A Month of Monitoring the New Magnetar Swift J1555.2−5402 during an X-Ray Outburst

The Swift BAT (Gehrels et al. 2004) detected a burst from an unknown source at 09:45:46 UT on 2021 June 3 (trigger number 1053220) and immediately pointed to the source direction (Palmer et al. 2021). The Swift XRT (Burrows et al. 2005) obtained X-ray data in the WT mode for 62 s from 97 s after the BAT trigger and then in the PC mode for ∼1.7 ks from 1.1 hr after the burst. The XRT observations determined the source position (J2000.0) to be R.A. =15^h 55^m 0866 and decl. = −54° 03' 411 with an uncertainty radius of 22 at a 90% confidence level (Evans 2021). We adopt this source position for all analyses in this Letter. The BAT detected another four short bursts from the same direction, as summarized in Appendix Table B1. We analyzed the BAT data using the standard HEASoft BAT pipelines (version 6.28), following the same procedure described in Lien et al. (2016); these results are presented in Section 3.3.

We analyzed the XRT data obtained on June 3, 4, 5, and 7. The observation IDs (ObsIDs) used in this Letter are listed in Appendix Table A1. The observations on June 4, 5, and 7 were carried out in the WT mode for a total of 8.9 ks. We processed the data through the standard procedure of FTOOLS xrtpipeline with the default filtering criteria. For PC mode data, we extracted source photons from a circular region with a 20 pixel radius (1 pixel = 236) centered at the target, whereas we collected background spectra from a source-free region with a similar (20 pixel) radius located far (>2') from the source. For the one-dimensional WT mode data, we extracted the source spectra from a rectangular strip of length 40 pixels (∼94''). The background was estimated using a source-free region of a similar size located ∼2' from the source position. We used the latest available RMF file in CALDB version 20210504 and generated the ARF files with the xrtmkarf tool.

NICER (Gendreau et al. 2016), on board the International Space Station, began X-ray observations of the source at 11:21:31 UT on 2021 June 3, 1.6 hr after the first short burst detected with Swift BAT. This initial NICER observation was 2.4 ks long in exposure (ObsID 4202190101), and it was followed by high-cadence monitoring (see Appendix Table A2) carried out almost daily for 29 days under an approved Cycle 3 proposal. Each ObsID had roughly 2 ks exposure and was divided into several continuous good time intervals (GTIs) with exposures of a few hundred seconds for each.

The NICER's X-ray Timing Instrument (XTI) has 52 on-orbit active modules, each of which consists of coaligned X-ray concentrators and silicon drift detectors. The XTI has a time resolution of <100 ns, and the total effective area is about 1,800 cm² at around 1.5 keV. We performed the standard analysis procedures using NICERDAS (version 2020-04-23_V007a) in HEASoft 6.27.2 and the NICER calibration database (version 20200722). We generated level 2 cleaned events with the nicerl2 command. For the barycentric correction, we used barycorr with Jet Propulsion Laboratory solar system development ephemeris DE405 for the source coordinates stated above (Standish 1998). For timing analyses and burst searches, we utilized all 52 active modules. For spectral studies, we further excluded module numbers 14 and 34 to avoid potential contamination by instrumental noise in the soft energy band. The background spectral model is generated using the 3C50 background model with the nibackgen3c50 command (Remillard et al. 2021). ³³

NuSTAR (Harrison et al. 2013) observed Swift J1555.2−5402 on June 5–6 for a 38.4 ks exposure and two additional contemporaneous observations with NICER on June 9 (25 ks) and 21 (29 ks). We processed and filtered the NuSTAR data following the standard procedures with HEASoft version 6.28 and CALDB version 20210524 using the nupipline and nuproducts commands. We extracted on-source and background spectra from circular regions of 80'' radius centered at the source position and in a source-free region, respectively. The background-subtracted source count rate of FPMA was about 1 count s⁻¹ in the 3–79 keV band. For the spectral fitting, we grouped source spectra using the grppha tool of HEASoft, such that each spectral bin would have a minimum of 50 counts.

