SXP7.92: A Recently Rediscovered Be/X-ray Binary in the Small Magellanic Cloud, Viewed Edge On

1.4 MB·

by  E. S. Bartlett

MNRAS000,1–12(2017) Preprint10January2017 CompiledusingMNRASLATEXstylefilev3.0
SXP 7.92: A Recently Rediscovered Be/X-ray Binary in
the Small Magellanic Cloud, Viewed Edge On
E. S. Bartlett,1,2(cid:63) M. J. Coe3, G. L. Israel4, J. S. Clark5, P. Esposito6, V. D’Elia4,7
and A. Udalski8
1ESO - European Southern Observatory, Alonso de Co´rdova 3107, Vitacura, Casilla 19001, Santiago de Chile, Chile
2Astrophysics, Cosmology and Gravity Centre, Department of Astronomy, University of Cape Town, Rondebosch 7701, South Africa
3School of Physics and Astronomy, University of Southampton, Highfield, Southampton, SO17 1BJ, United Kingdom
7 4INAF – Osservatorio Astronomico di Roma, via Frascati 33, I-00040 Monte Porzio Catone (Roma), Italy
1 5Department of Physical Science, The Open University, Walton Hall, Milton Keynes, MK7 6AA, United Kingdom
0 6Anton Pannekoek Institute for Astronomy, University of Amsterdam, Postbus 94249, NL-1090-GE Amsterdam, The Netherlands
2 7ASI Science Data Center (ASDC), via del Politecnico snc, I-00133 Roma, Italy
8Warsaw University Observatory, Aleje Ujazdowskie 4, 00-478 Warszawa, Poland
n
a
J
Accepted2017January05.Received2017January05;inoriginalform2016July26
6

]
E ABSTRACT
H
We present a detailed optical and X-ray study of the 2013 outburst of the Small
.
h Magellanic Cloud Be/X-ray binary SXP7.92, as well as an overview of the last 18
p years of observations from OGLE, RXTE, Chandra and XMM-Newton. We revise the
- position of this source to RA(J2000)=00:57:58.4, Dec(J2000)=-72:22:29.5 with a 1σ
o
uncertainty of 1.5(cid:48)(cid:48), correcting the previously reported position by Coe et al. (2009)
r
t by more than 20 arcminutes. We identify and spectrally classify the correct counter-
s
a part as a B1Ve star. The optical spectrum is distinguished by an uncharacteristically
[ deep narrow Balmer series, with the Hα line in particular having a distinctive shell

profile, i.e. a deep absorption core embedded in an emission line. We interpret this as
1
evidencethatweareviewingthesystemedgeonandareseeingselfobscurationofthe
v
circumstellar disc. We derive an optical period for the system of 40.0±0.3 days, which
9
2 we interpret as the orbital period, and present several mechanisms to describe the
7 X-ray/Optical behaviour in the recent outburst, in particular the“flares”and“dips”
1 seen in the optical light curve, including a transient accretion disc and an elongated
0 precessing disc.
.
1
Key words: stars: emission-line, Be – X-rays: binaries – Magellanic Clouds
0
7
1
:
v
1 INTRODUCTION that, since the launch of INTEGRAL a new type of sgXRB
i
X has been identified; the Supergiant Fast X-ray Transients
High-Mass X-ray Binaries (HMXBs) are stellar systems in
r (SFXTs,Negueruelaetal.2006)whosenumbersareincreas-
a which a compact object, usually a neutron star (NS) with ing rapidly. With binarity increasingly being recognised as
a strong magnetic field, orbits a massive star of spectral
playingamajorroleintheevolutionofmassivestars(Sana
type OB. HMXBs broadly fall into two categories: Be/X-
etal.2012),HMXBsrepresentapivotalpointintheirevolu-
ray binaries (BeXRBs) in which the primary is a rapidly
tionary track: Once one star has gone supernova but whilst
rotatingearlytypestarwithluminosityclassIII-V,andsu-
the other is still fusing hydrogen, either in its core or enve-
pergiant X-ray Binaries (sgXRBs), in which the primary is
lope.This,inturn,meansthatHMXBscouldprovidevalu-
a post main sequence star. Material is accreted onto the
able insight into many astrophysical phenomena, including
compact object, either via Roche-Lobe overflow in the case
corecollapsesupernovae,gamma-raybursts,magnetarsand
of the sgXRBs, or directly from the circumstellar disc that
gravitational waves.
surrounds the Be star in the BeXRBs (and leads to their
The‘e’designationstandsforemission:Bestarshave,at
‘e’ designation). BeXRBs are currently the most numer-
somepointintheirlives,displayedoneormoreoftheBalmer
ous subclass of HMXB (Liu et al. 2006), though we note
series in emission in their optical spectra. These lines origi-
nate from the equatorial outflow, or decretion disc, around
(cid:63) E-mail:[email protected](ESB) thestar.Thisdiscisfedinpartbythestar’srapidrotation
c 2017TheAuthors
(cid:13)
2 E. S. Bartlett et al.
(Townsend et al. 2004), though the creation mechanism is 2 OBSERVATIONS AND DATA REDUCTION
still unclear (possibly linked to the transient nature of the
TheregionoftheSMCoccupiedbythepositionreportedby
emission lines). The accreted material is funnelled on the
Israel etal.(2013)hasbeenstudiedextensivelyinboththe
magnetic poles of the NSs. These NSs are rotating, lead-
opticalandinX-raysforthelast∼18years.Hereweoutline
ing these systems to be coherently variable X-ray sources,
the data presented in this paper in approximately chrono-
i.e. pulsators. The orbits of the NSs in these systems are
logical order, though we note that due to the serendipitous
highly eccentric, and are loosely correlated with the spin
natureofthedatacollection,thisisnotstrictlyadheredto.
period (Corbet 1984, 1986). There are two types of X-ray
All X-ray detections and non-detections for XMM-Newton,
outbursttraditionallyassociatedwithBeXRBs,TypeIout-
Chandra and Swift refer to the position reported by Israel
bursts, which occur around the time of periastron passage,
last days-weeks and reach Lx∼1036−37 erg s−1and Type II et al. (2013)
“Giant”outbursts which last longer (up to months) reach
higher luminosities (Lx>1037 erg s−1) and have no correla-
tionwithorbitalphase(Stellaetal.1986).Morerecently,it 2.1 X-ray
has been recognised that BeXRB behaviour is less dichoto-
2.1.1 RXTE
mouswithmanyoutburstsnoteasilyclassifiableinthiscur-
rent framework (Kretschmar et al. 2012). Additionally, a The RXTE observations used in this paper come from the
subset of BeXRBs, displaying persistent low level X-ray lu- regular monitoring of the SMC carried out over the period
minosity of around 1035−36 erg s−1(e.g. Bartlett et al. 2013) 1997-2012.TheSMCwasobservedonceortwiceaweekand
with no evidence of outbursts, have been discovered. For a the X-ray activity of the system determined from timing
fullreviewonBeXRBsandtheirpropertiesseeReig(2011). analysis.SeeLaycocketal.(2005)andGalacheetal.(2008)
TheSmallMagellanicCloud(SMC)playshosttoalarge fordetailedreportsonthistechnique.Asummaryofmostof
numberofHMXBsallofwhich,withtheexceptionofSMC the years of observing SXP7.92 was presented in Coe et al.
