XMM-Newton and Chandra observations of the ultra-compact binary RX J1914+24

0.26 MB·English

by  Gavin Ramsay

Mon.Not.R.Astron.Soc.000,000–000 (0000) Printed5February2008 (MNLATEXstylefilev1.4)
XMM-Newton Chandra
and observations of the
ultra-compact binary RX J1914+24
Gavin Ramsay1, Mark Cropper1, Pasi Hakala2,3
1Mullard Space Science Laboratory, University College London, Holmbury St. Mary, Dorking, Surrey, RH56NT, UK
2Observatory,University of Helsinki, PO Box 14, FIN-00014 University of Helsinki, Finland
3Tuorla Observatory, V¨ais¨al¨antie 20, 21500 Piikki¨o, Finland
6
0
5February2008
0
2

n
ABSTRACT
a
J
ThenatureoftheX-raysourceRXJ1914+24hasbeenthesubjectofmuchdebate.
5
It shows a prominent period of 569 sec in X-rays and the optical/infra-red: in most

modelsthis hasbeeninterpretedasthe binaryorbitalperiod.We presentouranalysis
1
of new XMM-Newton and Chandra data. We find a longer term trend in the XMM-
v
4 Newton data and power at 556 and 585 sec in 5 sets of data. It is not clear if they
0 are produced as a result of a beat between a longer intrinsic period and the 569
1 sec modulation or if they are due to secular variations. We obtain a good fit to the
1 XMM-Newton spectrum with a low temperature thermal plasma model with an edge
0 at 0.83keV. This model implies an unabsorbed bolometric X-ray luminosity of 1×
6 1033 ergs s−1(for a distance of 1kpc) - this is 2 orders of magnitude lower than our
0 previous estimate (derived using a different model). If the distance is much less, as
h/ the absorptionderivedfrom the X-rayfits suggest,then it is even lowerat ∼3×1031
p ergs s−1.
-
o Key words: Stars:individual:– RXJ1914+24– Stars:binaries– Stars:cataclysmic
r variables – X-rays: stars
t
s
a
:
v
i
X 1 INTRODUCTION the UI model (Marsh & Nelemans 2005). This conclusion
is based on the X-ray luminosity determined using ROSAT
r
a The natureof theX-ray source RX J1914+24 has been the (Cropperetal1998)andXMM-Newtonobservations(Ram-
subject of much debate in recent years. Only one modula- sayetal2005a). TheROSATdatawerewellfittedusingan
tion period (569 sec) has been seen in both its X-ray and absorbed blackbody model while the higher signal to noise
opticallightcurves(Ramsayetal2000).Forthis,andother data of XMM-Newton found a poor fit using this model.
reasons, the period of 569 sec has been taken to be the bi- Ramsayetal(2005a)foundthebestfittothedatacouldbe
nary orbital period, which would make it the second most achievedusinganabsorbedblackbodyplusGaussianmodel,
compact binary system known. althoughtheydidnotexpectthatthismodelwasphysically
realistic.
Themodelswhichhavebeenputforwardtoexplainthe
observational characteristics of RX J1914+24 fall into two
broad categories: accretion and non-accretion driven mod- SincethosefirstXMM-Newtonobservationsweremade
els. The fact that the optical spectrum of RX J1914+24 in Nov 2003, a further two sets of observations made in
(also known as V407 Vul) shows no emission lines, puts a Sept/Oct 2004 have become publically available. These ob-
question mark over the accretion driven models. The non- servationswerelongerthantheinitialXMM-Newtonobser-
accreting model is the unipolar-inductor (UI) model of Wu vationsandtheparticle backgroundwaslower. Inaddition,
etal(2002).Whilethismodelhasbeenabletoexplainmany afurthersetofX-rayobservationsofRXJ1914+24madeus-
oftheobservedcharacteristicsofthissource,itappearsthat ingChandrahasbecomepublicallyavailable.Wehavethere-
its X-ray luminosity implies an induction torque which is fore extracted these data from the archive and use them to
much greater than the amount generated by gravitational make a more detailed investigation of its X-ray spectrum,
radiation.Thereissomeevidencethattheobservedspin-up to determine if the 569 sec is continuing to spin-up at the
rate of RX J1914+24 (Strohmayer 2002, 2004b, Ramsay et same rate and tosearch for evidencefor periods other than
al 2005a) is therefore much smaller than that predicted by the569 sec period.
(cid:13)c 0000RAS
2
XMM StartDate Duration MeanEPIC
Orbit (ksec) pn(Ct/s)
0880 2004-09-28 17.0 1.08
0882 2004-10-02 19.0 1.16
Chandra StartDate Duration MeanChandra
ObsId (ksec) Ct/s
4519 2004-04-05 15.4 0.44
4520 2004-06-13 14.7 0.37
4521 2004-09-05 15.2 0.34
Table 1. The log of the observations of RX J1914+24 reported
