High-cadence observations and variable spin behaviour of magnetar Swift J1818.0-1607 after its outburst
David Champion, Ismael Cognard, Marilyn Cruces, Gregory Desvignes, Fabian Jankowski, Ramesh Karuppusamy, Michael J. Keith, Chryssa Kouveliotou, Michael Kramer, Kuo Liu, Andrew G. Lyne, Mitchell B. Mickaliger, Brendan O'Connor, Aditya Parthasarathy, Nataliya Porayko, Kaustubh Rajwade, Ben W. Stappers, Pablo Torne, Alexander J. van der Horst, Patrick Weltevrede
MMNRAS , 1–15 (2020) Preprint 9 September 2020 Compiled using MNRAS L A TEX style file v3.0
High-cadence observations and variable spin behaviour ofmagnetar Swift J1818.0 − David Champion , Ismael Cognard , , Marilyn Cruces , Gregory Desvignes , ,Fabian Jankowski , Ramesh Karuppusamy , Michael J. Keith (cid:63) , Chryssa Kouveliotou , ,Michael Kramer , , Kuo Liu † , Andrew G. Lyne , Mitchell B. Mickaliger ,Brendan O’Connor , , Aditya Parthasarathy , Nataliya Porayko , Kaustubh Rajwade ,Ben W. Stappers , Pablo Torne , , Alexander J. van der Horst , Patrick Weltevrede Max Planck Institute for Radio Astronomy, Auf dem H¨ugel 69, Bonn D-53121, Germany Station de Radioastronomie de Nan ˜A˘gay, Observatoire de Paris, CNRS/INSU, Universit ˜Al’ d’Orl ˜Al’ans, 18330, Nan ˜A˘gay, France Laboratoire de Physique et Chimie de l’Environnement, CNRS, 3A Avenue de la Recherche Scientifique, 45071, Orl ˜Al’ans Cedex 2, France LESIA, Observatoire de Paris, Universit ˜Al’ PSL, CNRS, Sorbonne Universit ˜Al’, Universit ˜Al’ de Paris, 5 Place Jules Janssen, 92195, Meudon, France Jodrell Bank Centre for Astrophysics, University of Manchester, M13 9PL Manchester, UK Department of Physics, the George Washington University, 725 21st Street NW, Washington, DC 20052, USA Astronomy, Physics, and Statistics Institute of Sciences (APSIS), The George Washington University, Washington, DC 20052, USA Instituto de Radioastronom´ıa Milim´etrica (IRAM), Avda. Divina Pastora 7, Local 20, 18012 Granada, Spain East Asian Observatory, 660 N. Aˆa ˘A´Zohoku Place, University Park, Hilo, Hawaii 96720, USA
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
We report on multi-frequency radio observations of the new magnetar SwiftJ1818.0 − − − Key words: stars: magnetars - stars: pulsars: individual: Swift J1818.0 − (cid:63) Contact author: [email protected] † Contact author: [email protected]
Among the population of young Galactic neutron stars, mag-netars form their own class (for an overview, see Kaspi & Be-loborodov 2017). Most of them are persistent X-ray sources,however, several have been also observed at optical, IR, and © a r X i v : . [ a s t r o - ph . H E ] S e p Champion et al. radio wavelengths. Their most characteristic attributes arethe emission of repeated, very short ( ∼ s of ms) hardX-ray bursts during randomly occurring outbursts, and anX-ray luminosity higher than the spin-down luminosity. Ofthe ∼ magnetars known to date, only two have been lo-cated outside the Milky Way: one each in the Large andSmall Magellanic Clouds. Their periods, P , range between ∼ and their periodderivatives, (cid:219) P , imply magnetic field strengths, computed via B = . × √ P (cid:219) P , typically of the order of − Gauss(Kouveliotou et al. 1998). Magnetar emission is thus be-lieved to be powered by their strong magnetic fields causingneutron star quakes and magnetospheric phenomena, givingrise to the designation of these objects (Duncan & Thomp-son 1992). Using the expression for the “characteristic age”derived for radio pulsars, τ = P / (cid:219) P , one finds that magne-tars are expected to be young, with typical ages of a fewhundred to a few thousand years. Consequently, magnetarsoccupy the upper right corner of the “ P − (cid:219) P diagram”, relativeto radio pulsars. We refer to recent reviews for a detailed dis-cussion of magnetar properties (e.g. Kaspi & Beloborodov2017) or their relationship to pulsars and other neutron starpopulations (e.g. Kaspi & Kramer 2016).Only a handful of magnetars have been detected at ra-dio frequencies. The first radio detection of XTE J1810 − − − − −
197 recently switched onas a radio source again ten years after the cease of its radioemission in 2008 (Levin et al. 2019; Dai et al. 2019).The individual pulses of magnetars are also commonlyvery narrow (‘spiky’), spreading over an often wide pulse window (Camilo et al. 2007b; Kramer et al. 2007; Torneet al. 2015), and longitude-resolved modulation indices re-veal a high degree of intensity fluctuations on day-to-daytimescales and dramatic changes across pulse phase (Sery-lak et al. 2009). This variability in the individual pulses isalso reflected in the varying shapes of average pulse pro-files (Camilo et al. 2007b; Levin et al. 2019). The ‘spikiness’combined with the large variability, and the high degree ofpolarisation, resembles some of the properties of repeatingFast Radio Bursts (FRBs) (Spitler et al. 2016), and so itis not surprising that magnetars are among the models toexplain FRBs (e.g. Lyubarsky 2014; Maan et al. 2019).Unlike normal pulsars, the radio emission of magnetarstypically shows a flat spectrum (Camilo et al. 2006; Sery-lak et al. 2009; Torne et al. 2015; Dai et al. 2019), withvariation in the single pulse spectra (Serylak et al. 2009;Torne et al. 2015) that is larger than that for normal pulsars(e.g. Kramer et al. 2003). The flatness of the radio spectrumhas led to the detection of radio-loud magnetars up to a fre-quency of ∼