X-1,areallBeXRBs(Coeetal.2015).Thereasonforthisis (2009),herewereporthereonobservationswhichextendthe
not fully understood but is thought to be in part due to its published record on SXP7.92 by three further years - but
recent star formation history and its low metalicity (Dray withnofurtherRXTE detectionsinthattime.Asdiscussed
2006). Its small angular extent and its location in the sky, inLaycocketal.andGalacheetal.,thequalityofanysingle
outoftheplaneoftheGalaxyandsoreasonablyunaffected observation depends upon the significance of the detected
by attenuation by the Galactic interstellar medium, makes periodcombinedwiththecollimatorresponsetothesource.
theSMCanidealregionforpopulationstudiesofBeXRBs. We remove any period detections with a significance less
than 99%, and a collimator response less than 0.2.
1.1 SXP7.92
ThesubjectofthispaperisSXP7.92,a7.92spulsarinthe
2.1.2 Chandra
SMC. First discovered by Corbet et al. (2008) with RXTE,
it was subject to further scrutiny by Coe et al. (2009). The Chandra has observed the position of SXP7.92 four times,
lackofimagingcapabilitiesofRXTE meantitwasnotpos- once in 2002 and three times in 2013, see Table A1. The
sible to localise SXP7.92 any further than approximately observationsweretakenwiththeACISinstrumentinimag-
onequarteroftheSMC.However,theydeterminedthatthe ing(TimedExposure)mode,withreadouttimesofapprox-
likely optical counterpart to the pulsar was the early type imately 3.2 s. The data were reprocessed and analysed fol-
star AzV285, based on a single Swift X-ray source identi- lowingstandardprocedureswiththeciaosoftwarepackage
fied in a total of nine 1 ks Target of Opportunity (ToO) version4.8andthecalibrationdatabasecaldb4.7.Source
observations,whichcoveredtheRXTE FWHMzone.These spectra and event lists were extracted from regions of ∼6–
observations were made approximately one month after the 8 arcsec size (depending on the off-axis distance), while for
initial RXTE discovery. Whilst the lack of pulsations de- the background we used annuli with internal/external radii
tectedmadeitimpossibletodefinitivelysaythatthissource of 15/20 arcsec. The spectra, the redistribution matrices,
and the RXTE source were one and the same, Coe et al. and the response files were created with specextract. For
2009madeacompellingprobabilityargumentforthisbased thetiminganalysis,weappliedtheSolarsystembarycentre
onitsproximitytoAzV285,whichshowedclearevidencefor correction to the photon arrival times using axbary. The
binary modulation in the expected range (Corbet 1984). detection algorithm used is based on the one described by
Using Chandra, Israel et al. (2013) reported the de- Israel & Stella (1996).
tection of a 7.9 s pulsar in data from September of that SXP7.92 was not detected in the 2002 observation
year. Although the source was not consistent in position 2948. We set a 3σ upper limit on the source count rate
with AzV285, we regard this X-ray source as identical to of ∼2.2×10−3 counts s−1 (0.3–8 keV). The best fit model
SXP 7.92, based on the agreement in the pulse period. The parameters from Section 3.3 were used along with Chandra
same source was also serendipitously observed in the IN- redistribution matrix and response files, to translate this
TEGRAL/XMM-Newton monitoringcampaignofSXP5.05 intothe0.2–12.0keVfluxlimitincludedinTableA1,using
(Coe et al. 2015). Here we discuss this recent X-ray data in xspec’s fakeit task.
detail along with previous X-ray observations and the long The discovery of 7.9-s period pulsations from SXP7.92
term optical light curve of the newly determined counter- is part of the results obtained from a larger project, the
part, identified as a variable V=16.0 mag object by Israel Chandra ACIS Timing Survey at Brera And Rome astro-
et al. (2013). nomicalobservatories(CATS@BAR;Israeletal.2016).This
MNRAS000,1–12(2017)
SXP7.92 3
is a Fourier-transform-based systematic search for new pul- usingthebackscaltaskandresponsematrixfileswerecre-
satingsourcesintheChandraAdvancedCCDImagingSpec- ated for each source using the task rmfgen and arfgen.
trometerpublicarchiveandcarriedoutinanautomaticfash- Thesourcewasnotdetectedinobservation0700580101.
ion. Again, redistribution and response files were created using
the task rmfgen and arfgen and an upper limit was de-
rived by generating a sensitivity map using the esensmap
2.1.3 XMM-Newton task. As previously, xspec’s fakeit task was used to simu-
late the best fit spectrum to SXP7.92 with this count rate
XMM-Newton has observed the position of SXP7.92 on 12 to obtain an estimate on the upper limit of the source flux
occasions since 2000, including the three observations in- at this time.
cluded in Coe et al. (2015). Details of the observations are
found in Table A1. The pre-2013 (MJD ∼ 56290), 3σ up-
perlimitswereobtainedusingtheFLIXupperlimitserver1, 2.1.4 Swift
whichmakesuseofthealgorithmdescribedbyCarreraetal.
