Detection of Millihertz Quasi-Periodic Oscillations in the X-Ray Binary 1RXS J180408.9 − 342058
Kaho Tse, Duncan K. Galloway, Yi Chou, Alexander Heger, Hung-En Hsieh
MMNRAS , 1–6 (2020) Preprint 15 October 2020 Compiled using MNRAS L A TEX style file v3.0
Detection of Millihertz Quasi-Periodic Oscillations in the X-RayBinary 1RXS J180408.9 − Kaho Tse − ★ , Duncan K. Galloway , , , Yi Chou , Alexander Heger , − , andHung-En Hsieh School of Physics and Astronomy, Monash University, Victoria 3800, Australia Graduate Institute of Astronomy, National Central University, Jhongli 32001, Taiwan Department of Physics, National Central University, Jhongli 32001, Taiwan OzGrav-Monash, School of Physics and Astronomy, Monash University, VIC 3800, Australia Joint Institute for Nuclear Astrophysics - Center for the Evolution of the Elements (JINA-CEE), Monash University, Vic 3800, Australia ARC Center of Excellence for Astrophysics in Three Dimensions (ASTRO-3D), Australia
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
Millihertz quasi-periodic oscillations (mHz QPOs) observed in neutron-star low-mass X-raybinaries (NS LMXBs) are generally explained as marginally stable thermonuclear burning onthe neutron star surface. We report the discovery of mHz QPOs in an
XMM-Newton observationof the transient 1RXS J180408.9 − ∼ / 𝑓 power-law noise continuum. Neither the QPOsignals nor the power-law noise were present during the April observation, which exhibiteda 2 . × higher luminosity and had correspondingly more frequent bursts. When present, theQPO signal power decreases during bursts and disappears afterwards, similar to the behaviourin other sources. 1RXS J180408.9 − Key words: stars: neutron – X-rays: bursts – X-rays: individual: 1RXS J180408.9 − Type I X-ray bursts are thermonuclear-powered flashes on thesurface of neutron stars in low mass X-ray binaries. These ener-getic outbursts are triggered by a thermonuclear runaway process.Through Rocbe-Lobe overflow, material accretes from the donoronto the neutron star and is then compressed and heated understrong ( ∼ g cm − ) gravity. Upon reaching the ignition condi-tion, thermonuclear runaway is triggered, usually by the triple − 𝛼 reaction, and the 𝛼𝑝 process operates to burn helium up to the irongroup. Followed by the rp process (rapid proton capture process),hydrogen is burnt close to the proton drip line to produce heavierelements and release additional energy. The bursts typically reachabout 10 erg s − , and last for some ten to a hundred seconds, de-pending mainly on their hydrogen and helium mass fractions (seeStrohmayer & Bildsten 2003 and Galloway & Keek 2017 for re-views). ★ E-mail: [email protected]
Revnivtsev et al. (2001) first discovered ≈ − − . (cid:164) 𝑀 Edd , where (cid:164) 𝑀 Edd = . × − × . /( 𝑋 + ) M (cid:12) yr − is the Eddington-limitedrate with the hydrogen mass fraction 𝑋 , in their simulations usingthe 1D stellar hydrodynamics code Kepler (Weaver et al. 1978;Woosley et al. 2004). They interpreted this oscillation, which takesplace within a narrow range of accretion rates, as the transitionalphase between stable and unstable nuclear burning, as also pre-dicted by one zone models (e.g., Paczynski 1983). In this regime,the thermonuclear generation rate is comparable to the cooling rate,which leads to this oscillatory behaviour. The oscillation period, 𝑃 (cid:39) √ 𝑡 therm 𝑡 acc , where 𝑡 therm and 𝑡 acc represent the time for gasto cool (thermal) and for replenishing fuel (accretion) respectively.This period found in the Heger et al. (2007) simulations is consistentwith the reported range of frequencies of mHz QPOs. © a r X i v : . [ a s t r o - ph . H E ] O c t K. Tse et al.
Figure 1.