300 GHz (Torne et al. 2017, 2020), which is thehighest radio frequency of any neutron star detection so far.It is intriguing to contrast the radio emission of magne-tars with that of ‘normal’ radio pulsars. Overall, as the abovebrief description shows, one finds that in several respectsmagnetar emission shows similarities to the emission prop-erties of normal radio pulsars while simultaneously showingstriking differences (Kramer et al. 2007). Understanding theextent, and potentially the origin, of these differences andsimilarities promises to help solve the radio emission mys-tery of radio emitting neutron stars as a whole. Extendingthese studies with additional, new radio-loud magnetars istherefore extremely useful.Determining the relationship between rotation andmagnetic powered neutron stars is also important in un-derstanding the formation of magnetars and population ofneutron stars as a whole. As pointed out by Keane & Kramer(2008), the neutron star birthrate and population estimatesare not consistent with the Galactic supernova rate. Thisproblem would be alleviated if one considers a possibleevolutionary scenario between some of the known neutronstar classes. The possibility of such a scenario was demon-strated by Espinoza et al. (2011), who pointed out that PSRJ1734 − P = . sand slowing down with a period derivative (cid:219) P = . × − ,is located in the ‘ P - (cid:219) P -diagram’ midway between those ofnormal rotation-powered pulsars and magnetars. In case ofan unchanged braking index, this pulsar may soon have therotational properties of a magnetar. The existence of suchan evolutionary channel is supported by a few other casessuch as PSR J1119 − − − − MNRAS , 1–15 (2020) igh-cadence observations of Swift J1818.0 − known magnetars . About 35 hours after the X-ray burst,observations with the 100-m Effelsberg telescope discoveredradio pulsations from Swift J1818.0 − − (cid:219) P = . ( )× − and, hence, a derived characteristic age of 265(1) yrs assum-ing a constant (since birth) braking index of 3 and a birthperiod much less than the current period. This measuredspin-down also implies a characteristic surface dipole mag-netic field of B = . × G and a spin-down luminosityof (cid:219) E = . × erg s − (assuming a moment of inertia of kg m ). These values confirmed the magnetar natureof the detected source. As pointed out by Champion et al.(2020), the characteristic age is very similar to that of SGR1806 −
20 which currently has the smallest characteristic ageon record (240 years; Olausen & Kaspi 2014). The measure-ment of spin-down rate was later confirmed by X-ray timing(Hu et al. 2020; Esposito et al. 2020) which also revealed anearly timing anomaly (Hu et al. 2020).In the following, we present continued, extensive radioobservations of Swift J1818.0 − The results reported here come from observations with theEffelsberg 100-m, the Lovell 76-m and the Nan¸cay radio tele-scopes and are summarised in Table 1. While observationscontinued to the time of writing, this work covers the firstmonth since the magnetar’s outburst, i.e. from 2020 March14 to 2020 April 15.At Effelsberg, Swift J1818.0 − The magnetar-like pulsar PSR J1846 − processing for the measurements at 1.37, 2.55 and 4.85 GHzinvolved real-time coherent dedispersion followed by averag-ing in to sub-integrations of 10 s duration starting from thesecond session using the PSRIX pulsar backend (Lazaruset al. 2016). Where possible, additional baseband data wererecorded with 320, 400 or 500 MHz bandwidth for the threedifferent receivers used. These data were processed offlineto generate both single pulses and 10 s sub-integrations af-ter coherent dedispersion and folding. We used dspsr (vanStraten & Bailes 2011) to coherently dedisperse and fold thedata in both real-time and offline processing pipelines. Ob-servations at 6 GHz with the wide band receiver covering 4-8GHz are recorded in search mode with a time and frequencyresolution of 131 µ s and ∼ MHz, respectively (see e.g.Desvignes et al. 2018). In each session, we recorded a 1 Hzswitched noise diode for 90–120 s to allow for polarisationcalibration. The calibration was conducted for frequency-dependent gain and phase variations between the signals re-ceived by the pulsar instrument from the two orthogonalprobes of the receiver. Subsequently, Faraday rotation wasmeasured and corrected to reveal the polarisation profile. Inaddition, the data were manually checked to remove appar-ent radio interference (RFI). The software package psrchive (van Straten et al. 2012) was used for most of the data post-processing.At JBO, the source was observed with the Lovell Tele-scope at a centre frequency of 1.53 GHz, with 512 MHzof bandwidth divided into 1532 channels, over four days,starting on 2020 March 14. Complex voltage data from thetelescope were converted into Stokes parameters using aROACH-1 Field Programmable Gate Array board. Then,they were processed by the DFB backend (Manchester et al.2013), folding the data modulo the topocentric period anddedispersing them at the dispersion measure (DM) reportedin Karuppusamy et al. (2020). The resulting time-frequencydata were folded using the best ephemeris into multiple 8-second sub-integrations and saved to disk. The data weresubsequently visually inspected and manually cleaned to re-move strong radio frequency interference (RFI) using the psrchive package, leaving approximately 250 MHz of usablebandwidth. To calibrate the polarisation of the JBO data, weused RFI-cleaned datasets of a bright pulsar with well knownpolarization properties that was observed close in time toeach of our magnetar observations as a template for the po-larization calibration. We generated the receiver solutionsfor the Lovell Telescope signal chain using the MeasurementEquation Template Matching (METM) method (van Straten2013) by solving for the receiver solution of the bright pul-sar data, namely PSR B0919+06 using a fully calibrateddata set of the same pulsar obtained from the EPN pulsardatabase (Johnston & Kerr 2018). Under the assumptionthat there is no significant change in the signal chains wethen use the receiver solutions to calibrate the magnetar ob-servations. We fit for the differential gain and phase as afunction of time to obtain the best fit polynomial for thereceiver solution using the publicly available software
PSR-SALSA (Weltevrede 2016). Finally, we corrected the PA forFaraday rotation using the value of RM obtained from theEffelsberg data. These have led high consistency with the Ef-felsberg polarisation properties (see Section 3.3.3 for moredetails).At the NRT, the source was observed on seven epochs
MNRAS , 1–15 (2020)