here. Each observation was continuous with no data gaps. Top
panel: The XMM-Newton observations of RX J1914+24 made
in 2004. The earlier observations are detailed in Ramsay et al
(2005a). Bottom panel: The Chandra made in 2004. The earlier Figure 1. The phase of the sharp rise to maximum X-ray flux
observations aredetailedinStrohmayer (2004b). determinedfromROSAT,ChandraandXMM-Newtondata(see
Ramsay et al 2005a for details of these observations). The more
recentpointreferstothenewXMM-Newtondatapresentedhere,
with the preceding point refers to Chandra data taken in April
2 OBSERVATIONS AND DATA REDUCTION
2004.
2.1 XMM-Newton observations
RX J1914+24 was observed twice using XMM-Newton in P˙ =−3.31±0.09×10−12 s/s(theerrorbeingdeterminedus-
Nov 2003. The details of these observations has been re- ingabootstrap method),whichisconsistentwith thevalue
ported by Ramsay et al (2005a). It was observed again late we determined previously.
in2004andweshowinTable1thestarttime,durationand
mean X-ray count rate of the second epoch XMM-Newton
observations.Inthispaperwediscussthedatacollected us-
4 SEARCHING FOR PERIODS OTHER THAN
ingtheEPICX-raydetectors(pnandMOS)whicharesen-
THE 569 SEC PERIOD
sitive over the energy band 0.15-10keV. We extracted the
data from the public archive and processed the data using Strohmayer (2004b) detected sideband structure in the
theXMM-NewtonSASv6.5inasimilarmannertothatde- power spectrum of RX J1914+24 in observations made us-
scribed in Ramsay et al (2005a). Data were barycentrically ingChandrabutconcludedthatthiswaslikelytohavebeen
corrected and in unitsof Terrestrial Time (TT). caused by the dithering pattern of those observations. On
the other hand, longer term variations have been seen in
X-raydata (Ramsay et al 2000, Strohmayer2004b).
2.2 Chandra observations We have searched for evidence of periods other than
the 569 sec period in these new X-ray data. We performed
Observations of RX J1914+24 made using Chandra on 5
a Discrete Fourier Transform (DFT) on the combined data
occasions were reported byStrohmayer(2004b). Sincethen takeninXMM-Newtonorbits0718and0721andalsoorbits
another3observationshaveenteredthearchive.Allofthese 0880 and 0882 as well as each of theChandra observations.
observationsweremadeincontinuousclockingmode:inthis
Asexpected,the569secperiodanditsharmonicswereseen
mode a one-dimensional image of the sky is produced. We
stronglyinalltheamplitudespectra.Topre-whitenthedata
haveextractedalloftheavailabledatafromthearchiveand
we used two methods: we pre-whitened in the usual man-
corrected and analysed them in a similar manner to that
ner using a sinusoid at predicted period at the observation
carried out byStrohmayer (2004b).
epoch (based on the initial period of Strohmayer (2004b)
and the spin-up rate reported here). We also pre-whitened
the data by subtracting the mean X-ray pulse profile ap-
propriatetoeachindividualobservation.Thelatteris more
3 DETERMINING THE RATE OF SPIN-UP
appropriate since the X-ray pulse profile is non-sinusoidal.
Ramsay et al (2005a) folded all available X-ray data As expected most of the peaks are removed after this pro-
(ROSAT, ASCA, Chandra and XMM-Newton) on the best cedure. However, in the second epoch set of XMM-Newton
fit initial period of Strohmayer (2004b) (ie without the pe- data(0880 and0882), side-bandstructureisclearly present
riod derivative) and found that the sharp rise of the X-ray at 584.1 sec and 552.8 sec in the pre-whitened spectrum
flux arrives earlier in phase over time. A best fit to the (Figure 2). They are located in thewings of the main peak
data showed that RX J1914+24 is spinning up at a rate intheoriginalamplitudespectrumandhaveanamplitudeof
of 3.2×10−12 s/s, which was similar to that reported by ∼1/10 of the main peak. In the earlier XMM-Newton data
Strohmayer(2002,2004b).Wedidthesameanalysisasdone (0718 and 0721) there is only weak evidence of these peri-
in Ramsay et al (2005a) and show the phase of the rise in ods. In the Chandra dataset of 2004 June 13 we also find
X-rayfluxfortheChandradatatakeninApril2004andthe strongevidenceforperiodsverysimilartothoseseeninthe
new XMM-Newton data in Figure 1. We find a best fit of XMM-Newton data (Figure 2).In two others, (2004 Jan 31
(cid:13)c 0000RAS,MNRAS000,000–000
3