The 0 . − XMM-Newton observations on 2015 March 6 and 2015 April1, respectively. The thermonuclear bursts are evident as the brief increases of the count rate above the persistent level. The burst rate and persistent countrate increase by about two times in the second observation period ( bottom panel ). Significant periodic signals appear in the labeled segments from the firstobservation, whereas no periodicity was detected in the second observation. mHz QPOs have been detected in several other sources(Strohmayer & Smith 2011; Linares et al. 2012; Strohmayer et al.2018; Mancuso et al. 2019), that are evidently related to the ther-monuclear burning on the neutron star surface, as Type I X-raybursts are also present in the observations. Although there are dis-crepancies between the observations and theory, for instance thedisagreement in the accretion rate for the oscillations to exist, addi-tional samples could enhance our understanding of the marginallystable burning or even put constraints on NS parameters, such as itsequation of state (Stiele et al. 2016).1RXS J180408.9 − 𝑙 = . ◦ , 𝑏 = − . ◦ ) wasidentified as a transient X-ray binary in 2012 when a Type I X-rayburst from the source was detected by INTEGRAL . It is classified asan "atoll" source, which traces out that pattern in its colour-colourdiagram (Muno et al. 2002). The source is located at a distancethought to be at most 5 . = . × erg s − for a helium-rich burst (Chenevez et al.2012). Baglio et al. (2016) initially suggested that this source is anultra compact X-ray binary with a helium white dwarf companion,based on their estimation of the orbital period of this binary system,and the optical spectrum. The hydrogen rich bursts observed by NuS-TAR in 2015 (Marino et al. 2019), however, suggest that the sourceaccretes mixed H/He fuel rather than pure helium. The nature of thisbinary system is therefore needed to be further justified. The atollsource was in a hard spectral state during an
XMM-Newton obser-vation in March (Stiele et al. 2016), and showed similar variabilitycharacteristics to the “extreme island” state in its power densityspectrum (Wijnands et al. 2017). Nearly a month later, the source,with slightly higher accretion rate, was identified being in the islandstate during the second
XMM-Newton observation.In this paper, we report timing analysis of
XMM-Newton datafrom 1RXS J180408.9 − ∼ The transient X-ray binary 1RXS J180408.9 − XMM-Newton made two follow-up observations during thisperiod. The observations on 2015 March 6 (Obs. ID: 0741620101)and 2015 April 1 (Obs. ID: 0741620201) each have 57 ,
000 s on-time, from which we report our analysis in this paper. We used datataken from the timing mode of EPIC/pn (Strüder et al. 2001) andadded the barycenter correction for all events by applying the task barycorr with the Science Analysis System (SAS) version .We followed the standard filtering of events for EPIC/pn, selectingsingle and double events. We modelled a Gaussian function to thedistribution of photons over the CCD and included the photonswithin 3 𝜎 as coming from the source, which accordingly coversthe CCD columns 19 ≤ RAWX ≤
61 and 27 ≤ RAWX ≤
52 forthe first and the second observations, respectively. A RAWX ≤ . MNRAS000
52 forthe first and the second observations, respectively. A RAWX ≤ . MNRAS000 , 1–6 (2020)
POs in 1RXS J180408.9 − Figure 2.
Two Fourier power spectra from the first and second 4 ,
000 sintervals indicated with numbers in Figure 1. The dashed and dotted linesshow the fitted red noise continuum, and 99 % confidence level based on thenull hypothesis under the red noise background respectively. Strong peaksbetween 6–8 mHz appear on both spectra.
In order to explore the possible frequency drift of the QPOs, weapplied Lomb-Scargle periodograms (Lomb 1976; Scargle 1982),with 2 ,
000 s sliding windows and 20 s for each forwarding step, tocompute dynamical power spectra for all the burst-free segments.We searched the range 5 to 9 mHz and adopted the normalizationfrom Press et al. (2007) for the power spectra.