We carried out radio observations of Swift J1555.2−5402 for a total exposure of roughly 10.8 hr at five epochs during 2021 June 4–12 using different Deep Space Network (DSN) radio telescopes (see Table A4). Simultaneous dual-frequency bands, with center frequencies at 2.2 (S band) and 8.4 (X band) GHz, were used for all observations. We used a single circular polarization mode for DSS-34 and DSS-36, whereas a dual circular polarization mode was used at each frequency band for DSS-43.

We recorded the data in filter-bank mode with a time resolution of 512 μs and frequency resolution of 1 MHz using the pulsar machine in Canberra. The data processing procedure follows similar steps to those presented in earlier studies of pulsars and magnetars with the DSN (e.g., Majid et al. 2017; Pearlman et al. 2018, 2019). After first flattening the bandpass response in each data set, we removed the low-frequency variations in the temporal baseline of each frequency channel by subtracting the moving average from each data point with a time constant of 10 s. The sample times were then corrected to the solar system barycenter.

The 20 m diameter Ishigaki-jima station of the VLBI Exploration of Radio Astrometry (VERA) conducted a target-of-opportunity observation of this source at an observation frequency of 22 GHz (1.3 cm, K band) with a bandwidth of 512 MHz. The acquired data for 1 hr at 14:40–15:40 UT on 2021 June 6 were processed and folded to explore radio pulsations both with and without assuming the rotation period. Because the data quality was limited due to a low elevation of the object and bad weather conditions, we could only set an upper limit of the peak flux density of 1.02 Jy (1σ).

Figure 1 shows the time series of the physical parameters of Swift J1555.2−5402 during NICER monitoring for the 29 days from shortly after the onset of the outburst until July 1. In constructing the time series, we first derived the pulsar spin ephemeris, for which we used a Gaussian pulse template and constructed pulse times of arrival (TOAs) with an integration time of 300 s and a minimum exposure of 200 s contained in each bin, using the script photon_toa.py from NICERsoft. ³⁴ The timing analysis was carried out over 2–8 keV, where the energy range was determined from a Z_n ² search with n = 2 to optimize the pulse significance (Buccheri et al. 1983). We used the Python-based package for high-precision timing analysis "PINT" (Luo et al. 2021; version 0.8.2) to compute the best timing model through a weighted least-squares fit to the TOAs. The TOAs were found to be well described by a fifth-order polynomial model, as summarized in Table 1, for the spin evolution of Swift J1555.2−5402. The best-fit frequency and its derivatives are ν = 0.258997103(8) Hz, $\dot{\nu }=-2.04(5)\,\times {10}^{-12}$ Hz s⁻¹, $\ddot{\nu }=-4.50(13)\times {10}^{-18}$ Hz s⁻², ν⁽³⁾ = − 1.10(10) × 10⁻²³ Hz s⁻³, ν⁽⁴⁾ = 3.59(15) × 10⁻²⁹ Hz s⁻⁴, and ν⁽⁵⁾ = 1.59(14) × 10⁻³⁴ Hz s⁻⁵ at barycentric epoch T₀ = MJD 59,382.7549. NICER's sensitivity enables our measurement up to fifth order in frequency with TOAs of only 300 s.