SXP7.92hasbeenobservedwithXRTonboardSwift on129
(2007).Theappropriatecannedredistributionmatricesand
occasions since 2006 November, of which 118 observations
response files were then used, along with the best fit model
were in photon counting (pc) mode and 11 were in win-
parameters in Section 3.3, to correct these flux values from
theassumedinputspectrum(NH=3×1020cm−2andΓ=1.7) dow timing (wt) mode. The majority of these observations
to that of SXP7.92 using xspec’s fakeit. The flux limits were targeted at SXP5.05 and so are concentrated around
the 2013 outburst. We used the Swift XRT data products
given in Table A1 are the most sensitive obtained from any
generator2 (Evans et al. 2007, 2009) to obtain background
oftheEPICinstrumentsonboardXMM-Newton duringthe
subtracted count rates or upper limits of SXP7.92 in these
observation.
observations. Of the 11 XRT-wt observations, 9 were de-
The three most recent XMM-Newton observations of
clared“unsafe”as no centroid could be determined. These
SXP7.92, discussed extensively in this paper, are from the
observationswerenotincludedinouranalysis.TheXRT-pc
2013targetedmonitoringcampaignofSXP5.05.Fordetails
count rates for each observation were converted to source
ofthegeneraldatareductionstepsoftheseobservationssee
fluxvaluesusingtheSwift cannedresponsematrixandbest
Coe et al. (2015). SXP7.92 is located ∼5.5 arcminutes to
fit model to the source spectrum with xspec’s fakeit. The
the north east of SXP5.05 and fell into the field of view of
wt observations were not treated in this manner, as the 1
all observations, though we note that, due to the observing
dimensional observations are dominated by SXP5.05.
strategy employed, SXP7.92 was not visible to all three in-
TheXRT-wtandpcobservationsinwhichSXP7.92was
strument in every observation: In observations 0700580101
detected were downloaded and processed with xrtpipeline
and0700580601(thefirstandfinalobservations),theEPIC-
with the position of SXP7.92 specified. Event files were ex-
pndetectorwasinSmallWindowModetoavoidpileup,the
tracted with a 20 arcsec circular region and a barycentric
reducedfieldofviewofthismodemeantthatSXP7.92was
correction was applied using the mission independent tool
not visible to this instrument. In observation 0700580401
barycorr. Source lightcurves were created and searched for
(the second observation) the EPIC-pn detector was in full
periods, however none were detected due to a combination
framemode,andsoSXP7.92wasvisible,butthesourcefell
of low count rates (∼0.06 counts s−1 at peak for the XRT-
on a chip gap in the two EPIC MOS detectors.
pc observations), short exposures times (<5 ks with many
For observations where the source was detected (ob-
<1 ks) and, in the case of the XRT-pc data, low time reso-
servations 0700580401 and 0700580601), light curves and
lution (2.506 s).
spectra were extracted from the relevant instruments. The
extraction regions were defined by the xmmsas task ere- A combined spectrum of all the positive detections,
gionanalyse, which calculates the optimum extraction re- along with the associated redistribution matrix, response
and background files was created with the XRT data prod-
gion of a source by maximising the signal to noise ratio.
ucts generator and is included in Section 3.3.
This resulted in extraction radii ranging from 35 to 45 arc-
seconds. For the EPIC-pn light curves of SXP7.92 “Sin-
gle” and “double” pixel events were selected with quality
2.2 Optical
flag (FLAG=0), i.e. all bad pixels and columns were disre-
garded.FortheEPIC-MOSlightcurves,“single”to“quadru- 2.2.1 Photometry
ple”(PATTERN<=12)pixeleventswereselectedwithqual-
Data from the Optical Gravitational Lensing Experiment
ityflag#XMMEA EM.Photonarrivaltimeswereconverted
(OGLE) Phase III and IV (Udalski 2003; Udalski et al.
to barycentric dynamical time, centred at the solar system
barycenter,usingtheSAStaskbarycen. Backgroundlight 2015) are used to investigate the long-term behaviour of
the identified optical counterpart of SXP7.92. OGLE has
curveswereextractedfroma90arcsecondregiononaneigh-
been regularly monitoring this object since 2000 with the
bouring CCD free of sources. The background subtraction
was performed using the epiclccorr task which also cor- 1.3 m Warsaw telescope at Las Campanas Observatory,
Chile, equipped with two generations of CCD camera: an
rectsforbadpixels,vignettingandquantumefficiency.Spec-
eight chip 64 Mpixel mosaic (OGLE-III) and a 32-chip 256
trawereextractedusingthesameregionsasthelightcurves.
Mpixel mosaic (OGLE-IV). Observations were collected in
The area of source and background regions were calculated
the standard I-band.
1 http://www.ledas.ac.uk/flix/flix_dr5.html 2 http://www.swift.ac.uk/user_objects/
MNRAS000,1–12(2017)
4 E. S. Bartlett et al.
51000 52000 53000 54000 55000 56000 57000
15.5 10 10
−
15.6
10 11
−
15.7
10 12
15.8 −
15.9 10 13
−
16.0
10 14
−
1998-01 2001-01 2004-01 2007-01 2010-01 2013-01
15.5 10 10
−
15.6 X
de 10−11-ray
u15.7
t fl
gni 10 12ux
a15.8 − e
M r
g
d s
an15.9 10−13cm
b −
I-16.0 s2
10−14−1
56200 56400 56600 56800 57000
15.6
15.7 10 11
−
15.8
15.9
10 12
−
16.0
16.1
10 13
−
56600 56620 56640 56660 56680 56700 56720
Date(MJD)
Figure 1.LightcurveoftheopticalcounterpartofSXP7.92.Thegreenarrowsmarkthe3σX-rayupperlimitsfromChandra XMM-
Newton and Swift, the red points mark the X-ray detections. The epochs of the optical spectra are marked by the blue lines. The top
panel shows the entire ∼18 year light curve, and the lower panels subsequent small subsets of the most recent optical outburst. The
dottedlinesrepresenttheorbitalperiodproposedinSection3.5.AnX-rayfluxof2.3×10−12 ergcm−2 s−1 correspondstoaluminosityof
1036 ergs−1 atthedistanceoftheSMC(62.1kpc;Graczyketal.2014)
2.2.2 Spectroscopy usingthe standardIRAF packages. A He-Ar lamp spectrum
was acquired during the same night as the target and used
Optical spectra of the optical counterpart of SXP7.92 were for wavelength calibration.
taken on the nights of 2013 October 18th and 2013 Decem- The later observation was taken with Robert Stobie
ber 1st. The first observations were taken with the ESO Spectrograph (RSS, Burgh et al. 2003) on the Southern
Faint Object Spectrograph (EFOSC2, Buzzoni et al. 1984) AfricanLargeTelescope(SALT,Buckleyetal.2006)atthe
mountedattheNasmythBfocusofthe3.6mNewTechnol- South African Astronomical Observatory. Grating PG2300
ogyTelescope(NTT),LaSilla,Chile.Theinstrumentwasin wasusedwithagratingangleof30.5degrees,aslitwidthof
longslitmodewithaslitwidthof1.0arcsecandinstrument 1.5 arcsec and 2×4 (spectral × spatial) binning. The setup
binning 2×2. Grism #20 was used to obtain a spectrum described results in a dispersion of ∼0.25 ˚A pixel−1 at the
around the Hα region (6563 ˚A). Grism #20 is a Volume- centralwavelength,4600˚A,withchipgapsof∼16˚Acentred
Phase Holographic grism with 1070 lines pixel−1, blazed at at 4221 ˚A and 4587 ˚A.