The resulting light curves from the both observations show regularType I X-ray bursts (Figure 1). The burst rate in the second obser-vation period is about twice as high as during the first period, andshows an enhanced persistent luminosity. All bursts have a durationof ∼
150 s, suggesting that they have a hydrogen-rich compositionduring burst. We identified candidate signals in two intervals, as in-dicated in Figure 1, labeled 1 and 2. They show strong peaks in theirpower spectra at frequencies of ≈ . . ,
000 s continuous interval. Following the pro-cedure from Vaughan (2005) to test for the presence of periodicitywith limited observational data against a red noise background, wefirst fit the two periodograms with a power-law model 𝑃 𝑖 = 𝐴 𝑓 − 𝛼𝑖 with least squares regression to estimate the red noise continuumin the power spectra. The frequency range of interest, 4 to 10 mHz,was ignored for the fit, such that the model is independent of thesignals in this range of frequency. The resulting exponents, 𝛼 , are0 . ± .
06 and 0 . ± .
06 for the first and second intervals, respec-tively. For a given modeled noise power 𝐿 𝑖 , we determined a 99 %confidence level based on the fact that the ratio of twice the signalpower to the noise power, 2 𝑃 𝑖 / 𝐿 𝑖 , is distributed as a 𝜒 nature over the periodograms as well as taking the number of trials intoaccount. Both of the peaks we detected exceed the 99 % confidencelevel (Figure 2). We did not detect any periodic signals from thesecond observation where the 95 % confidence upper limit on thefractional RMS amplitude in the range 4 to 10 mHz is < .
62 %.The dynamical power spectra of the six burst-free segmentsfrom the first observation show that the oscillations are time-dependent in both frequency and amplitude (Figure 3). Other panelsalso show high power signals in the periodgrams, but these are rathersporadic and only last for brief times. To further investigated how themHz QPOs correlate to the bursts, we have computed power spec-tra for light curve segments before, during, and after the flashes.Figure 4 shows one example for the disappearance of mHz QPOsafter the onset of the first Type I X-ray burst in the first observation.Following the same normalization as above, the ∼ ,
000 s segments labeled with num-bers in Figure 1, in which the detected oscillations have relativelystable frequencies and last for longer duration, to study the energydependence on the oscillation amplitude. The events from the twosegments were divided into 9 energy bands, from 0 . .
57 mHz and 7 .
73 mHz from the result of the Fourier analysis re-spectively, to produce oscillation profiles and derived the fractionalRMS amplitudes for the two segments. The amplitude increasestowards lower energy ( (cid:46) . 𝑦 -axis (Fig-ure 6). For comparison, we phase-folded ≈ ,
000 s of simulationlight curve data with an oscillation frequency of ≈ .
37 mHz (localframe) and took the average luminosity for each bin ( right axis ).The luminosity is corrected by the assumed surface red shift, i.e., 𝐿 ∞ = 𝐿 /( + 𝑧 ) , where 1 + 𝑧 = / √︁ − 𝐺 𝑀 / 𝑐 𝑅 ≈ .
26 for1 . (cid:12) and 11 . ≈ . × erg s − for 0 . (cid:164) 𝑀 Edd local accretionrate with a 0 .
759 hydrogen mass fraction. For clarity, both 𝑦 -axesoffsets and scaling are adjusted arbitrarily for matching. Both asym-metric shapes of the oscillations show a steeper rise than decay. Thefractional RMS amplitudes of the observed and the simulated are ≈ . .
65 %, respectively. This discrepancy is discussed inthe following section. − −
536 ,4U 1608 −
52 and Aql X-1 (Revnivtsev et al. 2001), 4U 1323 − − − −
676 (Mancuso et al. 2019).Except IGR J00291+5934 (Ferrigno et al. 2017), the presence ofType I bursts shows direct evidence to support the mHz QPOs beingas a result of marginally stable burning. Ferrigno et al. (2017) re-
MNRAS , 1–6 (2020)
K. Tse et al.
Table 1.