Zoom In Zoom Out Reset image size

Table 1. Summary of the Timing and Spectral Properties of Swift J1555.2−5402

Parameter	Values
Timing Properties (NICER Monitoring)
MJD range	59,368.58–59,396.98
Epoch T0 (MJD)	59,382.7549
Spin frequency ν (Hz)	0.258997103(8)
Frequency derivative $\dot{\nu }$ (10−12 Hz s−1)	−2.04(5)
Second frequency derivative $\ddot{\nu }$ (10−18 Hz s−2)	−4.50(13)
Third frequency derivative ν(3) (10−23 Hz s−3)	−1.10(10)
Fourth frequency derivative ν(4) (10−29 Hz s−4)	3.59(15)
Fifth frequency derivative ν(5) (10−34 Hz s−5)	1.59(14)
rms residual (phase)	0.014
χ2/dof	117.863/126
Period P (s)	3.86104705(12)
Period derivative $\dot{P}$ (10−11 s s−1)	3.05(7)
Second period derivative $\ddot{P}$ (10−17 s s−2)	6.7(2)
Third period derivative P(3) (10−22 s s−3)	1.63(15)
Fourth period derivative P(4) (10−28 s s−4)	−5.3(2)
Fifth period derivative P(5) (10−33 s s−5)	−2.4(2)
Characteristic age τc (kyr)	2.01(5)
Surface magnetic field Bsurf (1014 G)	3.47(4)
Spin-down luminosity Lsd (1034 erg s−1)	2.09(5)
Spectral Properties (NICER+NuSTAR Joint Fit)
Joint observation numbers	1	2	3	Average
Observation date (MJD)	59,370	59,374	59,386	⋯
Column density NH (1022 cm−2)	8.72(8)			8.69(7)
Temperature kT (keV)	1.144(4)	1.148(4)	1.153(4)	1.149(3)
Radius R (km)	2.10(2)	2.05(2)	2.04(2)	2.06(3)
Photon index Γ	1.27(12)	1.68(17)	1.15(0.16)	1.35(9)
Absorbed 2–10 keV flux (10−12 erg s−1 cm−2)	${46.16}_{-0.38}^{+0.19}$	${45.41}_{-0.55}^{+0.14}$	${44.48}_{-0.45}^{+0.18}$	${45.81}_{-0.19}^{+0.04}$
Unabsorbed 2–10 keV flux (10−12 erg s−1 cm−2)	70.60(32)	69.39(35)	67.80(36)	69.5(4)
10–60 keV flux (10−12 erg s−1 cm−2)	9.32 ± 0.34	7.73 ± 0.35	8.72 ± 0.85	9.02 ± 0.24
2–10 keV luminosity Lx (1035 erg s−1)	8.45	8.30	8.11	8.32
10–60 keV luminosity Lx (1035 erg s−1)	1.11	0.93	1.04	1.08
Quiescent (Swift)
Absorbed 2–10 keV flux (10−12 erg s−1 cm−2)	<0.26 (3σ)
Unabsorbed 2–10 keV flux (10−12 erg s−1 cm−2)	<1.2 (3σ)
2–10 keV luminosity Lx (1035 erg s−1)	<0.15 (3σ)

Note. The column density of the three joint NICER and NuSTAR spectral fittings is fixed to the same value. The X-ray luminosity and radius are calculated on an assumption of a fiducial distance of 10 kpc. Quoted errors indicate the 68% confidence limit. The quiescent X-ray upper limits are derived with an assumed temperature of 0.45 keV (see the text).

Download table as:  ASCIITypeset image

In Figures 2(a)–(d), we present the energy-resolved and background-subtracted pulse profiles of Swift J1555.2−5402 in the 2–3 and 3–8 keV bands with NICER and 3–8, 8–12, and 12–20 keV with NuSTAR, where estimates of the background rates were made with the nibackgen3C50 tool for the NICER data and based on the measured rate in the background region for the NuSTAR data. The soft X-ray profile (2–12 keV) shows a single-peaked, nearly sinusoidal shape, while the pulsation in the hard X-ray band (≳12 keV) was not detected. We further divided the data into finer energy bands and calculated, in Figure 2(f), the time-averaged rms pulsed fraction (PF) as defined in Bildsten et al. (1997) and Woods et al. (2004). Both the NICER and NuSTAR observations suggest that the rms PF increases from ∼30% in the softer band (3–4 keV) to a maximum of ∼40% at around 7 keV and then decreases with energy to ≲10%. We also found that the PF in the 3–8 keV range remained almost constant during the observed period (Figure 1(e)).

