X-ray and Optical Study of the Gamma-ray Source 3FGL J0838.8-2829: Identification of a Candidate Millisecond Pulsar Binary and an Asynchronous Polar
PPreprint typeset using L A TEX style emulateapj v. 5/2/11
X-RAY AND OPTICAL STUDY OF THE GAMMA-RAY SOURCE 3FGL J0838.8 − Jules P. Halpern , Slavko Bogdanov , and John R. Thorstensen (Received 2016 November 8; Accepted 2017 January 13) ABSTRACTWe observed the field of the
Fermi source 3FGL J0838.8 − − γ -ray emitting millisecond pulsars(MSPs). We find that 1RXS J083842.1 − ≈ ◦ apart) when the accretionrate is at minimum. High-amplitude X-ray modulation at periods of 94 . ± . . ± . . ± . . ± .
004 minutes, which is consistentwith the orbital period inferred from the X-rays. In any case, this system is unlikely to be the γ -raysource. Instead, we find a fainter variable X-ray and optical source, XMMU J083850.38 − Fermi sources, withthe optical modulation due to heating of the photosphere of a low-mass companion star by, in thiscase, an as-yet undetected MSP. We propose XMMU J083850.38 − − Subject headings: cataclysmic variables — gamma rays: stars — pulsars: general — X-rays: individual (1RXS J083842.1 − − − INTRODUCTIONThe Large Area Telescope on the
Fermi
Gamma-rayObservatory has detected numerous young pulsars, aswell as recycled millisecond pulsars (MSPs) in close bi-nary systems. Most prominent of the new discoveries arethe black widow (BW) pulsars and so-called “redback”systems (Roberts 2013), which comprise a large fractionof the MSPs selected by
Fermi . The BWs are MSPs withsub-stellar mass, degenerate companions, while the red-backs generally have > . M (cid:12) evolved companions. Thelatter are usually close to filling their Roche-lobes, whichmakes them a direct link to the low-mass X-ray binary(LMXB) progenitors of MSPs.Recently, three redbacks have been observed to tran-sition between radio pulsar and accreting states ontimescales of years: PSR J1023+0038 (Archibald et al.2009), XSS J12270 − − γ -raysources. PSR J1023+0038 was initially misclassified asa cataclysmic variable (CV) (Bond et al. 2002, but seeThorstensen & Armstrong 2005 for a contrary interpreta-tion), as was XSS J12270 − Columbia Astrophysics Laboratory, Columbia Univer-sity, 550 West 120th Street, New York, NY 10027-6601;[email protected] Department of Physics and Astronomy, 6127 Wilder Labo-ratory, Dartmouth College, Hanover, NH 03755-3528 which show characteristic dips and flares that are uniqueto this class (Bogdanov et al. 2015; de Martino et al.2013). We employed this test to reevaluate two
ROSAT
All-Sky Survey X-ray sources in
Fermi error circles thatwere spectroscopically classified as CVs by Masetti et al.(2013). Using X-ray and optical time-series data for oneof these, 1RXS J154439.4 − − − − − Fermi error circle, XMMUJ083850.38 − γ -ray source. In Section 5 we show that a third X-raysource in the Fermi error circle is a QSO, probably unre-lated to the γ -ray source. Section 6 discusses the prop-erties of the CV and the MSP candidate in relation toother objects in their respective classes. X-RAY AND OPTICAL OBSERVATIONSThe field of 3FGL J0838.8 − XMM–Newton , on 2015 October 20 (ObsID0764420101, 53 ksec) and 2015 December 2 (ObsID0790180101, 77 ksec). The EPIC pn detector was used inlarge window mode, while the two MOS detectors were a r X i v : . [ a s t r o - ph . H E ] J a n Halpern et al.
