheemission
temperature exceeds 900,000 K only for WD mass around linesintheopticalspectraofnovaeafterthePCygprofiles
1.3 M⊙ (Yaron et al. 2006). havedisappearedusuallyappearflat-topped,indicatingthat
The models yield an absorbed WD luminosity around theexpandingmedium is optically thick.
7×1036,whichat12kpcdistanceisafactorof5-10smaller We measured the emission line fluxes by fitting the
thantheluminositysofarmeasuredinotherWDwitheven strongest lines with Gaussian profiles, and we attempted
lower peak temperature (e.g. Balman et al. 1998, Orio et a fit to a few weak lines with triangles. The photon flux is
al. 2003a, Nelson et al. 2008). For theothernovae,theWD given for each line in the three spectra in Table 1. We esti-
luminosity at peak was several times 1037 erg s−1. We ex- matethatourmeasurementsareaccuratetowithin10%for
plain also this puzzle, liketheone posed bytheapparent P isolatedlinesandfluxaboveapproximately0.4×10−4 pho-
Cygprofiles,ifweassumeUScowedidnotobservetheWD tons cm−2 ks−1 (which were fitted with a Gaussian). The
directly.ThereprocessingfactorfortheWDradiationmust accuracy is probably not more than ≃25% below thisvalue
havebeen of theorder of 10%. (where we fitted a few lines with a triangle, especially for
10 M. Orio et al.
theday35 spectrum). Wedid notattempt tomeasure line with the Table 2 models for days 18 and 23. However, the
fluxes below 0.12 photons cm−2 ks−1, because the uncer- BVAPEC collisional ionization model in XSPEC assumes a
tainty seems very large below this value, but in some cases low density limit and does not predict that the forbidden
we can still identify lines with probably almost an order of lineisquenchedcomparedtotheintercombination line. We
magnitude lower flux (e.g. Mg XI in the last spectrum). seeinFigures2and4thattheNVIλ29.5446forbiddenline
Theemission lines, muchlessstronginthiswavelength iitheNVItripletondays18and23,andtheOVIIλ22.097
range in other novae, in the U Sco spectra are very pro- line at day 23 are less strong than in the models shown in
nouncedandbroad(comparedwiththermalandinstrumen- theFigures.
tal broadening, see e.g. Brinkman et al. 2000, Ness et al.
We attribute this phenomenon to line formation in a
2003). We attribute the line broadening to the expansion
zone of such high density that the collisional de-excitation
velocityofthenovashell,likethebroadeningoftenobserved
rates are close to the radiative decay rates. When the two
intheopticalspectra.TherecurrentnovaeLMC2009(Orio
rates are comparable, the radiative decay fraction from
et al. 2011) and T Pyx (Tofflemire et al. 2012) have been
an excited level are reduced and so the emission becomes
the only other ones with broad emission features. We also
weaker. The electron density ne at which the N VII forbid-
note that the carbon lines are much weaker than those of den line is de-excited is a few times 109 cm−3 (Ness et al.
nitrogen.OnlytheCVILyαlineinemissionat33.90±0.12
2001, Chen et al. 2004). This density is not unheard of for
˚A (rest wavelength 33.7342 ˚A) is clearly detected.