6597˚A. This setup leads to a dispersion of ∼1.0 ˚A pixel−1 The data were reduced with the python based reduc-
and a resolution ∼7 ˚A (as determined from the FWHM of tion pipeline and software package for SALT, PySALT
theemissionlinesinthecomparisonspectrum).Thespectra (Crawfordetal.2010).Wavelengthcalibrationwasachieved
were extracted using standard procedures (bias and back- using comparison spectra of Copper and Argon lamps,
groundsubtraction,flatfieldingandwavelengthcalibration) taken immediately after the observation with the same in-
MNRAS000,1–12(2017)
SXP7.92 5
strument configuration. One dimensional, background sub- 200
tracted spectra were then extracted and normalised using
the standard IRAF packages. wer150
Po
3 RESULTS AND ANALYSIS mb-Scargle100
3.1 Overview Lo 50
Figure 1 shows the ∼18 years of OGLE I-band photometry,
0.10 0.15 0.20 0.25 0.30 0.35 0.40
aswellasanoverviewofthelongtermX-rayactivityofthe Frequency(s−1)
source,andsummarisestheopticalandX-raydatapresented
in this paper. SXP7.92 was detected on two occasions by
Figure 2. Example periodogram of SXP7.92. Figure shows the
RXTE: MJD = 53068 and MJD = 54630 – 54645. These
Lomb-Scargle periodogram of the 0.2–10.0 keV EPIC-pn light
two detections are indicated on Fig. 1 and clearly coincide
curveofSXP7.92fromobservation0700580401.
with the peaks of optical outbursts. Though these optical
outbursts each lasted 1 – 2 years, and the RXTE coverage
was extensive throughout these periods, SXP7.92 was only Table 1.SummaryoftheX-raytimingresultspresentedinthis
detectedonthosetwooccasions.Wecannotdefinitivelydraw paper.Asdiscussedinthetext,thelackofimagingcapabilitiesof
any conclusion as to the absolute flux of the source during RXTE meansthatitisimpossibletodeterminethebackground
these intervals, but any source with Lx∼5×1035 erg s−1 and levelandhencetheabsolutefluxandpulsedfractionofthesource
a pulsed fraction >0.3 would have been detected in these intheseobservations.
observations. Observation Date PulsedPeriod PulsedFraction
90086-01-01 2004-03-04 7.9±0.1 -
93037-01-49 2008-06-13 7.918±0.002 -
3.2 X-ray Position and Timing 93037-01-50 2008-06-20 7.9206±0.0001 -
93037-01-51 2008-06-28 7.92±0.05 -
Israel et al. (2013) reported the detection of an X-ray tran-
15504 2013-09-06 7.91807±.00001 0.50±0.02
sient in Chandra observation 15504 (MJD = 56541) at
0700580401 2013-11-20 7.9173±0.0001 0.60±0.02
a position RA(J2000)=00:57:58.4, Dec(J2000)=-72:22:29.5
0700580601 2013-12-26 7.9167±0.0002 0.34±0.01
witha1σuncertaintyof1.5(cid:48)(cid:48).Thesourcewassubsequently
searchedforperiodicsignalsbymeansofanadhocdetection
search, generated from the EPIC-pn light curve of observa-
algorithm and showed a large and highly significant (confi-
tion 0700580401. A strong signal is detected in both obser-
dence level above 12σ) signal at a period of about 7.9 s
vation0700580401andobservation0700580601ataperiods
(Nyquist period at 6.3 s) in its power spectrum, confirming
7.917 s.
it to be SXP7.92 and the same source as seen by RXTE.
Theerrorsonthedetectedperiodsweredeterminedvia
In order to reject the hypothesis of a spurious re-
a bootstrap-with-replacement method: 1000 artificial light
sult, the signal was automatically checked by means of the
dither region ciao task3 and confirmed to be intrinsic to the curves were generated for each observation by sampling
withreplacementfromtheoriginallightcurve.Lomb-Scargle
source. A more reliable and accurate period was derived
analysis was performed on each of these light curves in an
by performing a phase-fitting analysis by dividing the light
identical manner to that of the original light curve and
curve in five sub-intervals. We obtained a best period value
of P=7.91807±.00001 s and a pulsed fraction (defined as the period determined. The distributions of the resulting
periods detected are well characterised by Gaussian func-
the semi-amplitude of the sinusoid divided by the average
countrate)of50±2%.Thepulseshapeisalmostsinusoidal, tions with mean 7.9174 and 7.9167 and standard deviation
1×10−4and2×10−4 respectively. Thus we determine the pe-
within the relatively poor time resolution of the data (in-
riod of SXP7.92 to be 7.9174±.0001 s as of MJD = 56616
trinsic ACIS binning time of 3.14 s).
and7.9167±.0002sasofMJD=56652.Table1summarises
AsearchoftheChandra archiverevealedthatSXP7.92
these results.
is detected in two previous Chandra observations, albeit
Figure3showstheRXTE andXMM-Newtonpulsepro-
at a fainter level (see Table A1). Shortly after, SXP7.92
files (folded light curves) for a subset of observations, so
was detected in two of the three XMM-Newton obser-
chosenbecause(a)theydemonstratetherangeofpulsepro-
vations of Coe et al. (2015) (see Table A1) at posi-
files shapes shown by this sources and (b) because these
tions RA(J2000)=00:57:58.7, Dec(J2000)=-72:22:28.0 and
instrumentsandmodeshavehightimingresolution(i.e.sig-
RA(J2000)=00:57:58.7,Dec(2000)=-72:22:28.4respectively.
This position is ∼20.6 arcminutes to the south-east of nificantly less than the pulse period, unlike Chandra or the
EPIC-MOSinstrumentsonboardXMM-Newton.)Eachlight
the originally reported position and conclusively rules out
curve has been folded on the best period detected between
AzV285 as the optical counterpart to SXP7.92.
7.0 s and 8.5 s in each observation, with the phase = 0
The XMM-Newton light curves were searched for pe-
point defined by the start time in observation 93037-01-49
riods using the numerical implementation of the Lomb–
(though we note that we have made no correction to take
Scargle method (Lomb1976;Scargle 1982; Press &Rybicki
into account the changing pulse period between the Phase
1989). Figure 2 shows a resulting periodogram from this
= 0 point and each individual observation). The pulse pro-
files all appear to have a common double, possibly triple,
3 http://cxc.harvard.edu/ciao/ahelp/dither_region.html peaked structure, varying from symmetric to asymmetric.