TABLE 1Log of MDM Observatory Time-Series Photometry of 1RXS J083842.1 − Telescope/Detector Date (UT) Time (TDB) Filter Exposure (s) Conditions2.4 m/Templeton 2014 Mar 22 02:47–06:47 V 10 Clear2.4 m/Templeton 2014 Mar 23 02:40–06:51 V 10 Photometric1.3 m/Templeton 2015 Feb 17 06:12–09:19 BG38 20 Partly cloudy1.3 m/Templeton 2015 Feb 18 03:51–09:14 BG38 20 Photometric1.3 m/Andor 2016 Mar 16 04:31–06:31 GG420 5 Photometric1.3 m/Andor 2016 Mar 17 04:02–06:02 GG420 5 Photometric
NE acb
NE a----|b----| c----|
Fig. 1.—
Left:
XMM–Newton
EPIC pn image (0.3–10 keV) from 2015 December 2 (ObsID 0790180101) showing the
Fermi error circleof 3FGL J0838.8 − . (cid:48)
6. Right: Finding chart from an MDM 1.3m image taken through a BG38filter. The field is 4 . (cid:48) × . (cid:48)
3. Labeled sources in both images are (a) the CV 1RXS J083842.1 − − − TABLE 2Optical Positions
Label Source R.A. (h m s) Decl. ( ◦ (cid:48) (cid:48)(cid:48) )a 1RXS J083842.1 − − − Note . — Coordinates are equinox J2000.0 configured in small window mode, all with the thin filter.Time resolution is 48 ms for the pn, and 0.3 s for theMOS. In addition, the
XMM–Newton
Optical Monitor(OM) obtained 10 contemporaneous exposures of 5000 sor 4160 s in the V -band on 2015 October 20, and 15 con-temporaneous 4400 s exposures in the B -band on 2015December 2. Figure 1 (left) shows the Fermi
95% errorcircle superposed on the
XMM–Newton pn image of 2015December 2. In addition to 1RXS J083842.1 − − Chandra
ACIS on 2016 July7 for an exposure time of 30 ks (ObsID 17769).The S3 CCD only was operated in continuous clock-ing (CC) mode to avoid pileup of the bright source1RXS J083842.1 − − × − F [ e r g s c m ] Wavelength [ ] R a d i a l V e l o c i t y [ k m s ] Phase (P = 0.068342 d, T = HJD 2457408.8365 ) E m i ss i o n E W () Phase (P = 0.068342 d, T = HJD 2457408.8365 ) Fig. 2.—
Top: Mean spectrum of 1RXS J083842.1 − α emission line, folded on the best-fitting period, together with a sinusoidal fit. The error bars shownare based on signal-to-noise and do not include systematic effects.The data are repeated for a second cycle for continuity. Bottom:Equivalent width of the H α emission line. Templeton, and 5 s exposures with 12 ms dead-time usingthe Andor. A log of the observations is given in Table 1.The 1.3m Templeton images (Figure 1, right) coveredall three X-ray sources, while the Andor and 2.4m Tem-pleton images only included 1RXS J083842.1 − − × − and a FWHM resolution of ≈ . − Optical Spectroscopy
Our mean spectrum of 1RXS J083842.1 − I lines on a blue continuum. He II λ − ]01020304050607080 ( F l u x / C o n t i nuu m ) + O ff s e t − α Line Profile vs. Phase
Fig. 3.—
Rectified spectra of 1RXS J083842.1 − α ,as a function of orbital phase. Each trace is a weighted average ofspectra taken near the phase indicated. Successive traces have beenshifted upward by four times the continuum. of ∼ β (EW ≈
125 ˚A). The spectrum implies a synthetic V ∼ . α using a convolu-tion function tuned to be sensitive to the steep sides ofthe line profile, which are about 1500 km s − apart. Ta-ble 3 lists the resulting radial velocities and uncertainties.To find the period we constructed a dense grid of trialfrequencies and fitted the velocities with least-squaressinusoids at each frequency. This yielded a period near98.4 minutes, with no ambiguity in the daily or month-to-month cycle count. A least-squares best fit of the form v r ( t ) = γ + K sin[2 π ( t − T ) /P ] has T = BJD 2457408 . ± . P = 0 . K = 274 ±
16 km s − (3) γ = 49 ±
11 km s − (4)with an RMS scatter of 36 km s − for the 59 data points.The precise period is 98 . ± .