Champion et al. from 2020 March 28 to 2020 April 15 at a central frequencyof 1.48 GHz. The Nan¸cay Ultimate Pulsar Processing In-strument (NUPPI) backend was used to record the data inPSRFITS search mode format, with 512 MHz bandwidthand 4-MHz frequency resolution. The data were then foldedoffline using the pulsar ephemeris determined from obser-vations before 2020 March 28, into 10-s sub-integrations.Each sub-integration was visually inspected and manuallycleaned to remove RFI using the psrchive software package,which leaves an overall effective bandwidth of approximately420 MHz. Preliminary results of the first week of the observations re-ported here were already presented in Karuppusamy et al.(2020), Rajwade et al. (2020) and Champion et al. (2020).Here we provide further data to conduct a comprehen-sive study of the spin and emission properties of SwiftJ1818.0 − In the report of the discovery of radio emission, Karup-pusamy et al. (2020) presented a DM of ( ) cm − pc anda rotation measure (RM) of + ± rad m − , based on Ef-felsberg observations at 1.37 GHz. The same day, Rajwadeet al. (2020) measured DM = ± cm − pc from JBO ob-servations at 1.53 GHz. A few days later, Champion et al.(2020) presented a refined DM value of ( ) cm − pc, basedon the early observations from both telescopes.Lower et al. (2020b) determined a DM of 699.5(3)cm − pc, using the MeerKAT telescope at a central frequencyof 1.28 GHz, over a 856 MHz bandwidth. This analysis usedscatter-broadened templates to account for scattering at thelower edge of the band, and reported a characteristic scat-tering time of τ SC , = ( ) ms at 1 GHz. Later obser-vations with the Parkes telescope derived a similar value of τ SC , = + − ms, with a scattering index of α SC = − . + . − . (Lower et al. 2020a). They also infer a mean DM of . ( ) cm − pc and RM = . ( ) rad m − from observationsacross the 0.8 to 4.0 GHz band.As indicated by the analysis undertaken by Lower et al.(2020b) the measured DM value may be affected by the pres-ence of interstellar scattering effects. Hence, we have used acode to model the DM and scattering at the same time, whilerepresenting the intrinsic pulse profile with Shapelets, sim-ilarly to Lentati et al. (2017). We divided the available 512MHz-bandwidth NRT data into four subbands and by apply-ing the Shapelet model we obtained a DM value of . ( ) pc cm − and a scattering timescale of τ SC , = ( ) ms(assuming the scattering index α SC as determined by Loweret al. 2020a). Although our scattering timescale agrees withthe result reported by Lower et al. (2020a), we find that ournew DM value is significantly smaller, closer to the value reported by Lower et al. (2020b) using MeerKAT, and con-sistent with the value by Champion et al. (2020). We believeour estimate of the DM is robust and attribute the varia-tion in the reported DM values to different (or absence of)methods of accounting for the bias in DM estimation fromscattering, rather than intrinsic variation in the DM. TheseDM values correspond to a distance in the range from 4.8 to8.1 kpc, as determined using the YMW16 (Yao et al. 2017)and NE2001 (Cordes & Lazio 2002) electron density models,respectively, and such a large inferred distance is consistentwith the observed interstellar scattering. Each observation described in Section 2 was averaged in fre-quency and time to form sub-integrations of approximately200s in length. As reported in Section 3.3.1, the radio emis-sion from Swift J1818.0 − psrchive and fittingwas carried out using a Markov-chain Monte-Carlo (MCMC)approach with emcee (Foreman-Mackey et al. 2013).The spin frequency of the magnetar evolves significantlyover time, changing from ν = . to . overour observing span of 32 days, resulting in a mean frequencyderivative (cid:219) ν = − . × − Hz , equivalent to a period deriva-tive (cid:219) P = . × − . However, it is immediately apparentthat the spin-down rate also varies significantly over the ob-serving span, ranging between − and − × − Hz whenaveraged over several day spans, and clear signs are seenof discrete timing events in the data potentially associatedwith step changes in spin parameters. A thorough timinganalysis was performed using a pipeline based on tempo2 (Lentati et al. 2014) and enterprise (Ellis et al. 2019)software packages to perform a Bayesian timing analysis,with sampling performed using emcee . As with other mag-netars, Swift J1818.0 − α , and log-amplitude, log ( A red ) , of the power-law Gaussian process such that thepower-spectral density of the red noise is given by P ( f ) = A π (cid:18) f yr − (cid:19) − α yr . MNRAS , 1–15 (2020) igh-cadence observations of Swift J1818.0 − Table 1.
Details of the observations carried out during the first month after the magnetar activation.
UTC start Duration (min) Telescope Centre Frequency (GHz) Bandwidth (MHz) Details
In order to explore the evidence for discrete timing events inthe data, we developed a large number of competing mod-els with between 0 and 4 discrete state switches, where thestate switches may include step changes in ν , (cid:219) ν and/or (cid:220) ν . Theepochs of the transitions between states had uniform priorswithin large, non-overlapping windows around the suspectedevents determined by eye. We also explored models includ-ing a fixed value of (cid:220) ν over the observing span, but withoutstep changes in (cid:219) ν . Model selection was performed using atrans-dimensional MCMC approach with enterprise , andverified by direct computation of evidence using tempon-est . The preferred models have 4 discrete changes in (cid:219) ν withor without a step change in ν at the first transition, and allother models were rejected with an odds ratio of less than0.01. Table 2 shows the results for the model both with andwithout a step change in ν at the first transition, and boththe pre-fit and post-fit residuals for the maximum-likelihood model are shown in Figure 1. The parameters for the com-plete maximum-likelihood timing model are given in Table3. Regardless of the choice of model, the period evolution iswell constrained, with only subtle differences in how much ofthe frequency evolution is attributed to “noise” versus “de-terministic” changes. Although ‘by eye’ there seems to be asuggestion of cubic behaviour in the residuals, this is likelya result of our subtraction of a quadratic model from thedata and the algorithms do not find sufficient evidence tosupport including the (cid:220) ν term in the model.The first timing event at MJD 58928 (2020 March 20)stands out from the others as modelling it also requires asignificant change in spin-frequency, ∆ ν = . ( ) µ Hz. Thisvalue is somewhat smaller but largely consistent with thevalue obtained from NICER observations around the sameday (Hu et al. 2020), although the change in spin-down rate
MNRAS000
MNRAS000 , 1–15 (2020)
Champion et al.