584.1" 552.8" XMM-Newton
2004-09-28+
e 2004-10-02
d
u
plit
m
A

584.1" 55 2.8" Chandra
2004-01-31

e
d
u
plit
m
A

Chandra
584.1" 552.8" Figure 3. The XMM-Newton EPIC pn spectra from the data
2004-06-13 takenin2004.Thesolidlineisthebestfittothemodelconsisting
de ofanabsorbedthermalplasmamodelwithanedge.
u
plit
m
A
5 SPECTRA

584.1" 552.8" Chandra Ramsay et al (2005a) showed the X-ray spectrum of RX
2004-09-05
J1914+24 was poorly modelled using an absorbed black-
de body model. Their best fit was obtained using an absorbed
u
plit blackbody plus a broad Gaussian line near 0.6keV. This fit
m
A was still poor (χ2ν=1.63) and the model was unlikely to be
physically realistic.
WiththebetterqualitydataaffordedbythenewXMM-

0 0.002 0.004 Newton data we have re-visited the X-ray spectrum of RX
Freq (1/sec) J1914+24. WefittedtheEPIC-pndatafromXMM-Newton
orbits0880and0882separately.Giventheunphysicalnature
Figure 2. The pre-whitened amplitude spectra of 5 sets of X- ofthepreviousfitsandworkingfromthesatisfactory fitsto
raydata. Theyhavebeenpre-whitenedbysubtractingthemean our X-ray observations of other compact binaries (the AM
X-rayprofileappropriatetoeachepoch. CVnstars,Ramsayetal2005b) wetried fittingthespectra
with an absorbed thermal plasma model with variable ele-
mentalabundance(vmekalinXSPEC).Thismodelstillgave
apoorfittothedata(χ2ν=4.6 forthe0880 data).However,
and 2004 Sept 5) there are peaks in the amplitude spectra when we added an edge to the model we obtained a good
which correspond to these periods, although they are only fit to the 0880 data (χ2ν=0.97, 122 dof; Figure 3). The fit
marginally significant compared tootherpeaks in thespec- to the 0882 data using the same model was slightly poorer
tra. There is no evidence for these periods in the longest (χ2ν=1.21, 141 dof). Elements which were required in the
Chandra dataset which lasted 35 ksec. We also determined fit include N, O, Ne, Ca, Fe for the 0880 data, with some
howtheX-raypulseprofilechanged overtheindividualob- evidence for Na in the 0882 data. The edge is located at
servation duration.We findthat themean profile brightens 0.83keVwhichisconsistentwithitbeingduetoOVIII.The
over the duration of both the 0880 and the 0882 observa- absorption derived from the fits is ∼ 1.2×1021 cm−2. We
tions. note we cannot obtain a good fit with this model when we
We believe that 553 and 584 sec periods could result fixtheabsorptionat8.3×1021 cm−2,thelargervaluedeter-
when a modulation signal (ie the 569 sec period) is mixed minedfrom opticalinterstellarabsorption features(Steeghs
withanotherlongertermvariation(eitheraperiodicornon- et al 2006).
periodic modulation). Ifit was a periodic modulation, then The observed flux in the 0.1-10keV energy band
itsperiodwouldbethedifferencebetweenthe569secperiod (1.30×10−12 ergs s−1 cm−2for the orbit 0880 data and
and the 553 and 584 sec periods: just over 6 hrs. Alterna- 1.48×10−12 ergs s−1 cm−2for the orbit 0882 data) is very
tively if a modulation signal is mixed with either a 553 sec similartothatfoundusingtheblackbodyplusGaussianline
or584secperiodthenthiswouldgiverisetothelongerterm ofRamsayetal(2005a).However,theunabsorbedbolomet-