004 minutes. The middlepanel of Figure 2 shows the sine fit superposed on theradial velocities, and Figure 3 shows the H α line profilesas function of orbital phase.The lower panel of Figure 2 shows the equivalent width(EW) of the H α emission line as a function of spectro-scopic phase. A periodogram of the EW has a peak atexactly the spectroscopic period. Although noisy, the Halpern et al. Fig. 4.—
MDM light curves of 1RXS J083842.1 − TABLE 3H α Spectroscopy of 1RXS J083842.1 − Date Exposure v r σ H α EW(BJD) a (s) (km s − ) (km s − ) (˚A)2457402.8178 900 −
119 6 1742457402.8286 900 234 8 1622457403.8843 480 −
110 6 2352457403.8903 480 −
204 6 2372457403.8962 480 −
287 7 2522457403.9021 480 −
244 6 2172457403.9081 480 −
170 6 2612457403.9140 480 −
19 6 2462457403.9199 480 115 8 2142457403.9259 480 220 9 1672457405.7628 480 −
97 23 2182457405.7687 480 83 15 1882457405.7806 480 268 16 1232457405.7865 480 257 14 1582457405.7924 480 137 15 1602457405.7983 480 − −
120 18 2452457405.8102 480 −
282 14 2132457405.8161 480 −
247 13 2222457405.8221 480 −
187 12 2242457405.8280 480 −
101 9 2692457405.8339 480 44 10 2432457405.8399 480 188 9 2002457405.8458 480 263 9 1842457405.9192 480 279 8 1692457405.9251 480 214 7 1882457405.9310 480 55 7 2252457405.9370 480 −
137 7 2192457405.9429 480 −
267 8 2342457405.9488 480 −
315 7 2772457405.9548 480 −
259 8 3082457407.7556 480 217 12 2862457407.7615 480 243 16 2242457407.7734 480 125 20 1992457407.8030 480 −
223 17 1822457408.7526 480 −
188 14 2202457408.7586 480 −
189 13 3332457408.7645 480 −
28 17 2002457408.7704 480 47 21 1592457408.7763 480 255 16 2002457408.7823 480 251 17 1622457408.7882 480 286 17 2132457408.7941 480 258 17 1402457408.8001 480 50 16 1852457408.8060 480 −
115 15 1942457408.8119 480 −
232 19 2222457408.8178 480 −
305 20 2052457408.8238 480 −
268 16 2652457408.8297 480 −
219 16 2702457408.8356 480 −
37 19 1892457430.8336 480 −
183 4 2672457430.8396 480 −
59 5 2292457430.8455 480 62 5 2012457430.8515 480 201 5 2002457438.7866 480 222 6 1572457438.7925 480 202 8 2582457438.7984 480 122 6 2352457440.8426 600 255 6 2052457440.8499 600 180 6 196 a Barycentric Julian day of mid-integration in the UTC sys-tem.
EW varies sinusoidally with a phase lag of ≈ . Time-series Optical Photometry
The MDM light curves of 1RXS J083842.1 − − C oun t s s − k e V − RX J083842.1−282723XMM−Newton EPIC−pnObsID 079018010110.5 2 5−4−2024 χ Photon energy (keV)0.010.1 C oun t s s − k e V − RX J083842.1−282723Chandra ACIS−S10.5 2 5−202 χ Photon energy (keV)
Fig. 5.—
X-ray spectra of 1RXS J083842.1 − α are broadband filters chosen for their higher throughput,but they do not have standard calibrations. We approx-imated their magnitudes using V and R , respectively,from stars in the UCAC4.The individual time series all display a broad oscilla-tion with a period of ∼ .
07 d, similar to the spectro-scopic period. Although some of the light curves in Fig-ure 4 show a hint of a faster oscillation, a power-spectrumanalysis does not reveal any shorter coherent period. Inaddition, in all three years there is a large change of ≈ . ∼ .
07 d oscillation has large amplitude; whenit is faint the relative amplitude is smaller. This behaviormatches the X-ray (presented in the next section) verywell, with the night-to-night changes explained as mod-ulation of the accretion rate on the beat period betweenthe spin and the orbit of a stream-fed system.3.3.
XMM–Newton Analysis
We fitted the pn X-ray spectrum from thelonger
XMM–Newton observation with a thermalbremsstrahlung model in XSPEC. The temperature is 11.7 keV (Figure 5 and Table 4), and residualscorresponding to Fe K α mekal hotplasma model gives a similar temperature. There is noevidence of a soft blackbody from the white dwarf (WD)surface that is sometimes but not always seen in polars.X-ray photon arrival times were transformed toBarycentric Dynamical Time and extracted from a 20 (cid:48)(cid:48) radius around 1RXS J083842.1 − ≈ .