novae shells in the first few weeks after the outburst. ne of
order of 1012 cm−3 was derived by Neff et al. (1978) from
5.1 A comprehensive spectral fit: days 18 and 23 theanalysis of the optical spectra of V1500 Cyg.
Thishighdensityishigherthanthevaluederivedfrom
In Fig. 2, 4 and 5 we plotted in red the fit to the lines and
continuumobtainedwithphysicalmodels.Thesemodelscan the emission measure in the fit in Table 2, ne = a few 107
with the assumptions of the model, if the ejected shell vol-
beregardedasasortoffirstorderapproximationtothecor-
ume is uniformly filled. Since estimates of the ejecta mass
rect model, to explore the relevant physical mechanisms in
thesespectra.Wehavenotreachedaperfectfit,asthevalue of U Sco converge vary from a few 10−6 M⊙ to a few 10−7
of the reduced χ2 in these fits varies from 2.7 (Fig. 2) to M⊙ (seeSchaefer2011,Starrfieldetal.1988),assuminguni-
4 (Fig. 5). Rather than obtaining a statistically significant form filling of the shell, we expect that ne does not exceed
a few times 105 at day 18, and it should be decreasing as
fit, we focused on a good qualitative agreement and on re-
theejectaexpand.Thisdiscrepancyimpliesanemittingvol-
producing the emission line ratios as well as possible. We
ume of the order of only 10−4 the volume of the shell. The
think that, as novae observations are continued in the next
explanation seems to be that line emission occurs in dense
few years, we should be able to gather many more details
condensations. We note that the large clumps that explain
of the nova physics, that will allow to fine tune these mod-
the light curve ≃3 hours period in the Ness et al. (2012)
els of the very complex phenomena occurring in novae WD
model occupy a much smaller fraction of the shell volume
atmospheres and in their ejected shells.
corresponding to only a few 10−9. Clumpiness is not com-
Ondays18and23,therelativelinestrength,combined
pletelyunexpected,andithasoftenbeeninvokedtoexplain
with the lack of a broad radiative recombination RRC fea-
ture of N VI at 22.5 ˚A, which should be clearly present novae optical spectra, starting with the pioneering work of
Gallagher&Anderson(1976).Interestingly,alreadyin1992
in the case of photoionization (e.g. X-ray spectra of plan-
Lloyd et al. discussed how the X-ray emission from a nova
etary nebulae or some AGN, see Fig. 5 of Kinkhabwala et
shell, albeit observed only with a broad band instrument
al.,2002),indicatethatthetransitionsareduetocollisional
(ROSAT),impliescompactclumpsintheejectaasemitting
ionization. We used the XSPEC package, v12.6, adding to
regions.
theatmosphericmodelofthepreviousSectionandacompo-
nent of collisional ionization, namely the BVAPEC model, Inourspectralfits,whoseparametersaregiveninTable
thatallowstoincludevelocitybroadening (1σ value,which 2,theappearanceofoxygenlineson day23dependson the
correspondstothehalfwidthathalfmaximumofthelines). interplay of plasma temperature and emission measure. In-
Table 2gives relevant parameters of thespectral fitsshown creasing theplasma temperaturebyonly ≃30%,or increas-
intheFigures.Weassumedthatheliumintheplasmacom- ing the emitting volume by a factor of a few (assuming for
ponent is enhanced 10 times with respect to solar values. instance that additional mass is being ejected) the oxygen
In order to improve the fit to the emission lines, we need a lines are predicted, but then it becomes difficult to repro-
large overabundance of nitrogen not only in the also in the duce the other lines, correctly modeled only with a plasma
BVAPEC component,at least 10times thesolar value,and temperatureinthe100-120eVrange.ThestrengthoftheO
even up to 70 times like in thefit we show in Fig. 2 for the VIII Lyα feature is much stronger than all the lines of the
Chandraspectrum(thisisthebestfitwithonlytwocompo- O VII He-like triplet, indicating an overlapping component
nents, although we obtained sightly worse, but similar fits at higher temperature. In conclusion, in order to reproduce
with anenhancementbyafactor of10andaslightly differ- both the soft and hard portion of the spectrum, we need
ent set of parameters). We note that the WD unabsorbed atleasttwocomponentsatdifferenttemperatures.Anaddi-
luminosityresultingfromthesefitsisreducedbyupto30% tionalcomponentat180eVreproducestherelativestrength
with respect tothevaluescontained inthe“experiment”in oftheoxygenfeatures.Thecomponentthatexplainstheni-
Section 4.1, in which we fitted only the continuum, cutting trogen lines, at 120-130 eV, fails instead to reproduce the
theprominent emission features. “harder”lines.Itisthuslikelythatonday23thecondensa-
Figures2and4showthatwecanreproducetherelative tionsinwhichemission linesoriginatedwerenotatuniform
strengthofmostlinesbyassumingcollisionalionizationlines temperature.