MNRAS000,1–12(2017)
6 E. S. Bartlett et al.
93037-01-49 Table 2. Best fit parameters for the phabs*vphabs*powerlaw
3.8
model fit to the data, both to individual observations and si-
3.6 multaneously. The phabs component corresponds to NH,Gal, and
1)− was fixed at 6×1020 cm−2. Errors, where reported, are the 90%
Rate(s 3.4 cfiotn.fidenceintervals.“dof”referstothedegreesoffreedomofthe
unt 3.2
Co Observation Instrument NH,i Γ χ2/dof
3.0 ×1022cm−2
15504 ACIS-I 0.4+0.3 0.40+0.09 100.0/118
2.8 −0.2 −0.08
0.0 0.5 1.0 1.5 2.0 0700580401 EPIC-pn 0.29+0.09 0.57±0.06 100.9/122
Phase −0.08
5.4 93037-01-50 0700580601 EEPPIICC--MMOOSS21 0.31+−00..0098 0.55±0.05 186.5/185
5.2 Simultaneous 0.32+0.07 0.49±0.03 506.4/485
−0.06
1)−5.0
(s 4.8
ate caused by changes in accretion mode and geometry due to
R 4.6
nt variations in the mass accretion rate. The pulsed fractions
ou 4.4 of the Chandra and XMM-Newton light curves, included in
C
4.2 Table1,werecalculatedbyintegratingoverthepulsedpro-
4.0 fileandcomputingthefractionoftheprofilethatpulsedvs.
3.8 the total flux.
0.0 0.5 1.0 1.5 2.0
Phase
3.2 93037-01-51 3.3 X-ray Spectra
3.0 The spectral analysis reported here was performed using
1)− XSPEC (Arnaud 1996) version 12.9.0i. The spectra of the
(s 2.8
ate recentChandra andXMM-Newton observations(i.e.there-
R 2.6
nt cent period of activity) were fit both individually, to see if
Cou 2.4 there was any evidence for variability, and simultaneously
with the Swift combined spectrum, to better constrain the
2.2
model parameters. The spectra were fit with an absorbed
2.0 power law (phabs*vphabs*powerlaw in XSPEC); the typical
0.0 0.5 1.0 1.5 2.0
Phase model for BeXRBs (Haberl et al. 2008). The photoelectric
1.0 absorptionwassplitintotwocomponents,aGalacticcompo-
0700580401
nent,NH,Gal (thephabs component),toaccountforGalactic
0.8 foregroundextinction,fixedto6×1020cm−2(Dickey&Lock-
1)− man1990)withSolarabundancesfromWilmsetal.(2000),
ate(s 0.6 and an intrinsic absorption, NH,i to account for attenuation
R localtothesourceandtheSMCwithelementalabundances
ount 0.4 of 0.2 solar (for elements heavier than He; the vphabs com-
C
ponent). Table 2 shows the model parameters for both the
0.2 individual and simultaneous fits to the data. All the obser-
vations are well fit by this simple model with no evidence
0.0 0.5 1.0 1.5 2.0
forvariabilityinthespectralshapebetweentheobservations
Phase
(i.e.theabsorbingcolumnandthephotonindexareconsis-
Figure 3.RXTE andXMM-Newton pulseprofilesofSXP7.92, tent with being constant) despite the variability in source
so chosen as these instruments have the best timing resolution flux, represented by a variable normalisation parameter in
andsoofferthebestinsightintothevariationintheshapeofthe the spectral fit (see middle panel of Fig. 4).
pulse profile. The count rate for the RXTE profiles (top three Figure 4 shows the five spectra, the Chandra ACIS-
panels) represent the count rate for the entire observation, i.e.
I spectrum from observation 15504 (black), The XMM-
across the whole field of view and without any background fil-
Newton EPIC-pn spectrum from observation 0700580401
tering applied. As such, these count rate values are inclusive of
(red),theEPIC-MOS1andMOS2spectrafromobservation
theX-raybackgroundandmaycontaincontributionsfromother
sourcesinthefieldofview. 0700580601(greenandbluerespectively)andthecombined
Swift XRT spectrum (cyan), along with the simultaneous
model fit.
Thetriplepeakshapeappearstobemoreprominentinthe
final, most recent pulse profile, but making such inferences
3.4 Spectral Classification
across instruments with different resolutions and different
signaltonoisedatashouldbedonewithcaution.Theshape Figure5showsthenormalisedSALTspectrumoftheoptical
and evolution of the RXTE pulse profiles are discussed at counterpart of SXP7.92 taken on MJD = 56639 (2013-12-
length in Coe et al. (2009), who conclude they are likely 12). This spectrum corresponds to the second blue line in
MNRAS000,1–12(2017)
SXP7.92 7
0.1 There appears to be some evidence for Siiv lines, particu-
larly at λ4088, suggesting a spectral type of B1-B2. The
strength of the Siivλ4088 line relative to the neighbouring
V−1 0.01 Oiiλ4097 line in particular points to a B1.5 spectrum. The
e
−1unts s k 10−3 Hsotreaarii(λsiλu.e4b.1dl4uw4ma/4rinf1/o2ds1witrayartcfiloastsisasriVn(d)IiVicfa-wVtei)vaeisfosBufma2.meTaahisnepsefeacqtrruraeilngchcletasdpswaBna1erfl,
o
c of Figure 5 shows the NTT spectrum of SXP7.92 around
0.01 theHαλ6563regionofthespectrum.Thereisunambiguous
evidence of emission, justifying an ‘e’ classification for the
V)−1 10−3 optical counterpart.
e10−4 Zaritsky et al. (2002) quote a visual magnitude for the
k
2keV −2ms −1 10−5 oinpgticaaldcisotuanntceerptaortthtoe SSXMPC7.o9f26o2f±V2=k1p5c.72(G±r0a.0c3zy.kAsestumal-.
otons c1100−−76 2in0t1e4r)staenlldaruesxintigntchtieonrevlaatliuoenooff6G×u¨1v0e2r0&cmO¨−2z(eDl(ic2k0e0y9)&anLdocakn-
Ph man 1990) to calculate AV gives us an absolute magnitude
(10−48 MV=−3.51±0.07.Comparingthisvaluetothetablesofab-
solute magnitudes for OeBe IV/V stars in Wegner (2006;
2 B0.5Ve=−3.63, B1Ve=−3.35, B1.5Ve=−3.05, B2Ve=−2.72)
supports a spectral classification of B1Ve. We note, how-
2χ 0 ever, that our calculated absolute magnitude may be mis-
∆
leadingasitdoesn’ttakeintoconsiderationanylocalextinc-
−2
tion (which may vary throughout the orbit) and the error
doesn’t take into account the depth of the SMC, estimated
−4
to be as much as 6.2±0.3 kpc (Haschke et al. 2012) or any
0.2 0.5 1 2 5 10
local extinction.