07 d, as wellas a broader modulation on a timescale of ∼ . Z periodogram of the longer (December2) observation (Figure 8a) shows three peaks, at 88.0minutes, 98.3 minutes, and 14.7 hr, where the latter isconsistent with the beat between the two shorter peri-ods. However, a periodogram of the shorter observation(October 20; Figure 8c) has only a single peak at a pe-riod of 95 minutes, which falls in between the pair at 88.0minutes and 98.3 minutes.Inspection of the light curves reveals what is responsi-ble for the difference in the power spectra. At the mini-mum of the 14.7 hr cycle, the phase of the shorter periodjumps by ≈ ◦ . This is especially clear in Figure 7,where the periodic tick marks switch from marking fluxminima before day 7360 to nearly flux maxima after day7360. This phase jump causes the peak in the Fouriertransform to split into two, straddling the true peak. Byanalogy with amplitude modulation of a carrier signal,one should still see the carrier in the power spectrum,with symmetric sidebands on either side. However, if thecarrier (or any signal) experiences a phase jump, then itspower will be split into two frequencies, neither of whichis the true one. The frequency splitting of the signalshould be equal to the frequency with which the phasejumps, in this case 1 / . − , which is consistent withthe observed splitting of 1 / . − .We tested this interpretation by measuring the powerspectrum of the first part of the December 2 observation,before the minimum of the beat cycle at day 7360 in Fig-ure 7. This restricted power spectrum has a single peakat 94.6 minutes (Figure 8b), consistent with the periodfrom October 20. The same period is also confirmed ina Chandra observation described below. Therefore, weadopt the average value of 94 . ± . XMM–Newton observations as the probable spin pe-riod of the WD.We further confirm this interpretation using simulatedlight curves and power spectra. As an approximation ofthe light curve, a squared sinusoid with a 94.8 minuteperiod was amplitude modulated at the 14.7 hr period,and a phase jump of varying angle was introduced at day7360. This reproduced well the splitting of the spin signalin the observed power spectrum. The best match to theobserved light curve and power spectrum was achievedfor a phase jump of ≈ ◦ , with an uncertainty of ∼ ◦ .We hypothesize that the WD spin modulates the ap-parent X-ray flux by self occultation of the base of acolumn that accretes onto a magnetic pole in an AMHer-like system. The 14.7 hr amplitude modulation su- Halpern et al. TABLE 4X-ray Spectral Fits
Label Source N H (cm − ) kT br (keV) Γ F x (0.3–10 keV) a χ ν (dof) Chandra
ACIS-S, ObsID 17769, 2016 July 7a 1RXS J083842.1 − . +1 . − . × . +1 . − . ... (9 . ± . × − XMM–Newton
EPIC-pn, ObsID 0790180101, 2015 December 2a 1RXS J083842.1 − . +0 . − . × . ± . . ± . × − − . +0 . − . × ... 1 . ± . . ± . × − − . ± . × . +11 . − . ... (1 . ± . × − − . +0 . − . × ... 2 . ± . . +0 . − . × − Note . — Uncertainties are 90% confidence. a Unabsorbed flux in units of erg cm − s − . Fig. 6.—
Background-subtracted
XMM–Newton light curve of 1RXS J083842.1 − ≈ ◦ phase jump after minimum light of the beat cycle. Middle: Hardness ratio of counts in the (2–10 keV)/(0.2–2 keV)bands, in 300 s bins. Bottom: XMM–Newton
OM magnitudes from 5000 s or 4160 s exposures in the V -band. bservations of 3FGL J0838.8 − Fig. 7.—
Background-subtracted
XMM–Newton light curve of 1RXS J083842.1 − ≈ ◦ phase jump after minimum light of the beat cycle. Middle: Hardnessratio of counts in the (2–10 keV)/(0.2–2 keV) bands, in 300 s bins. Bottom: XMM–Newton
OM magnitudes from 4400 s exposures in the B -band. There is a gap in the data at day 7360.05. Halpern et al.