Zoom In Zoom Out Reset image size

The right panels of Figure 1 show the long-term spectral properties of Swift J1555.2−5402 obtained with Swift, NICER, and NuSTAR data. Here we applied a single-temperature blackbody multiplied by the Tuebingen–Boulder interstellar absorption model (tbabs∗bbodyrad in Xspec terminology) to derive the physical parameters at each epoch (Wilms et al. 2000). The first data point in each of the right panels corresponds to the initial 1.7 ks Swift PC spectrum obtained ∼1.1 hr after the BAT trigger, which is well fitted with the model above (χ² of 113 for 128 degrees of freedom, dof). We derived a best-fit hydrogen column density of ${N}_{{\rm{H}}}={6.8}_{-1.7}^{+2.0}\times {10}^{22}$ cm⁻² and blackbody temperature of ${kT}={1.26}_{-0.16}^{+0.20}$ keV. The absorbed 2–10 keV flux is ${6.8}_{-0.8}^{+0.9}\times {10}^{-11}$ erg s⁻¹ cm⁻². This flux is significantly higher than those in the following observations with the Swift/XRT in the WT mode and NICER monitoring.

The subsequent daily NICER spectra (1.7–10 keV) were systematically fitted with the same model with the hydrogen column density tied to be the same value among all NICER spectra at N_H = (8.88 ± 0.12) × 10²² cm⁻². Each observation has ∼4 counts s⁻¹ (Figure 1(f)). The reduced χ² values were ∼0.9–1.2 for ∼100–400 dof. No spectral variation during the initial monitoring was found. The absorbed and unabsorbed 2–10 keV fluxes were ∼4.3 × 10⁻¹¹ (Figure 1(g)) and ∼7.5 × 10⁻¹¹ erg s⁻¹ cm⁻², respectively. The derived temperature and emission radius were constant at ∼1.1 keV (Figure 1(h)) and ∼2 d₁₀ km (Figure 1(i)), respectively (Section 3.5). These NICER parameters are consistent with those obtained with the WT mode data of Swift/XRT.

We performed joint spectral fits of three observations of NICER and NuSTAR on June 5–6, 9, and 21. Since the NICER spectra showed no significant time variation, we extracted the NICER spectra for the period on the same day as NuSTAR and regarded them as simultaneous, even if their observation periods were not fully simultaneous. The column densities among the three epochs are tied to the same value. The best-fit spectral model is shown in Figure 3. In addition to the soft X-ray blackbody component, a hard X-ray component above 10 keV was detected with 3σ significance extending up to at least 40 keV. The hard X-ray flux, when fitted by the power-law model, was (7–9) × 10⁻¹² erg s⁻¹ cm⁻² in the 10–60 keV band with a power-law photon index of 1.2–1.7. We also performed a combined fit of all three epochs, given that no significant spectral change was observed between them. The resultant average ν F_ν is shown in the right panel of Figure 3 and Table 1. The ν F_ν plot shows that the hard X-ray component above 10 keV exceeds the high-energy tail of the soft blackbody component.

Zoom In Zoom Out Reset image size

Swift BAT detected five short bursts, as summarized in Appendix Table B1. For the first detected burst on June 3 (an onboard trigger), we used the data from T − 240 to T + 100 s, where T is the burst detection time, whereas we used the ∼3 s interval events collected through subthreshold triggers ³⁵ in our analyses of the other bursts. The BAT event data have a time resolution of ∼100 μs (Barthelmy et al. 2005). The BAT temporal analysis utilizes light curves binned in 1, 2, and 4 ms. Our BAT spectral analysis was performed using spectra created with the T₉₀ duration of each burst, which is the duration that covers 90% of the burst emission. All of the spectra were successfully fitted with a single blackbody model (bbodyrad model in Xspec), except for the one on June 7, the statistics of which were too poor to give meaningful constraints.