Energy (keV)
3.5 Optical Period
Figure 4. The 0.2–12.0 keV spectrum of SXP7.92 from the It is clear from Figure 1 that the optical counterpart of
various X-ray telescopes. From top to bottom in the top panel SXP7.92 spends a large fraction of its time in a quiescent
the XMM-Newton EPIC-pn (red), XMM-Newton EPIC-MOS1
state, with V∼15.9, but with several major outbursts each
(green)andEPIC-MOS2(blue),theChandraACIS-I(black)and
lasting 1–2 years. Also shown in more detail in this Figure
the Swift combined XRT (cyan) spectrum. The top two pan-
is the most recent outburst. The data from epoch MJD =
els show the background subtracted spectrum with the best fit
phabs*vphabs*powerlaw model,fittoallobservationssimultane- 56450 – 56710 and epoch MJD= 56800 – 57100 were de-
ously. The top panel shows this convolved with the relevant in- trendedwithsimplesecondandthirdorderpolynomialsre-
strumentalresponses,themiddlepanelshowsthesameinν−Fν spectively.Theseintervalsweresearchedforperiodicsignals,
form.Thebottompanelshowstheresiduals. both as separate intervals and as an entire MJD = 56450 –
57100 period. This was performed in the same manner as
the X-ray light curve: Using the numerical Lomb-Scargle
Fig.1andwastakenjustafterthesharpminimum,between method.TheresultingperiodogramsareshowninFigure6.
the two positive XMM-Newton detections (MJD = 56616 All three epochs show a strong peak at ∼40 d, as reported
and 56663). A box filter smoothing of 3 pixels has been ap- by Schmidtke & Cowley (2013).
plied4 as well a redshift correction of -158km s−1(Richter Error determination was, again, performed via boot-
et al. 1987) corresponding to the recessional velocity of the strapping.AswiththeX-raylightcurve,1000artificiallight
SMC. The absorption lines used to classify early type stars curvesweredrawnfromthedetrendedMJD=56500–57100
arelabelled.Evansetal.(2004)presentadetaileddiscussion data set, by sampling with replacement, and were searched
ofthedifficultiesinclassifyinglowluminosity(i.e.subgiant) forperiodicitiesinthesamewayastheoriginallightcurve.
B-typestarsintheSMC:Itisnearimpossibletodetermine The resulting distribution of periods recovered is well de-
whether the absence of a metal line in a stellar spectrum scribed by a Gaussian function with mean 40.04 and stan-
is due to temperature or the result of the low luminosity darddeviation0.32,thuswewedeterminetheopticalperiod
and/or low metallicity environment. As such, we use the of SXP7.92 to be 40.0±0.3 days.
methoddescribedbyEvansetal.(2004)andfurtherimple- Thedetrendeddataforeachoftheseperiodswerefolded
mented by Evans et al. (2006) and adopted a conservative on the proposed binary period and are shown in Fig. 6. We
approach in our classification. note that there appears to be an evolution from “flaring”
The lack of evidence for Heiiλλ4200,4541or4686 above MJD = 56450 – 56710, to dips in epoch MJD= 56800 –
thenoiseofthedata,impliesaspectralclasslaterthanB0.5. 57100, offset from flare phase. The phases of these dips are
indicated in Fig. 1 by the vertical dashed lines. We note
that this is not the first report of“dips”in the optical light
4 http://docs.astropy.org/en/stable/convolution/index. curveofaBeXRB:similarsizeddipswereseenintheOGLE
html light curve of LXP169 (Maggi et al. 2013) and SMCX-2
MNRAS000,1–12(2017)
8 E. S. Bartlett et al.
1.1 ǫHNiIIHeI+IIOIIISiIVOIIδHSiIVHeIHeI HeII γH HeIOIIOIIHeIMgIIHeIINiIISiIIISiIIISiIIIOIINiIIOIIOIIOIIHeIIHeI βH OIII SiIVOII δH SiIVHeI HeI αH 1.25
1.0 1.20
ux0.9 1.15
Fl 1.10
d0.8
alise0.7 1.05
m 1.00
Nor0.6 0.95
0.5 0.90
0.4 0.85
4000 4200 4400 4600 4800 4050 4100 4150 6500 6550 6600
WWaavveelleennggtthh˚A˚A
Figure5.LefthandpanelshowstheSALTspectrumofSXP7.92inthewavelengthrange3900–4900˚A.Thespectrumhasbeennormalised
toremovedthecontinuumandredshiftcorrectedby-158kms−1.Atomictransitionsrelevanttospectralclassificationhavebeenmarked.
Themiddlepanelshowsasubsetofthedatathatisdiscussedexplicitlyinthetexttofacilitateidentificationoftheweakermetallines.
TherighthandpanelshowstheEFOSC2HαspectrumofSXP7.92.Again,thespectrumhasbeennormalisedtoremovedthecontinuum
andredshiftcorrectedby-158kms−1.Thebrokenlineat6563˚AistherestwavelengthofHαandiscentredonthemiddledepression
has shown a similar evolution from“flares”to dips (Rajoe- dromedae (Clark et al. 2003). Whilst determining whether
limananaetal.2011).ThiswillbefurtherdiscussedinSec- a double peaked emission line is a genuine shell line, rather
tions 4.1 and 4.2. than due to V/R variability or another mechanism is non
trivial, Hanuschik (1996) note that only shell-type absorp-
tion can cause the central peak to dip below the adjacent
continuum, as seen in SXP7.92. Such an inclination would
4 DISCUSSION
also lead to deep narrow Balmer profiles (again, as seen in
SXP7.92)asinsteadofseeingphotosphericabsorptionpro-
4.1 The Circumstellar Environment
files from the rapidly rotating star, we are in fact seeing
Figure 7 shows the SALT spectrum of the optical counter- absorption from circumstellar material at large radii (and
part of SXP7.92 alongside a selection of isolated SMC B hence low velocities).