Table 2.
Best-fit values from the MCMC analysis. Upper panelshows results from a model with a step change in ν at the firstepoch. Lower panel is the results without a step change in ν .The last column gives the inferred characteristic age. Values arequoted as the mean of the posterior with a 1- σ range in the lastdigit given in parenthesis.Epoch Start ∆ ν ∆ (cid:219) ν (cid:219) ν AgeMJD ( µ Hz ) ( µ Hz ) ( µ Hz ) (yr)0 58922.4 – −
45 2601 58928.3(1) 1.2(5) 19(5) −
25 4602 58930.9(7) − −
53 2203 58935.9(5) 44(8) − − −
26 4400 58922.4 – −
44 2601 58927.5(4) 21(5) −
23 5102 58930.8(7) − −
53 2203 58935.9(5) 45(9) − − −
25 470 determined from our dense radio observations is a few timeslarger than they reported.Even when removing the deterministic spin-downmodel, there is a strong red-noise process remaining in theresiduals. From the best-fitting model with 4 discrete stateswitches we estimate α = . ( ) and log ( A red ) = − . ( ) which implies spin variations 3 orders of magnitude largerthan the long-term red noise processes seen in normal pul-sars (Parthasarathy et al. 2019), or alternatively that SwiftJ1818.0 − ν and (cid:219) ν changes as α = . ( ) and log ( A red ) = − . ( ) , althoughthis model is disfavoured in our analysis.The frequency evolution, ν ( t ) can be investigated by an-alytically taking the derivative of the Fourier-basis Gaus-sian process and adding this to the deterministic spin-downmodel. Determining (cid:219) ν ( t ) is more difficult as the power-lawred noise has a spectral index less than 4, so the secondderivative of the Gaussian process is not a smooth process.Therefore, we approximate the derivative of ν ( t ) by comput-ing the average gradient of ν ( t ) over windows of one day.This process was repeated on 256 samples from the Markov-chain to give an impression of the allowed range of valuesfrom our model. These results are shown in Figure 2, over-laid with the epochs of our observations. In particular, thefigure’s lower panel shows the variation of the characteristicage as determined at a given epoch, demonstrating the im-practicality of attempting to measure the characteristic ageof a magnetar from short term period evolution. Radio-loud magnetars are known to show drastic change intheir radio pulse profile within a few weeks from the onsetof the X-ray outburst (e.g. Camilo et al. 2007a; Levin et al.2019), and this is no different for Swift J1818.0 − Table 3.
Maximum likelihood timing model. Parameters wereeither fixed (F), solved with linear least-squares (L) or from theMCMC analysis (M). Parameters fitted with least squares aregiven with an error in the last digit, uncertainties on other valuesshould be taken from Table 2. The parameters are as reported by tempo2 in Barycentric Coordinate Time (TCB) units. A red and α red are the amplitude at yr − and index of the power-law rednoise model as defined in the text.Parameter Value TypeRA (J2000). . . . : : . FDec (J2000) . . . − : : . F ν ( t ) . . . . . . . . . . . . ( ) Hz L t (MJD) . . . . . . . F (cid:219) ν . . . . . . . . . . . . . . − . ( ) × − Hz LDM . . . . . . . . . . . . − pc FEpoch (MJD) 58928.30 MEpoch (MJD) 58931.00 MEpoch (MJD) 58936.05 MEpoch (MJD) 58947.22 M ∆ ν . . . . . . . . . . . . . µ Hz M ∆ (cid:219) ν . . . . . . . . . . . . . µ Hz M ∆ (cid:219) ν . . . . . . . . . . . . − . µ Hz M ∆ (cid:219) ν . . . . . . . . . . . . . µ Hz M ∆ (cid:219) ν . . . . . . . . . . . . − . µ Hz M log ( A red ) . . . . − . ( ) M α red . . . . . . . . . . . . . ( ) M scopes already reported significant changes (Karuppusamyet al. 2020; Rajwade et al. 2020; Champion et al. 2020). Wehave seen profile shape changes on timescales from hours todays, while, as shown below, in general only two main typesof average pulse shapes are observed which the magnetar isswitching in between over the course of our observations.Initial observations with the Lovell Telescope on MJDs58922 (2020 March 14) and 58923 (2020 March 15) show awider profile that transitioned into a narrower profile withina span of a few hours (top panel of Figure 3) where the aver-age pulses from individual 20 s sub-integrations switch fromwider to narrower profiles between the two observations. Toquantify this transition, we fitted a Gaussian function to thepulse profile of each sub-integration. The 20-second lengthof the sub-integrations was chosen to optimise the signal-to-noise ratio (S/N) for the function fitting and to preservethe best possible time resolution to capture the transition.For each sub-integration, we measured the Full Width HalfMaximum (FWHM) of the best-fit Gaussian to the pulseprofile. Finally, we generated a histogram of the measuredwidths of all sub-integrations. We repeated the analysis ontwo epochs; one where the profile was wider and; the otherright after the transition to a narrower pulse profile. Thechange in profile is evident from the distribution of pulsewidths shown in Figure 3.Further profile changes seem to be associated with thetiming events discussed in Section 3.2. Within a couple ofdays’ window around the epoch MJD 58928 (2020 March20), i.e. the first timing event, there is a trend of profilevariation which is shown in Figure 4. While the pulsar ex-hibited a single-component structure on MJD 58925 (2020March 17), a secondary component started to be visible fromMJD 58927 (2020 March 19). The leading edge of the maincomponent also started to show up again after fading on MNRAS , 1–15 (2020) igh-cadence observations of Swift J1818.0 − R e s i d u a l ( s ) (a) 321012 R e s i d u a l ( t u r n s ) R e s i d u a l ( s ) (b) 0.100.050.000.050.100.150.20 R e s i d u a l ( t u r n s ) R e s i d u a l - M o d e l ( s ) (c) 1050510 R e s i d u a l - M o d e l ( m illi t u r n s ) Figure 1.