We also searched the 2–8 keV NICER event data for short bursts using the Bayesian block technique ³⁶ (Scargle et al. 2013) and the data filtering criteria in Appendix B. The blocks with high backgrounds and durations longer than ∼1 s are further filtered out on the basis of comparison between the housekeeping data (mkf files), multiple blocks in one burst, and blocks close to the GTI boundaries. We used the Poisson probability to determine the significance of detecting a number of photons in a block, where the nonburst count rate was calculated from 1 s intervals close in time to the bursts. We identified 45 short bursts exceeding a 5σ detection significance, as summarized in Appendix Table B2. The average duration of the bursts was 23 ± 17 ms, and 13 photons were detected in a burst, on average. Note that burst 27–1 on the list occurred during the tail of burst 27, and we included it in burst 27 in the calculation.

We stacked the detected bursts to obtain an average spectrum and found it to be equally well fitted with both a single blackbody (tbabs ∗ bbodyrad) and a power-law (tbabs ∗ pegpwrlw) with Cash statistics of C-stat = 261.1 and 264.0, respectively, with 310 dof. When fixing the absorption column density at N_H = 8.72 × 10²² cm⁻² (Table 1), the former model gave a blackbody temperature of ${2.9}_{-0.5}^{+0.8}$ keV, whereas the latter gave a photon index of 0.0 ± 0.2. Using the blackbody model, we find an average unabsorbed flux of (1.3 ± 0.1) × 10⁻⁸ erg s⁻¹ cm⁻² in the 2–8 keV range. Assuming a distance of 10 kpc, the blackbody radius is estimated to be 7.4 ± 1.6 km. The fluences of detected bursts are calculated to be in the range of (1–13) × 10⁻¹⁰ erg cm⁻² with an assumed blackbody spectrum of kT = 2.8 keV using the WebPIMMs Appendix (Appendix Table B2). One of the NICER bursts was simultaneously detected with Swift BAT (Appendix Figure B2).

We dedispersed the data of each epoch with trial Dispersion Measures (DMs) between zero and 5000 pc cm⁻³ and subsequently searched each resulting time series for both periodic and single-pulse emission. We found no statistically significant periods with a signal-to-noise ratio (S/N) above 7.0 after folding individual dedispersed time series modulo period candidates from PRESTO's accelsearch package (Ransom et al. 2002). In addition, we folded the dedispersed time series at each DM trial using the timing model from NICER in Table 1 but found no evidence of radio pulsations at the S or X band during any of our observations. For each epoch, we place 7σ upper limits on the magnetar's flux density, assuming a duty cycle of 10% (see Table A4). Based on our longest observation with DSS-43, the 70 m radio telescope in Tidbinbilla, Australia, we obtain 7σ upper limits on the magnetar's flux of <0.043 mJy at the S band and <0.026 mJy at the X band.

We also searched the dedispersed time series at each frequency band for radio bursts using a matched filtering algorithm, where the time series was convolved with boxcar functions with logarithmically spaced widths between 512 μs and 150 ms. Candidates with a detection S/N above 7.0 were saved and classified using the FETCH software package (Agarwal et al. 2020). The dynamic spectra of the candidates were also visually inspected for verification. We detected no radio bursts during the radio observations and placed 7σ upper limits on the fluence of individual bursts during each epoch at both the S and X bands (see Table A4). On 2021 June 5 and 6, we detected two X-ray bursts, with widths of w = 15.91 (see Table B2; burst 8) and 0.27 ms (see Table B2; burst 10), during overlapping radio and X-ray observations. However, no prompt radio emission (within ±10 s of the X-ray burst time) was detected above a 7σ fluence detection threshold of $1.6\sqrt{w/1\,\mathrm{ms}}$ and $0.61\sqrt{w/1\,\mathrm{ms}}$ Jy ms at the S and X bands, respectively.