Key figures and properties of mHz QPOs from different sources.Source Accretion luminosity(10 erg s − ) QPOs frequency ( mHz ) R.M.S ( % ) QPO Disappearafter bursts a References4U 1636 −
536 [0.6–3 . ] ( −
150 keV ) . ] ( . − ) Y(D) Revnivtsev et al. (2001)Altamirano et al. (2008)4U 1608 − − ∼ . . ] ( −
20 keV ) − ∼ ∼ . ] ( − ) − Y − Revnivtsev et al. (2001)4U 1323 −
619 0 . ( −
25 keV ) b . . ( −
20 keV ) Y Strohmayer & Smith (2011)IGR J17480 − ] ( −
50 keV ) . ] ( −
60 keV ) ? Linares et al. (2012)GS 1826 − ∼ ( . − ) ∼ . ] ( . − . ) − Strohmayer et al. (2018)EXO 0748 −
676 [ ∼ . . ] ( −
50 keV ) ∼ ∼ ( − ) Y(D) Mancuso et al. (2019)1RXS J180408.9 − . ( . −
50 keV ) c ∼ . ( . −
10 keV ) Y This work a D = Frequency of the oscillations systematically drops before being undetectable in some cases. Indeterminate for IGR J17480-2446 as bursts and QPOsappear in separate observations. b Persistent flux obtained from Galloway et al. (2020), assuming the source distance to be 4 . c Persistent luminosity estimated by Ludlam et al. (2016) with the assumed source distance = . Figure 3.
Dynamical power spectra showing the evolution of the mHz QPOsfor all burst-free segments from the first observation using Lomb-Scargleperiodograms with the normalization of (Press et al. 2007), 2 ,
000 s slidingwindows, and 20 s steps. The end of each PDS is within a step of the onset ofa burst. For clarity, we plot only power ≥
25 to filter out noise. The 𝑥 -axisshows the start time of the current 2 ,
000 s spectrum from the beginningof each segment. The common color bar shown on the right hand siderepresents the signal power for all panels. ported that the mHz signal found in IGR J00291+5934 may be trig-gered by the so-called "heartbeat model" (Altamirano et al. 2011)which is associated with the movement of the inner accretion disk.The correlated increases in the blackbody radius and the normal-ization of the Comptonization component, along with QPO flux intheir QPO phase-resolved spectrum analysis support this attribu-tion. Moreover, the lack of Type I X-ray bursts during when theQPOs occur, and the very high RMS amplitude of QPOs ( ≥
30 %)also contra-indicate thermonuclear burning. We summarized thekey figures and properties of the marginally stable nuclear burningfrom the eight sources, excluding IGR J00291+5934, in Table 1. Itincludes the accretion luminosity for the occurrence of mHz QPOs,the discovered frequency range and fractional RMS amplitude ofthe oscillations, as well as whether the signal disappears after bursts.In the first
XMM-Newton observation of1RXS J180408.9 − (cid:46) .
4) atpersistent luminosity = . × erg s − (Ludlam et al. 2016).The source had transited to hard/island state, accompanied with ≈ . XMM-Newton observation. Withthis moderate change of accretion rate, the mHz oscillationsbecame undetectable, along with the power-law noise in the lowfrequency range. In contrast, in the previous detections of mHzQPOs, 4U 1636 − − − −
676 were reported to be in a relatively soft state,unlike 1RXS J180408.9 − −
619 (Strohmayer & Smith 2011), 4U 1636 − − − − MNRAS000
619 (Strohmayer & Smith 2011), 4U 1636 − − − − MNRAS000 , 1–6 (2020)
POs in 1RXS J180408.9 − Figure 4.