Timing Residuals for Swift J1818.0 − MJD 58923 (2020 March 15). The profile stayed approxi-mately the same, until MJD 58930 (2020 March 22), whenthe second component became barely visible. Then a single-component profile was seen from MJD 58936 (2020 March28) onward.The flux density of Swift J1818.0 − MNRAS000
Timing Residuals for Swift J1818.0 − MJD 58923 (2020 March 15). The profile stayed approxi-mately the same, until MJD 58930 (2020 March 22), whenthe second component became barely visible. Then a single-component profile was seen from MJD 58936 (2020 March28) onward.The flux density of Swift J1818.0 − MNRAS000 , 1–15 (2020)
Champion et al. ( m H z ) P P (a)86420 / ( H z ) (b)58925 58930 58935 58940 58945 58950MJD2.02.53.03.54.0 l o g ( c / y r ) (c) Figure 2.
Spin evolution of Swift J1818.0 − P and P ) were observed.MNRAS , 1–15 (2020) igh-cadence observations of Swift J1818.0 − Figure 3. Top Panel:
Waterfall plots of 20-second sub-integrations of 1.4 GHz Lovell data clearly showing the width ofthe pulses before and after the transition.
Bottom Panel:
Nor-malized width distribution of 20-second sub-integrations of SwiftJ1818.0 − discuss single-pulse properties next and refer to the flux den-sity spectrum in Section 3.3.4. The flux density of the magnetar is sufficiently high to en-able us to undertake a detailed study of its single pulse prop-erties. It was found that, as in other radio-loud magnetars,Swift J1818.0 − N o r m a li s e d a m p li t u d e Pulse phase
Figure 4.
Profile variations over 11 days which span 2020-03-20,the epoch of the first timing event. The amplitudes in each profilewere normalized to the peak intensity of the profile. larger at the secondary component. This behaviour was alsoseen at the other epochs when the secondary component waspresent. The overall single-pulse energy typically fluctuatesup to a factor of 5-7 times the average, with however onesingle pulse seen with energy approximately 14 times theaverage. The polarisation profile of this pulse is presentedin Figure 6. Most of the pulse energy is concentrated at thephase of the secondary component, with a high degree oflinear polarisation. Only a very weak detection can be seenat the phase of the primary component.
The first polarisation profile of Swift J1818.0 − MNRAS000
The first polarisation profile of Swift J1818.0 − MNRAS000 , 1–15 (2020) Champion et al. A r b . a m p . Pulse energy P u l s e nu m b e r Pulse phase m I Figure 5.
A single-pulse sequence from the 20200320 L-bandobservation which contains 1,000 rotations. Top right panel: Av-erage profile (solid line) normalised to its peak amplitude, and thelongitude-resolved modulation indices (points, without removal ofwhite noise). Bottom left panel: Energy of the single pulses nor-malized by the mean. Bottom right panel: Colour map of pulseintensity as function of pulse phase and pulse number.
We estimated the pulse-averaged flux densities of the Lovelland Effelsberg Telescope multi-frequency data to charac-terise the magnetar’s temporal evolution in flux density andits radio spectrum from quasi-simultaneous observations.For the data from the Lovell Telescope, we used a previ-ously determined analytical parameterisation for its system-equivalent flux density (SEFD) as a function of elevation toreference our measurements to a known flux density scale.
Figure 6.
The brightest single pulse from the 20200320 L-bandobservation. The solid, dashed and dotted line correspond to thetotal intensity, linear and circular polarisation, respectively. Thepulse energy hits approximately 14 times the average, as seen inFigure 5 (near pulse number 320).
Observations of high-DM pulsars with well-known absolutecalibrated flux densities at L-band (from Jankowski et al.2018), that were performed close in time to the magnetarobservations, were used to ascertain the fidelity of our cali-bration method. The Effelsberg data are from observationsat L, S and C-band, for which the calibration methods differ.Both the S and C-band data were referenced to an absoluteflux density scale based on observations of a primary fluxdensity calibrator, the planetary nebula NGC 7027. We em-ployed a standard parameterisation for its radio spectrum.For those data, as well as the ones from the Lovell Tele-scope, we visually identified on-pulse phase gates per epochthat included all pulse profile components and measured theband-integrated pulse-averaged flux densities after baselinesubtraction. The uncertainties were derived from the RMS ofthe off-pulse regions. The L-band measurements were basedon the radiometer equation (Dewey et al. 1985), extrapo-lated sky temperatures to the reference frequencies (Haslamet al. 1982; Lawson et al. 1987) and known receiver perfor-mance parameters . We assumed a 25 per cent uncertaintyfor those measurements.In Fig. 8, we show a timeline of our flux density mea-surements together with a selection of data points and upperlimits at various radio frequencies reported in the literature.Where literature data are available at epochs close to ourmeasurements (and frequencies), they generally agree well.The middle panel shows the same multi-epoch data set ina spectrum. Finally, the bottom panel presents the best-fitting spectral indices from fitting a simple power law ofthe form S ( ν ) ∝ ν α , where S is the pulse-averaged flux den-sity at frequency ν and α is the spectral index, to quasi-contemporaneous band-integrated data from the Effelsbergtelescope obtained at two or three frequency bands. While https://eff100mwiki.mpifr-bonn.mpg.de MNRAS , 1–15 (2020) igh-cadence observations of Swift J1818.0 − P A ( ◦ ) . . . − . . . . F l u x ( a r bun i t s ) . . . . . . . . . − P A ( ◦ ) . . . . . . . . . . . F l u x ( a r bun i t s ) . . . . . . . . . . . . Figure 7.