We searched the Swift archival data for a serendipitous detection of Swift J1555.2−5402 in its quiescent state. Two observations in 2012 (ObsIDs 00042728001 and 00042729001) covered the location of this source as part of the Swift XRT Galactic plane survey program (Reynolds et al. 2013). The total exposure is 1055 s. We find no X-ray source at this position within a 30'' radius. The 3σ upper limit of the count rate is estimated to be 5.0 × 10⁻³ counts s⁻¹ within this radius. Then, the 3σ upper limits of the 2–10 keV absorbed and absorption-corrected fluxes are calculated to be 5.0 × 10⁻¹³ and 9.0 × 10⁻¹³ erg s⁻¹ cm⁻², respectively, on the assumption of an absorbed blackbody spectrum with N_H = 8.72 × 10²² cm⁻² and kT = 1.1 keV. Yet the magnetar surface temperature in quiescence is considerably smaller than that during outburst epochs, with an average of about 0.45 keV among the population (Olausen & Kaspi 2014). Hence, we estimate 2–10 keV absorbed and absorption-corrected fluxes of about 2.6 × 10⁻¹³ and 1.2 × 10⁻¹² erg s⁻¹ cm⁻², respectively. This corresponds to an upper limit of the 2–10 keV quiescent X-ray luminosity of 1.5 × 10³⁴ erg s⁻¹ (bolometric thermal luminosity of 4.8 × 10³⁴ erg s⁻¹) at 10 kpc.

The characteristic age of Swift J1555.2−5402 is inferred to be 2.01 kyr (see Table 1). If we assume that its true age is comparable to its characteristic age, we expect to find a young supernova remnant (SNR) surrounding the neutron star. Detection of an associated SNR, combined with a possible proper-motion measurement, would be helpful and can be important for constraining the magnetar's true age given that the characteristic age may be unreliable due to underlying assumptions about the neutron star rotation period at the birth and braking index. Our search of archival radio, infrared, and X-ray data (e.g., Green 2019) for an SNR or a pulsar-wind nebula coinciding at the position of Swift J1555.2−5402, (l, b) = (327.872, −0.335) in the galactic coordinates using SkyView ³⁷ fails to yield any convincing candidates.

High-cadence monitoring with NICER over 1 month allows us to measure the spin ephemeris of the new source Swift J1555.2−5402 (Table 1). Refined from the initial reports in GCNs and ATels (Coti Zelati et al. 2021b; Ng et al. 2021), the period and its time derivative are measured to be 3.86104705(12) s and 3.05(7) × 10⁻¹¹ s s⁻¹, respectively. The combination of the two values falls within the distribution of known magnetars on the P–$\dot{P}$ diagram (Figure 4(a)). Assuming the standard rotating magnetic dipole model and a braking index of n = 3, these timing parameters correspond to a characteristic age of τ = 2.01(5) kyr, surface magnetic field strength of B_surf = 3.47(4) × 10¹⁴ G, and spin-down luminosity of L_sd = 2.09(5) × 10³⁴ erg s⁻¹. This source was classified as a magnetar based on the measured strong magnetic field. The derived characteristic age suggests that it is one of the youngest magnetars among the known ones. The suggestion is supported by the observed strong timing noise, which requires a model with high-order polynomials (Section 3.1), similar to that of the young magnetar Swift J1818.0−1607 (Hu et al. 2020). We caution that the derived pulsar parameters during the outburst may deviate from those in the quiescent state (e.g., Younes et al. 2017a; Archibald et al. 2020). Frequency derivatives are known to fluctuate during magnetar outbursts, with variations of a factor of 1–50 (see, e.g., Dib et al. 2012; Dib & Kaspi 2014; Levin et al. 2019). Thus, the accuracy of the inferred parameters ${B}_{\mathrm{surf}}\propto \sqrt{\dot{P}}$, $\tau \propto {\dot{P}}^{-1}$, and ${L}_{\mathrm{sd}}\propto \dot{P}$ relative to the quiescent values still has uncertainties due to this $\dot{P}$ variation.