stars. Be stars, both isolated and in BeXRBs, are charac-
The width of the Hα profile can be used to calculate
terised by Doppler broadened Balmer series due to their
the radius of the circumstellar disc (Coe et al. 2006). Using
rapid rotation, yet the Balmer series of SXP7.92 appear the peak separation ( 11±1 ˚A) of the line profile as a char-
comparabletothoseinisolatedBstars,suggestingSXP7.92
acteristic width and assuming the circumstellar disc lies in
has a low vsini. One explanation is that we are viewing our line of sight (i=90◦), and a stellar mass, M∗, of 18.0M(cid:12)
the star near pole on (sini∼0). Such a scenario, however,
(correspondingtoaB0Vstar,Cox2000),theradiusisgiven
wouldleadtosinglepeakedemissionlinesfromthecircum-
by:
stellar disc whereas the Hα profile is unambiguously double
peaked.DoublepeakedHαprofilesarecommoninBeXRBs;
t1h9e6i1r)sihsaapneaarneda ovfaraicatbiivleityres(“eVa/rcRh.vaTrhiaebseiliatyre”;eMxpclLaianuegdhlbiny Rcirc=G(0M.5∗∆sivn)22i (1)
aninclinedcircumstellardiscwithblueandredshiftedemis-
where ∆v is the peak separation in velocity space. This
sion at alternate ends. In such cases the central depression
resultsinaradiusof(3.8±1.3)×1010m(thoughclearlyitcan
cases tends to be shallow. On occasion, the Balmer series
be smaller if the inclination is different). We note here that
of BeXRBs show evidence of“infilling”, i.e. a superposition
theOBstarsintheSMCareoftenoverlyluminouscompared
of a broad absorption line and a narrow emission feature,
toGalacticsystemsandhencethemasscouldbesomewhat
SXP7.92seemstoshowtheoppositeofthis:emissioninthe
lowerand,again,theresultingdiscsizecomparablysmaller.
wings of the line and a deep narrow core - a shell profile.
Shelllinesareobservedinsubset(∼10-20%)ofisolated If the dips in I-band flux, discussed in Section 3.5, are
Bestarsandarecharacterisedbyadeepabsorptioncoreem- entirelyduetosomeobscuringobjectpassinginfrontofthe
bedded in an emission line (Hanuschik 1996; Porter 1996). circumstellar disc, then we can estimate its size using the
Thisisexplainedbyabsorptionoflightfromthediscandthe relation:
central star by the circumstellar disc itself and is a natural ∆F R2
consequence of viewingthe Be star equator or near equator = (2)
F AR2
on.Hanuschik(1996)usethefractionofBeshellstarstocal- circ
culateahalfopeningangleofBestardiscsof13◦ andshow where F is the combined flux of the star and circumstellar
that these discs must be thin at the inner edge, increasing disc, ∆F is the drop in flux caused by the occultation, R is
in scale height at greater radii. Along with the small half the radius of the obscuring object, Rcirc the radius of the
opening angle, this implies a concave (i.e. hydrostatically circumstellar material and A is a scale factor to account for
balanced)disc.Theyshowthat,foranarrowrangeofi,the thedifferentgeometriesbetweentheemittingandobscuring
shell phenomenon can be transient, as seen in e.g. o An- bodies.RearrangingforRandsubstitutinginformagnitude
MNRAS000,1–12(2017)
SXP7.92 9
ω 2ω 3ω 4ω 5ω 6ω
0.4
1.005 56450<Time(MJD)<57100
0.3
1.000
0.2
0.995
0.1
0.990
e
ud 0.4
agnit 1.005 56450<Time(MJD)<56710 0.3 Lom
bandM 10..090905 0.2 b-Scarg
I- le
d 0.1 P
e 0.990 o
d w
etren 1.008 0.4 er
D 1.006 56800<Time(MJD)<57100
1.004 0.3
1.002
0.2
1.000
0.998 0.1
0.996
0.0 0.5 1.0 1.5 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
BinaryPhase Frequency(s−1)
Figure6.ThefoldedOGLEIVdataandLomb-ScargleperiodogramsfromtheperiodsMJD=56500–57100(toppanels),MJD56450
–56700(middlepanels)andMJD56800–57100(bottompanels)respectively.Thelightcurvesfoldedattheproposedbinaryperiodof
40.0d.AlllightcurveshavethesamezerophasepointT0=56850,andallhavebeenbeende-trendedbeforefolding.Anapparentshift
inthephaseoftheflareswithrespecttothedipsisevident.
gives:
(cid:112)
R=ARcirc 1−10−0.4∆m (3)
where ∆m is the change in magnitude. From Fig. 1 we can
B1V
seethat∆I variesfrom∼0.04–0.12,whichleadstoasizees-
timate of ∼(5−10)A×109 m using our estimates above for
Offset B2V tAheiscinroctumpostseslilbalredwisicthsoizuet.mDeotreerminifnoirnmgatthioenspoencifithcevanlauteuroef
bitrary astnedll/aorrdoisrciebnutattiinonthoefstihmepolecsctulctainsegwobhejercetb,oatnhdotbhjeecctisrchuamve-
+Ar B3V thesamegeometry, A=1.Onepossibilityisthatthisobject
ux is an inhomogeneity in the material in the equatorial plane
Fl
d (i.e. a lump), however, it is difficult to understand how a
malise lump of material could be supported for several orbits in a
Nor differentially rotating disc.
SXP7.92
If we interpret the optical period as an orbital period
then it follows that some feature associated with the tran-
sit of the NS is an obvious explanation for these regular
features. As discussed in Section 1, the orbital periods of
4000 4100 4200 4300 4400 4500 4600 4700 4800 4900 BeXRBarecorrelatedwiththeirpulseperiods(Corbet1984,
Wavelength˚A 1986).The40.0±0.3disinthetheexpectedregionofanor-
bitalperiodforan∼8spulsarandwouldlocatethissource
alongsideothersimilarperiodsourcessuchasSXP7.78(Porb
Figure7.spectrumofSXP7.92alongsidethoseofisolatedSMC = 44.8 d) and SXP8.80 (Porb = 33.4 d) on the Corbet Di-
B stars. Despite usually being characterised by Doppler broad- agram. Bird et al. (2012) present an extensive study on the
enedBalmerlines,theBalmerseriesofSXP7.92appearcompa- periodicitiespresentintheOGLEI-bandlightcurvesofthe
rabletothoseinisolatedBstars. SMC BeXRBs. They show that aliasing between the non-
radialpulsationsoftheBestarandtheOGLEsamplingcan
leadtoapparentperiodsinthe2-100dregion,however,the
strong correlation between the X-ray and optical behaviour
MNRAS000,1–12(2017)