Average pulse profile of Swift J1818.0 − + ( ) rad m − . it would be possible to split some of the data into frequencysub-bands, this is referred to future work. Our spectral fitscover the frequency ranges L to S ( ∼ . GHz), S to C( ∼ . GHz), or the full L to C-band range ( ∼ . GHz).In total, we derive spectral indices at seven epochs, and thevast majority of those observations are nearly contiguousin time. We performed robust and uncertainty-weighted pa-rameter estimation using MCMC techniques, for which weemployed the emcee software. The start parameters for theMCMC runs were set based on initial maximum likelihoodfits. We measured spectral indices that vary between − . and − . , with a mean spectral index over all epochs of − . ± . , where the uncertainty is the standard error of themean. Our measurements agree well with estimates from theliterature obtained close to our observing epochs, e.g. thosefrom the Parkes telescope on MJD 58939 (2020 March 31)between about 0.7 to 4 GHz (Lower et al. 2020a) and the onefrom the Deep Space Network on MJD 58947 (2020 April 08)between 2.3 and 8.4 GHz (Majid et al. 2020a), which fur-ther reassured us of the fidelity of our calibration methods.The magnetar’s radio spectrum was therefore surprisinglysteep at the times of measurement and showed a signifi- MNRAS000
Average pulse profile of Swift J1818.0 − + ( ) rad m − . it would be possible to split some of the data into frequencysub-bands, this is referred to future work. Our spectral fitscover the frequency ranges L to S ( ∼ . GHz), S to C( ∼ . GHz), or the full L to C-band range ( ∼ . GHz).In total, we derive spectral indices at seven epochs, and thevast majority of those observations are nearly contiguousin time. We performed robust and uncertainty-weighted pa-rameter estimation using MCMC techniques, for which weemployed the emcee software. The start parameters for theMCMC runs were set based on initial maximum likelihoodfits. We measured spectral indices that vary between − . and − . , with a mean spectral index over all epochs of − . ± . , where the uncertainty is the standard error of themean. Our measurements agree well with estimates from theliterature obtained close to our observing epochs, e.g. thosefrom the Parkes telescope on MJD 58939 (2020 March 31)between about 0.7 to 4 GHz (Lower et al. 2020a) and the onefrom the Deep Space Network on MJD 58947 (2020 April 08)between 2.3 and 8.4 GHz (Majid et al. 2020a), which fur-ther reassured us of the fidelity of our calibration methods.The magnetar’s radio spectrum was therefore surprisinglysteep at the times of measurement and showed a signifi- MNRAS000 , 1–15 (2020) Champion et al. F l u x d e n s i t y ( m J y ) This workATel 13560 (GBT)ATel 13562 (MeerKAT)ATel 13575 (GBT) ATel 13580 (uGMRT)ATel 13587 (Parkes)ATel 13649 (DSN)lsjb20 (Parkes) F l u x d e n s i t y ( m J y ) S p e c t r a l i n d e x mean: linear fit Figure 8.
Top panel: Timeline of the pulse-averaged flux densityof Swift J1818.0 − − σ uncertainties from both the literature and our work, the meanspectral index computed over all epochs, and the best-fitted lin-ear function to the spectral indices in time. The spectral indiceswere obtained from fitting a simple power-law model to quasi-simultaneous band-integrated data from the Effelsberg telescopeat two or three frequency bands ( ∼ . to ∼ . GHz frequencycoverage). cant amount of variability, as can be seen in the early databefore MJD 58925 in particular. Additionally, from purelyvisual inspection it seemed that the spectrum became flatterover time (bottom panel of Fig. 8). To test that hypothesis,we fit both a constant and a linear function to the spectralindex time series using the same techniques as before andperformed model selection using the Akaike information cri-terion (e.g. Burnham et al. 2010). We found that there isstrong statistical evidence for a spectral flattening over time.The slope of the best-fitting linear function has a formalstatistical significance in excess of σ , and the probabilitythat the linear function is the best-fitting one among the twomodels tested is about 70 per cent. We conclude that we sawan initial transition of the source to a more magnetar-like,i.e. flatter, radio spectrum over a few weeks. Interestingly,Majid et al. (2020b) recently reported the measurement ofan inverted spectrum of the magnetar with spectral index of+0.3 between 2.3 and 8.4 GHz in mid July 2020. We have studied the newly discovered magnetar SwiftJ1818.0 − (cid:219) ν = − × − Hz , implying a characteristic age of about 500years. We consider this a more reliable characteristic age es-timate. Nevertheless, one should be cautious on whether thecalculated characteristic age (based on τ = P / (cid:219) P ) reflectsthe true age of the magnetar, because the braking index canwell be deviated from 3 or even time-varying (e.g., John-ston & Karastergiou 2017), and the birth period could besignificantly less than the current period. In particular, abraking index less than 3, which has already been observedin a small number of magnetars (Gao et al. 2016), could welllead to an underestimation of the true age while using thevalue of the characteristic age. Such caveats also apply toprevious claims of the age of the magnetar stated in Cham-pion et al. (2020) and Esposito et al. (2020). In fact, there isno reported coincident supernova remnant that could helpto further constrain the age of the magnetar (Green 2019;Lower et al. 2020a), though a non-detection may suggest MNRAS , 1–15 (2020) igh-cadence observations of Swift J1818.0 − a true age of a higher value. The combination of the spinperiod and spin-down rate locates Swift J1818.0 − P - (cid:219) P -diagram (see Fig. 10) which expandsthe current boundary of magnetars. In fact, it would be thefastest spinning magnetar known if PSR J1846 − − (cid:219) ν in addition to power-law rednoise, we caution that this does not rule out the generalclass of smooth models of (cid:219) ν variations, and that as seen inFigure 2(b), there is significant smooth variation in (cid:219) ν at ascale only slightly below that of the discrete events.The rapid onset of radio emission, and the high cadencetiming observations we have performed, have allowed us tostudy the coherent timing behaviour of this source shortlyafter its outburst. The bulk timing behaviour shows varia-tions which are similar (in a relative sense) to those seen inthe X-rays in the early days after the outburst from XTEJ1810 −
197 (Ibrahim et al. 2004) but with more detail. Thesevariations may also provide vital information on what is hap-pening to the underlying neutron star immediately after itreturns to quiescence.We note that the exact epoch of the first timing eventis estimated to coincide with Effelsberg observations at 2.7GHz. We have investigated the timing behaviour as well asthe emission properties around the estimated epoch, but therelatively low strength of the magnetar emission at this fre-quency on this day prevents us from studying informationbased on single pulses, which may have revealed magneto-spheric changes (cf. Palfreyman et al. 2018). Hence, giventhe available time resolution and signal strength, we do notdetect significant changes.Similarly to other radio-loud magnetars, SwiftJ1818.0 − -0.0002-0.00015-0.0001-5x10 -5 -5 ν - . ( H z ) MJD - 58945 ν . = -23.7 × -12 , τ c = 491 yr Figure 9.