trend in the data. The question remains if this longer term ricfluxinthefitpresentedhere(1.14×10−11 ergss−1 cm−2;
trend is periodic or due to the secular variability that have 0880 and 1.24×10−11 ergs s−1 cm−2; 0882) is substantially
alreadybeenobservedinthissystem:wediscussthisfurther reducedcomparedtotheearlierresults(3.2×10−10 ergss−1
in §6.1. cm−2) where a blackbody spectrum was used.
(cid:13)c 0000RAS,MNRAS000,000–000
4
6 DISCUSSION
33..00
6.1 The periods in RX J1914+24
Chandra
Until now there has been no convincing evidence for any 22..55
XMM-Newton
otherperiodsinRXJ1914+24apartfromthe569secperiod.
In the case of 5 X-ray observations we have found evidence
for power at 552 and 584 sec. This changes our perception V) 22..00 WXP
ke XP
ofRXJ1914+24. Thedataareclear-theX-rayfluxclearly 0
increasesintheXMM-Newtondatatakeninbothorbit0880 4-1. 11..55 WXPXP
0. WXP
andinorbit 0882. Thiscouldbeconsideredtobetheresult x (
u XP
of an underlying ∼6 hour period, or it could be the result Fl 11..00
WXP
ofsecularvariabilityknownalreadyinthissystematlonger
timescales (Ramsay et al 2000, Strohmayer 2004b). Both
00..55
could producetheobserved amplitudespectrum.
To explore this question, we have obtained the mean
00..00
X-ray flux in each of the Chandra and XMM-Newton ob- 22660000 22770000 22880000 22990000 33000000 33110000 33220000 33330000
servations. Since the Chandra ACIS-I detector has a low HJD - 2450000
effective area at energies less than ∼0.4keV, we determined Figure4.Theobservedfluxinthe0.4-1keVenergyband(unitsof
the observed mean flux in the 0.4-1.0keV energy band. We 10−12ergss−1cm−2)determinedinChandraandXMM-Newton
showhowthefluxinthisbandvariesovertimeinFigure4. observations. ‘XP’ denotes the presence of periods at 552 and
Wealsoshowthoseepochsatwhichwedetectedthe552sec 583 sec: in the case of the most recent Chandra observation the
and 584 sec periods which could be signatures of a longer significanceismarginal-‘WXP’weakX-rayperiods.
periodtrendintheobservation.Althoughnotconclusive,it
appears that at the higher flux levels there is no (or weak)
evidence of these other periods, so that shorter (<1 day)
timescale variations may take place only when the system the new model is an optically thin plasma, a correction ef-
is emitting at intermediate or low levels. The detection of fect for projection is now no longer applicable. Moreover,
power at 552 and 584 sec does not appear to be related absorption derived from the X-ray fits is ∼15% of that de-
to theobservation length since thelongest observation (the rived from the optical data of Steeghs et al (2006) based
Chandra observation of Feb 2003 which lasted 35 ksec) did on a G9 spectral type. If the G star is not the secondary,
not show obviouspower at theseperiods. thenwecan reducethedistancebyasimilar factor and the
Atthisstage, wecannot distinguish betweentheX-ray luminosityistherefore 3×1031 ergss−1.Inthiscase, theX-
longterm behaviourbeingduetorandom changesin either ray luminosity of RX J1914+24 is more than ∼3 orders of
the accretion rate (in the case of the accreting models) or magnitudeless than that found by Ramsay et al (2005a).