10 E. S. Bartlett et al.
of SXP7.92 in the period MJD = 56600 – 56660 supports dips in the light curve occur over a period of ∼200 d, sug-
the hypothesis that this period is orbital in origin. gestingthatthediscisconsistentlypresentthroughoutthis
Rajoelimanana et al. (2011) attribute the evolution interval, however, the 2013 XMM-Newton non-detection at
from flares and dips seen in SMCX-2 to a precessing elon- MJD = 56597 occurs 56 days after the Chandra detection
gateddisc:Whenthesemi-majoraxisoftheBediscextends and suggests a drop in flux of at least two order of mag-
toeithersideofthestar(withrespecttothelineofsightof nitude, followed by similar increase just 15 days later. The
the observer) we see flares when the NS interacts with the nondetectionandlargedynamicrangesuggeststhatnoac-
disc. When the semi-major axis of the disc points towards cretion disc exists during this period. We note that several
theobserver,theinteractionoftheNSanddiscblockslight (though not all) BeXRBs in the SMC have been detected
fromtheBestar,creatingdips.Whilsttherearesomesimi- between outbursts with Lxin the range 1033−34erg s−1(e.g.
laritiesinthebehaviourofSMCX-2andSXP7.92(e.g.the Laycock et al. 2010) implying quiescent accretion, in one
dips are seen as the source fades) there are some clear dif- form or another, can occur in a subset these systems.
ferences - most noticeably, that the flares are much more An alternate interpretation is that the flares and dips
uniforminSMCX-2(i.e.weseeseveralcycles)andofasim- arecausedbyaprecessingelongatedcircumstellardisc,like
ilar amplitude to the dips, whereas we see only a few flares that inferred by Rajoelimanana et al. (2011) for SMCX-2.
in SXP7.92 with ∆I ranging from 0.07–0.27. Again, this scenario is difficult to reconcile with the data,
ThevalueofRderivedaboveisnotcompatiblewiththe with the circumstellar disc appearing to disappear almost
radius of a NS alone (generally assumed to be 10–15 km). completely(asevidencedbythedroptotheI-bandbaselevel
Maggi et al. (2013) invoke a similarly large sized obscuring coinciding with little to no Balmer emission in the optical
object,≥3.5×109 m,toexplainthedipsinthelightcurveof spectra)mid-outburst,andre-establishing∼200dlater.Itis
LXP169.Theyclaimthatthesedipsareduetothetransitof unclearhowsuchadramaticallyvariabledisccouldsupport
theNSsystemacrosstheBestar,ratherthananobscuration a stable super-orbital period.
of part of the circumstellar disc, and that the size can be
understood in terms of the inflow of material onto the NS
from the stellar wind which would create a total accreting
4.3 Theoretical Prospects
systemoftherequisitesize.Itseemsprobable,whateverthe
detailedexplanation,thatSXP7.92andLXP169couldhave Statistically, identifying a BeXRB in a shell orientation
much in common. should be quite a rare occurrence. At the time of writing
we are not aware of any other such system in the SMC or
theLMC,thoughwenotethatwewouldexpectLXP169to
4.2 Global Behaviour
besuchasystemiftheorbitalplaneisperpendiculartothe
The long term interplay between the optical and X-ray ac- rotational axis of the Be star (i.e. the NS orbits in plane of
tivity of SXP7.92 can be easily understood within our cur- circumstellar disc). To date, no optical spectra of LXP169
rent framework of BeXRBs. When the optical counterpart have been published.
of SXP7.92 is in a quiescent state there is little to no cir- Reig et al. (2016) present the long term spectro-
cumstellar disc present (as evidenced by the low levels of scopic variability of several Galactic HXMBs with data
Balmer emission in the optical spectra at minimum). No spanning the period 1987 – 2014 (though this varies
material is available for accretion onto the NS and so the for individual sources). In their sample, the BeXRBs
system is similarly dormant in X-rays. During the 1–2 year 4U0115+63/V635 Cas, RXJ0146.9+6121/LSI+61◦235,
outburststhediscsgrowsconsiderably,providingareservoir RXJ0440.9+4431/LSV+4417, 1A0535+262/V725Tau,
of material for the NS to accrete from. Within the most re- IGRJ06074+2205, GROJ2058+42, SAXJ2103.5+4545,
cent outburst we are seeing the same behaviour on much IGRJ21343+4738and4U2206+54,haveatsomepoint,had
shorter timescales: We see rapid optical flares, occurring at their Hα line profiles characterised as a shell profile. Their
∼Porb.FromthelowerpanelsofFig.1wecanseethatX-ray behaviours vary dramatically, with e.g. 4U0115+63 having
activity closely tracks the optical during this episode, with shell phases of < 2 months whilst 4U2206+54 appears to
thesourceindeclineintheX-rayastheflareatMJD∼56610 be in a near constant shell configuration over the entire
abates,fallingbelowthedetectionthresholdandthenrising 24yearsofmonitoring-highlightingbothhowvariableand
with the subsequent flare. The apparent peculiarities in the how stable the circumstellar discs in these systems can be.
opticalspectraofSXP7.92(verynarrowBalmerseries,deep If SXP7.92 is indeed in a shell orientation, this sys-
central depression in the Hα ) can be explained by the op- tem has the potential to be a unique laboratory for un-
ticalcounterpartofSXP7.92beingashellstarviewededge derstanding BeXRB systems. The very narrow line profiles
on, though we note that the low resolution of the EFOSC2 of SXP7.92 could be used to look for motion in the Be
spectrummeansthisconclusionshouldbetreatedwithcau- star itself. Combined with X-ray pulse timing analysis (see
tion. Townsend et al. 2011), this could be used to determine the
Anopenquestionisthecauseoftheflaresanddipsseen orbital parameters of the system and hence the masses of
in the optical light curve of the 2013 activity. One possibil- both the components of the binary, as the inclination of
ityisthatthedipsareoccultationsofthecircumstellardisc the system, i, is already somewhat constrained. Continued
by an accretion disc that has built up around the NS. We OGLEmonitoringofSXP7.92willallowustoidentifywhen
wouldexpectsuchanaccretiondisctobetransient,forming thedisciscompletelyabsent(e.g.MJD∼56630and57030).
duringaccretionepisodesanddepletingwiththecircumstel- Anopticalspectrumtakenatsuchapointwouldallowusto
lar disc as it is no longer renewed. This scenario seems as determinetherotationalvelocityoftheBestar(again,since
oddswiththeXMM-Newton andSwift non-detections:The we already have a rough idea of the inclination) which, in
MNRAS000,1–12(2017)