Measurement of rotational frequency from individualepochs, together with a linear fit to the overall variation trend.The characteristic age was calculated with the (cid:219) ν obtained fromthe fit. Figure 10. P - (cid:219) P diagram with the position of Swift J1818.0 − a frequency-dependent emission height, a shrinkage ofthe beam could explain the unusually steep flux densityspectrum, as the emission would recede from our view athigher radio frequencies. The steep spectrum is indeed instark contrast with measurements from other radio-loudtransient magnetars that show high radio emission variabil-ity and generally flat ( α ≥ − . ; e.g. Camilo et al. 2007c;Keith et al. 2011; Torne et al. 2015; Dai et al. 2019) oreven slightly inverted spectra up to millimetre wavelengths(Torne et al. 2017). It is interesting to note that FRBs alsotend to show flat position angle swings (e.g. Fonseca et al.2020), and indeed the single pulse shown in Figure 6 or theaverage profiles shown in Figure 7 do resemble observedFRB pulse shapes. Given the short rotational period by MNRAS000
Measurement of rotational frequency from individualepochs, together with a linear fit to the overall variation trend.The characteristic age was calculated with the (cid:219) ν obtained fromthe fit. Figure 10. P - (cid:219) P diagram with the position of Swift J1818.0 − a frequency-dependent emission height, a shrinkage ofthe beam could explain the unusually steep flux densityspectrum, as the emission would recede from our view athigher radio frequencies. The steep spectrum is indeed instark contrast with measurements from other radio-loudtransient magnetars that show high radio emission variabil-ity and generally flat ( α ≥ − . ; e.g. Camilo et al. 2007c;Keith et al. 2011; Torne et al. 2015; Dai et al. 2019) oreven slightly inverted spectra up to millimetre wavelengths(Torne et al. 2017). It is interesting to note that FRBs alsotend to show flat position angle swings (e.g. Fonseca et al.2020), and indeed the single pulse shown in Figure 6 or theaverage profiles shown in Figure 7 do resemble observedFRB pulse shapes. Given the short rotational period by MNRAS000 , 1–15 (2020) Champion et al. magnetar-standards (Fig. 10), this source may provide yetanother indication that both source populations are related.To summarise, we have demonstrated the importanceof complementing X-ray with rapid radio follow-up observa-tions for studying the initial state of magnetars after theiractivation. In order to provide an unbiased view on the spin-evolution of these young neutron stars, a high cadence isrequired. A study of the associated radio emission proper-ties does not only help to unambiguously identify sources asmagnetars with their distinct features, but may also connectthe spin behaviour to magnetospheric processes.
ACKNOWLEDGEMENTS
FJ, KR and BS acknowledge funding from the European Re-search Council (ERC) under the European Union’s Horizon2020 research and innovation programme (grant agreementNo. 694745). GD, RK, KL, PT acknowledge the financialsupport by the European Research Council for the ERC Syn-ergy Grant BlackHoleCam under contract no. 610058. Pulsarresearch at Jodrell Bank Centre for Astrophysics and JodrellBank Observatory is supported by a consolidated grant fromthe UK Science and Technology Facilities Council (STFC).This publication is based on observations with the 100-mtelescope of the MPIfR (Max-Planck-Institut f ˜Aijr Radioas-tronomie) at Effelsberg. We thank Alex Kraus for schedul-ing our observations at the 100-m telescope so flexibly. TheNan¸cay Radio Observatory is operated by the Paris Obser-vatory, associated with the French Centre National de laRecherche Scientifique (CNRS).
DATA AVAILABILITY
The data underlying this article will be shared on reasonablerequest to the corresponding author.