theelectricalcurrentdissipation(inthecaseoftheunipolar- In the unipolar-inductor model of Wu et al (2002) the
inductormodel).Weurgethesearchforradialvelocityvari- amount of power which is emitted is determined by vari-
ations on a timescale of less than 1 day,or dedicated X-ray ous factors, with the degree of asynchronisity being one of
and/oropticalobservationsofthissystemtosearchforvari- the key determinants (along with component masses and
ations on this period. magnetic field strength). However, to produce theobserved
IfthelongertermvariationintheX-rayfluxisperiodic, luminosity,theasynchronisitywouldhavetobe0.001orless
thenweshowedearlier thatitsperiod wouldapproximately (Wuetal2002).ThisreducestheobjectionstotheUImodel
be 6 hrs. Smith & Dhillon (1998) show that for a binary onenergeticgroundsraisedbyMarsh&Nelemans(2005)to
systemwiththisorbitalperiodthenitsspectraltypewould a significant degree.
beexpectedtobeamidK-type,althoughtherangeisupto5 The abundances are clearly non-solar, which is consis-
sub-classes.WenotewithinterestthediscoveryofSteeghset tent for an evolved secondary as is seen in the fits to spec-
al(2006)offeaturesintheopticalspectrumofRXJ1914+24 tra of AM CVn systems (eg Marsh, Horne & Rosen 1991,
whichappeartoresembleaG9/K0spectraltype.Therefore, Strohmayer 2004a, Ramsay et al 2005b). If this model re-
an orbital period of 6 hrs would be marginally consistent flects the properties of the material leaving the secondary
with this. star then it rules out the G9/K0 star seen in the optical
spectrum being thesecondary star.
6.2 The X-ray spectrum of RX J1914+24
WehaveshownthatthederivedluminosityofRXJ1914+24
is highly model dependent. Further, the distance to RX
J1914+24 is still rather uncertain, but Steeghs et al (2006)
7 ACKNOWLEDGMENTS
suggest a distance of 1kpc based on its optical spectrum.
Foradistanceof1kpcthemeanbolometricluminosityfrom This is work based on observations obtained with XMM-
the XMM-Newton orbits 0880 and 0882 data is therefore Newton, an ESA science mission with instruments and
1.4×1033ergss−1-anorderofmagnitudelowerthanthepre- contributions directly funded by ESA Member States and
viousluminosityof3.8×1034 ergss−1.Ramsayetal(2005a) the USA (NASA). The XMM-Newton data were extracted
wentfurtherandappliedacorrection toaccountforprojec- from the archive at Vilspa and the Chandra data from the
tionaffectsgivingaluminosityof∼1×1035 ergss−1.Since HEASARCarchiveat Goddard Space Flight Center.
(cid:13)c 0000RAS,MNRAS000,000–000
5
REFERENCES
Cropper, M., Harrop-Allin, M. K., Mason, K. O., Mittaz, J. P.
D.,Potter,S.B.,Ramsay,G.,1998,MNRAS,293,L57
Marsh,T.R.,Nelemans,G.,2005,MNRAS,363,581
Ramsay,G.,Cropper,M.,Wu,K.,Mason,K.O.,Hakala,P.,2000,
MNRAS,311,75.
Ramsay, G., Hakala, P., Wu, K., Cropper, M., Mason, K. O.,
Co´rdova,F.A.,Priedhorsky,W.,2005a, MNRAS,357,49
Ramsay, G., Hakala, P., Marsh, T., Nelemans, G., Steeghs, D.,
Cropper,M.,2005b,A&A,440,675
Smith,D.A.,Dhillon,V.S.,1998, MNRAS,301,767
Steeghs, D.,etal,inprep
Strohmayer,T.E.,2002,ApJ,581,577
Strohmayer,T.E.,2004a,ApJ,608,L53
Strohmayer,T.E.,2004b,ApJ,610,416
Wu,K.,Cropper,M.,Ramsay,G.,Sekiguchi,K.,2002,MNRAS,
331,221
(cid:13)c 0000RAS,MNRAS000,000–000