Three pairs of light curve and the corresponding power spectrumfor selected time ranges. The left panel, from top to bottom, shows three1 ,
400 s windows in dark color before, during and after the first observedType I X-ray burst from the first observation, respectively. A strong peakstands out at ∼ of burst rate. Furthermore, the accretion luminosity at which themHz QPOs were observed is ∼ erg s − , which is closer to thetheoretical value ( ∼ (cid:164) 𝑀 Edd by, e.g., Bildsten 1998) compared to thereported range of luminosity ∼ –10 erg s − in other sources.Moreover, the presence of bursts and mHz QPOs are in separateobservations with distinct luminosities, unlike in the more rapidly-spinning sources, where the bursts and oscillations occur within thesame observations.Independent of the variations in mHz QPO properties, such asthe accretion rate for the oscillations to exist as well as whether theydisappear after bursts (refer to Table 1), remains some uncertaintiesas to whether the involved nuclear processes in different sourcesare the same. In fact, the actual nuclear burning depends on variousconditions such as fuel compositions, crustal heating, and neutronstar rotation rate, etc. Strohmayer et al. (2018) pointed out thatfast rotating NS has a significant difference of surface gravity fromthe pole and equator. For example, the 582 Hz rotating NS in theLMXB 4U 1636 −
536 has surface gravity at the pole ≈
11 % higherthan at the equator. This variation may lead to different burningregimes across the entire NS surface as they are sensitive to thesurface gravity (Heger et al. 2007). Besides, Galloway et al. (2018)provided additional evidence in support of the influence of NS spinon the burning regime. They found that the regime of maximumburst rate is at a lower accretion rate if the NS has a higher spinrate. Once the burst rate reaches a maximum, it starts to decreasewith further increases of accretion rate, supposedly because of theenhanced stabilization of nuclear burning (van Paradijs et al. 1988).Thus, the occurrence of mHz QPOs, triggered by marginally stableburning, shifts to a lower accretion rate for a NS with faster rotation.Interestingly, IGR J17480 − ,
000 s interval from the first segment) and the simulation
Figure 5.
Fractional RMS amplitudes of mHz QPOs over energy bandsfrom 0 . blue circle ) and thesecond ( orange square ) 4 ,
000 s continuous intervals in Figure 1 into 9 energybands and folded them with their corresponding frequency, 6 .
57 mHz and7 .
73 mHz respectively (see Figure 2). Results from the two intervals showincreases towards lower energy ( (cid:46) result from Heger et al. (2007). Both results show asymmetric os-cillation profiles with a steeper rise than decay. We speculate thatthe shape of the oscillations is mainly dependent on the burningfuel compositions. On the other hand, their fractional RMS ampli-tudes are distinctly different, ≈ . ≈ .
65 % from the simulation. For gravitational energy releasedfrom accretion ≈
200 MeV comparing to the energy released fromthermonuclear fusion ≈ / = . ∼ . ACKNOWLEDGEMENTS
Based on observations obtained with
XMM-Newton , an ESA sci-ence mission with instruments and contributions directly fundedby ESA Member States and NASA. This work was supported inpart by the National Science Foundation under Grant No. PHY-1430152 (JINA Center for the Evolution of the Elements). Parts ofthis research were conducted by the Australian Research CouncilCentre of Excellence for Gravitational Wave Discovery (OzGrav),through project number CE170100004. AH has been supported, inpart by the Australian Research Council Centre of Excellence for AllSky Astrophysics in 3 Dimensions (ASTRO 3D), through projectnumber CE170100013. Y.C. and H. -E. H. especially acknowledgesthe support from Ministry of Science and Technology of Taiwanthrough grant MOST 109-2112-M-008-004-.
MNRAS , 1–6 (2020)
K. Tse et al.
Figure 6.
A comparison between the mHz QPOs profiles of1RXS J180408.9 − ,
000 s interval in Figure 1 were folded with a frequency 6 .
57 mHz foundin Figure 2. A ∼ ,
000 s light curve is extracted from simulation (Hegeret al. 2007) and folded with a local oscillation frequency ≈ .