REFERENCES
Burnham K. P., Anderson D. R., Huyvaert K. P., 2010, BehavioralEcology and Sociobiology, 65, 23Camilo F., Ransom S. M., Halpern J. P., Reynolds J., HelfandD. J., Zimmerman N., Sarkissian J., 2006, Nature, 442, 892Camilo F., Reynolds J., Johnston S., Halpern J. P., Ransom S. M.,van Straten W., 2007a, ApJ, 659, L37Camilo F., et al., 2007b, ApJ, 663, 497Camilo F., et al., 2007c, ApJ, 669, 561Champion D., et al., 2020, The Astronomer’s Telegram, 13559, 1Cordes J. M., Lazio T. J. W., 2002, arXiv e-prints, pp astro–ph/0207156Dai S., et al., 2018, MNRAS, 480, 3584Dai S., et al., 2019, ApJ, 874, L14Desvignes G., et al., 2018, ApJ, 852, L12Dewey R. J., Taylor J. H., Weisberg J. M., Stokes G. H., 1985,ApJ, 294, L25Duncan R. C., Thompson C., 1992, ApJ, 392, L9Eatough R. P., et al., 2013, Nature, 501, 391Ellis J. A., Vallisneri M., Taylor S. R., Baker P. T., 2019, EN-TERPRISE: Enhanced Numerical Toolbox Enabling a RobustPulsaR Inference SuitE (ascl:1912.015)Enoto T., et al., 2020, The Astronomer’s Telegram, 13551, 1Espinoza C. M., Lyne A. G., Kramer M., Manchester R. N., KaspiV. M., 2011, ApJ, 741, L13 Esposito P., et al., 2020, ApJ, 896, L30Evans P. A., et al., 2020, GRB Coordinates Network, 27373, 1Fonseca E., et al., 2020, ApJ, 891, L6Foreman-Mackey D., Hogg D., Lang D., Goodman J., 2013,PASP, 125, 306Gao Z. F., Li X. D., Wang N., Yuan J. P., Wang P., Peng Q. H.,Du Y. J., 2016, MNRAS, 456, 55Gavriil F. P., Gonzalez M. E., Gotthelf E. V., Kaspi V. M., Liv-ingstone M. A., Woods P. M., 2008, Science, 319, 1802Green D. A., 2019, Journal of Astrophysics and Astronomy, 40,36Halpern J. P., Gotthelf E. V., Becker R. H., Helfand D. J., WhiteR. L., 2005, ApJ, 632, L29Haslam C., Salter C., Stoffel H., Wilson W., 1982, A&AS, 47, 1Hu C.-P., et al., 2020, The Astronomer’s Telegram, 13588, 1Ibrahim A. I., et al., 2004, ApJ, 609, L21Jankowski F., van Straten W., Keane E., Bailes M., Barr E.,Johnston S., Kerr M., 2018, MNRAS, 473, 4436Johnston S., Karastergiou A., 2017, MNRAS, 467, 3493Johnston S., Kerr M., 2018, MNRAS, 474, 4629Karuppusamy R., et al., 2020, The Astronomer’s Telegram,13553, 1Kaspi V. M., Beloborodov A. M., 2017, ARA&A, 55, 261Kaspi V. M., Kramer M., 2016, in Proceedings of the 26th SolvayConference of Physics. World Scientific. p. arXiv:1602.07738( arXiv:1602.07738 )Keane E. F., Kramer M., 2008, MNRAS, 391, 2009Keith M. J., Johnston S., Levin L., Bailes M., 2011, MNRAS,416, 346Kouveliotou C., et al., 1998, Nature, 393, 235Kramer M., Karastergiou A., Gupta Y., Johnston S., BhatN. D. R., Lyne A. G., 2003, A&A, 407, 655Kramer M., Stappers B. W., Jessner A., Lyne A. G., Jordan C. A.,2007, MNRAS, 377, 107Kuiper L., Hermsen W., Dekker A., 2018, MNRAS, 475, 1238Lawson K. D., Mayer C. J., Osborne J. L., Parkinson M. L., 1987,MNRAS, 225, 307Lazarus P., Karuppusamy R., Graikou E., Caballero R. N., Cham-pion D. J., Lee K. J., Verbiest J. P. W., Kramer M., 2016,MNRAS, 458, 868Lentati L., Alexander P., Hobson M. P., Feroz F., van HaasterenR., Lee K. J., Shannon R. M., 2014, MNRAS, 437, 3004Lentati L., Kerr M., Dai S., Shannon R. M., Hobbs G., Os(cid:32)lowskiS., 2017, MNRAS, 468, 1474Levin L., et al., 2010, ApJ, 721, L33Levin L., et al., 2012, MNRAS, 422, 2489Levin L., et al., 2019, MNRAS, 488, 5251Livingstone M. A., Ng C. Y., Kaspi V. M., Gavriil F. P., GotthelfE. V., 2011, ApJ, 730, 66Lower M. E., Shannon R. M., Johnston S., Bailes M., 2020a, ApJ,896, L37Lower M. E., Buchner S., Johnston S., Parthasarathy A., GeyerM., Bailes M., 2020b, The Astronomer’s Telegram, 13562, 1Lyne A., Hobbs G., Kramer M., Stairs I., Stappers B., 2010, Sci-ence, 329, 408Lyubarsky Y., 2014, MNRAS, 442, L9Maan Y., Joshi B. C., Surnis M. P., Bagchi M., Manoharan P. K.,2019, ApJ, 882, L9Majid W. A., Pearlman A. B., Nimmo K., Hessels J. W. T., PrinceT. A., Naudet C. J., Kocz J., Horiuchi S., 2020a, ApJ, 897,L4Majid W. A., Pearlman A. B., Prince T. A., Naudet C. J., BansalK., 2020b, The Astronomer’s Telegram, 13898, 1Manchester R. N., et al., 2013, Publ. Astron. Soc. Australia, 30,e017Olausen S. A., Kaspi V. M., 2014, ApJS, 212, 6Palfreyman J., Dickey J. M., Hotan A., Ellingsen S., van StratenW., 2018, Nature, 556, 219 MNRAS , 1–15 (2020) igh-cadence observations of Swift J1818.0 − Parthasarathy A., et al., 2019, MNRAS, 489, 3810Rajwade K., et al., 2020, The Astronomer’s Telegram, 13554, 1Serylak M., et al., 2009, MNRAS, 394, 295Spitler L. G., et al., 2016, Nature, 531, 202Torne P., et al., 2015, MNRAS, 451, L50Torne P., et al., 2017, MNRAS, 465, 242Torne P., et al., 2020, A&A, 640, L2Weltevrede P., 2016, A&A, 590, A109Yao J. M., Manchester R. N., Wang N., 2017, ApJ, 835, 29van Straten W., 2013, ApJS, 204, 13van Straten W., Bailes M., 2011, Publ. Astron. Soc. Australia,28, 1van Straten W., Demorest P., Oslowski S., 2012, AstronomicalResearch and Technology, 9, 237This paper has been typeset from a TEX/L A TEX file prepared bythe author.MNRAS000