Fig. 8.—
Power spectra of the X-ray light curves of 1RXS J083842.1 − XMM–Newton observation of Figure 7, the94.6 minute spin period is split into peaks at 98.3 minutes and 88.0 minutes by the phase jump after the minimum of the beat cycle atday 7360. The 14.7 hr period is inferred to be one-third of the beat cycle because the phase jump indicates that there is pole switchingspanning ≈ ◦ . The spin frequency is ω and the inferred orbital frequency is Ω. (b) When the power spectrum is restricted to the firstbeat cycle, before day 7360 in Figure 7, a single peak at the true period of 94.6 minutes is recovered. (c) The power spectrum of the XMM–Newton observation of Figure 6, restricted to the first beat cycle (before day 7316.4). The 95 minute period is consistent with theother X-ray observations. (d) The power spectrum of the
Chandra observation of Figure 9, which does not span a minimum of the beatcycle. The 93.9 minute period is consistent with the other X-ray observations. We adopt 94 . ± . perposed on this oscillation occurs as the secondary starmigrates in the rotating frame of the WD. The ≈ ◦ phase jump of the 94 . . ± . ± . Chandra Analysis
We fitted the
Chandra
X-ray spectrum with a ther-mal bremsstrahlung model in XSPEC, finding a tem-perature of 11.2 keV (Figure 5 and Table 4), consis-tent with the
XMM–Newton results. X-ray photon ar-bservations of 3FGL J0838.8 − Fig. 9.—
Top: Background-subtracted
Chandra
ACIS-S3 light curve of 1RXS J083842.1 − XMM–Newton data. Bottom: Hardness ratio of countsin the (2–10 keV)/(0.2–2 keV) bands, in 300 s bins. rival times were transformed to Barycentric Dynami-cal Time and extracted from a 2 (cid:48)(cid:48) radius around 1RXSJ083842.1 − XMM–Newton ones, the be-havior of 1RXS J083842.1 − XMM–Newton values. Since this short observa-tion does not span a minimum of the beat cycle, it doesnot clearly show a phase jump in the spin cycle. Howeverthere is an “extra” dip at day 7578.37 in Figure 9 thatmay be due to partial pole switching of the accretion. XMMU J083850.38 − Fermi γ -ray source, and theabsence of a plausible physical mechanism connect-ing 1RXS J083842.1 − − − XMM–Newton Analysis
We extracted pn light curves of XMMUJ083850.38 − ≈ . ∼
100 s or less.A power-spectrum analysis, excluding the flare, doesnot indicate any periodicity. Upper limits on orbital modulation in the 0.5–10 hr period range are ∼ − . ± . . +11 . − . keV). A blackbody fit is unacceptable, withreduced χ ν = 2 . N H = 0.On December 2, the OM observed XMMUJ083850.38 − ∼ − Time-series Optical Photometry
MDM 1.3m images from 2015 February 17 and 18 arethe only ones that cover this source. Because of itsfaintness and the short individual exposure times, webinned the images in groups of 10. Due to weather,only the 2015 February 18 observation yields a usefullight curve (Figure 12). It shows a broad minimumthat is characteristic of the heating light curve of a BWcompanion to an MSP. The X-ray and optical behav-ior of XMMU J083850.38 − Fermi sources, suchas PSR J1311 − Fig. 10.—
Background-subtracted
XMM–Newton light curves of the MSP candidate XMMU J083850.38 − XMM–Newton
OM magnitudes from 4400 s exposures in B -band on 2015 December 2 (ObsID 0790180101). There is a gap in the data atday 7360.05. the possibility of flaring, which may mask a second dipnear the end of the time series. XMMU J083842.85 − − − . N H = 1 . × cm − is consistent with the total 21 cm N H = 1 . × cm − on the line of sight (Kalberla etal. 2005). This object was also listed as a low-probabilityblazar candidate for 3FGL J0838.8 − Wide-field InfraredSurvey Explorer data. Lacking a radio detection in theNVSS, it is unlikely to be a blazar. We obtained an opti- cal spectrum of XMMU J083842.85 − II λ z = 0 . ± . ≈
45 ˚A (Figure 13). This confirmsthat it is a non-blazar AGN, and probably unrelated tothe
Fermi source. DISCUSSION6.1. − The X-ray light curve of 1RXS J083842.1 − . ± . − Fig. 11.—
XMM–Newton OM B -filter image cutouts around the MSP candidate XMMU J083850.38 − (cid:48)(cid:48) × (cid:48)(cid:48) . Fig. 12.—