37 mHz. Theaverage luminosity from the folded light curve is corrected by surface redshift, and offset by a persistent value (plotted referenced to the right axis).Both 𝑦 -axes offsets and scaling are adjusted arbitrarily for matching. Resultsfrom the observed and simulation similarly show a shape with a steep riseand shallower decline, while the oscillation amplitudes are notably different. DATA AVAILABILITY
The
XMM-Newton data underlying this article are available inthe High Energy Astrophysics Science Archive Research Center(HEASARC), at ( https://heasarc.gsfc.nasa.gov ). REFERENCES
Altamirano D., van der Klis M., Wijnands R., Cumming A., 2008, ApJ, 673,L35Altamirano D., et al., 2011, ApJ, 742, L17Baglio M. C., D’Avanzo P., Campana S., Goldoni P., Masetti N., Muñoz-Darias T., Patiño-Álvarez V., Chavushyan V., 2016, A&A, 587, A102Bildsten L., 1998, in Buccheri R., van Paradijs J., Alpar A., eds, NATOAdvanced Science Institutes (ASI) Series C Vol. 515, NATO AdvancedScience Institutes (ASI) Series C. p. 419 ( arXiv:astro-ph/9709094 )Chenevez J., et al., 2012, The Astronomer’s Telegram, 4050, 1Degenaar N., et al., 2015, The Astronomer’s Telegram, 7352, 1Ferrigno C., et al., 2017, MNRAS, 466, 3450Galloway D. K., Keek L., 2017, arXiv e-prints, p. arXiv:1712.06227Galloway D. K., et al., 2018, ApJ, 857, L24Galloway D. K., et al., 2020, arXiv e-prints, p. arXiv:2003.00685Gambino A. F., et al., 2016, A&A, 589, A34Heger A., Cumming A., Woosley S. E., 2007, ApJ, 665, 1311Huppenkothen D., et al., 2019, ApJ, 881, 39Krimm H. A., et al., 2015, The Astronomer’s Telegram, 6997, 1Leahy D. A., Darbro W., Elsner R. F., Weisskopf M. C., Sutherland P. G.,Kahn S., Grindlay J. E., 1983, ApJ, 266, 160Linares M., Altamirano D., Chakrabarty D., Cumming A., Keek L., 2012,ApJ, 748, 82Lomb N. R., 1976, Ap&SS, 39, 447Ludlam R. M., et al., 2016, ApJ, 824, 37Lyu M., Méndez M., Altamirano D., Zhang G., 2016, MNRAS, 463, 2358Mancuso G. C., Altamirano D., García F., Lyu M., Méndez M., Combi J. A.,Díaz-Trigo M., in’t Zand J. J. M., 2019, MNRAS, 486, L74Marino A., et al., 2019, MNRAS, 490, 2300Muno M. P., Remillard R. A., Chakrabarty D., 2002, ApJ, 568, L35Paczynski B., 1983, ApJ, 264, 282Press W. H., Teukolsky S. A., Vetterling W. T., Flannery B. P., 2007,Numerical Recipes 3rd Edition: The Art of Scientific Comput-ing, 3 edn. Cambridge University Press,
Revnivtsev M., Churazov E., Gilfanov M., Sunyaev R., 2001, A&A, 372,138Scargle J. D., 1982, ApJ, 263, 835Stiele H., Yu W., Kong A. K. H., 2016, ApJ, 831, 34Strohmayer T., Bildsten L., 2003, arXiv e-prints, pp astro–ph/0301544Strohmayer T. E., Smith E. A., 2011, The Astronomer’s Telegram, 3258, 1Strohmayer T. E., et al., 2018, ApJ, 865, 63Strüder L., et al., 2001, A&A, 365, L18Vaughan S., 2005, A&A, 431, 391Weaver T. A., Zimmerman G. B., Woosley S. E., 1978, ApJ, 225, 1021Wijnands R., Parikh A. S., Altamirano D., Homan J., Degenaar N., 2017,MNRAS, 472, 559Woosley S. E., et al., 2004, ApJS, 151, 75van Paradijs J., Penninx W., Lewin W. H. G., 1988, MNRAS, 233, 437This paper has been typeset from a TEX/L A TEX file prepared by the author.MNRAS000