MDM 1.3m light curves of source “b” (MSP candidateXMMU J083850.38 − − Fig. 13.—
Optical spectrum of XMMU J083842.85 − biting secondary star or the accretion stream, becauseeither of those would be much narrower. Eclipse ingressor egress of the entire WD would take only a few seconds,less than one bin of the light curves, based on the orbitalvelocities estimated below. The eclipse duration wouldbe < . ≈ ◦ at the minimum of the 14.7 hr modulation. Thisis interpreted as switching of the accretion between notquite antipodal regions, when the accretion almost stops.The effect on the power spectrum is exactly analogous toamplitude modulation of a carrier signal, with a phasejump, which splits the signal into two periods straddlingthe true one. We have recovered the true period, whichwe interpret as the WD spin, in the three X-ray observa-tions reported here. There is no direct manifestation ofthe orbit in the power spectrum, but the 14.7 hr periodmost likely represents the time it takes for the companionstar to migrate ≈ ◦ around the WD, i.e., one-third ofthe beat period of the spin and the orbit. The impliedorbital period is 98 . ± . . ± .
004 minutes, which is consistent with the X-ray inferred orbital period, and not with the spin. Butthe radial velocity amplitude of 274 km s − is too largeto be an orbital velocity of the WD, and the lines are toobroad, ∼ − to be coming from the heated faceof the secondary star. A typical CV with orbital periodof 100 minutes has a 0 . M (cid:12) WD and a 0 . M (cid:12) secondary(Savoury et al. 2011). Typical orbital velocities are then ≈
56 km s − for the WD and ≈
445 km s − for thesecondary. On the other hand, free-fall velocity onto theWD is up to 3900 km s − . So we conclude that theemission lines are located high in the accretion column.We also examined the EW of the H α emission line asa function of spectroscopic phase. The EW varies ≈ . < ∼ − . − XMMU J083850.38 − None of the properties of the CV 1RXSJ083842.1 − − − γ -ray spectralshape and variability index. Accordingly, we proposeXMMU J083850.38 − − − − XMM–Newton observa-tion of XMMU J083850.38 − − Fermi
MSPs are at distances of ∼ × erg s − , in the range of both BWs and redbacks in theradio pulsar state (Roberts et al. 2015). In contrast, ac-creting redbacks have L x ∼ × erg s − (Bogdanovet al. 2015).This would be the first time that a simultaneous X-ray and optical flare is seen from a BW. It may be ashort-period system similar to PSR J1311 − − <
35% amplitude for an assumed sinusoid. The cadence ofthe 4400 s OM exposures is too long to test for such aperiod, which may, in addition, be masked by the flaringbehavior. The MDM optical data, on the other had, haveadequate cadence, and a dip that is characteristic of aheating light curve, but the dip does not repeat withinthe 5.4 hr time series (Figure 12), which suggests thatthe period is > . − ≈ . γ -raypulsar detection, could confirm the identification, andresolve the remaining questions about the basic parame-ters of the binary system. CONCLUSIONSThe γ -ray properties of 3FGL J0838.8 − − − − Fermi sources. A binary period, an im-portant test of this hypothesis, is not yet revealed bythe available data, but follow-up time-series photome-try, optical spectroscopy, and/or a radio pulsar detec-tion, should be able to determine the orbital parametersof the system. ACKNOWLEDGEMENTSWe thank Eric Alper for obtaining time-series pho-tometry of 1RXS J083842.1 − − Chandra
X-ray Observatory. Support for this work was providedby the National Aeronautics and Space Administrationthrough
Chandra
Award Number SAO GO6-17027X is-sued by the
Chandra
X-ray Observatory Center, which isoperated by the Smithsonian Astrophysical Observatoryfor and on behalf of the National Aeronautics Space Ad-ministration under contract NAS8-03060. This investiga-tion also uses observations obtained with
XMM–Newton ,an ESA science mission with instruments and contribu-tions directly funded by ESA Member States and NASA.,an ESA science mission with instruments and contribu-tions directly funded by ESA Member States and NASA.