Zoom In Zoom Out Reset image size

We detect a hard X-ray tail above 10 keV with NuSTAR, extending up to at least 40 keV with 3σ significance. The spectral energy distribution shows that the hard X-ray component is distinguished from the blackbody component, which should originate from the stellar surface (Figure 3). The existence of the distinctive hard tail is further supported by the steep drop in the energy-dependent PF above 10 keV (Figure 2(f)). Two-component spectra of this kind are reported from other persistently bright and transient magnetars (Kuiper et al. 2006; Enoto et al. 2010a; Younes et al. 2017b). The lack of pulsations (PF < 10%) in the hard X-rays may suggest that the hard tail originates in magnetospheric emission that does not have much anisotropy and higher emission altitude than emission from the stellar surface. This may imply a low magnetic impact parameter (e.g., Wadiasingh et al. 2018) for the observer across the pulse for resonant Compton scattering. The 15–60 keV flux of Swift J1555.2−5402, F₁₅₋₆₀ = 6.21 ± 0.23 × 10⁻¹² erg s⁻¹ cm⁻², is lower than the absorbed 1–10 keV flux ${F}_{1-10}={4.65}_{-0.27}^{+0.19}\times {10}^{-11}$ erg s⁻¹ cm⁻². Accordingly, the broadband hardness ratio of the magnetospheric to surface thermal emissions is η = F₁₅₋₆₀/F₁₋₁₀ = 0.13. The hardness ratio η of known magnetars (η = 0.1–4) is suggested to be correlated with the surface magnetic field. Figure 4(d) plots the η values of known sources and Swift J1555.2−5402. It shows that the η of Swift J1555.2−5402 is not largely off, though apparently smaller than, the proposed correlation (see Equation (3) of Enoto et al. 2017). We note that the measured η has some systematic uncertainty. If the yet-unknown long-term frequency derivative of Swift J1555.2−5402 is lower than the value measured during the current outburst, the magnetic field strength in the quiescent state is weaker than our estimate, which places the η of Swift J1555.2−5402 closer to the known correlation.

Figure 4(b) compares pulsar X-ray luminosity L_x with spin-down luminosity L_sd. Immediately after the outburst, the surface emission (2–10 keV) and total (2–60 keV) luminosity of Swift J1555.2−5402 were ${L}_{{\rm{x}}}=8.5\times {10}^{35}{d}_{10}^{2}$ and $9.6\,\times {10}^{35}{d}_{10}^{2}$ erg s⁻¹, respectively. At the assumed distance of 10 kpc, these are, respectively, 40 and 46 times larger than its spin-down luminosity L_sd = 2.1 × 10³⁴ erg s⁻¹. The observed X-ray luminosity is similar to those of past reported outbursts of transient magnetars. The upper limit on the quiescent X-ray luminosity of Swift J1555.2−5402 is ${L}_{\mathrm{quie}}\approx 1.5\times {10}^{34}{d}_{10}^{2}$, corresponding to ${L}_{\mathrm{quie}}/{L}_{\mathrm{sd}}\lesssim 0.71{d}_{10}^{2}$, which is still up to 2−3 orders of magnitude higher than those of the rotation-powered pulsars (see Figure 4(b) and Figure 12 of Enoto et al. 2019). The upper limit on the quiescent bolometric thermal luminosity $\approx 4.8\times {10}^{34}{d}_{10}^{2}$ erg s⁻¹ also makes Swift J1555.2−5402 one of the coolest young magnetars (see, e.g., Potekhin et al. 2015) and is compatible with the L ∼ 2 × 10³³ erg s⁻¹ of PSR J1119−6127 (Gonzalez et al. 2005; Ng et al. 2012; Blumer et al. 2021). The upper limit on the quiescent bolometric thermal luminosity $\approx 4.8\times {10}^{34}{d}_{10}^{2}$ erg s⁻¹ erg s⁻¹ is compatible with L ∼ 2 × 10³³ erg s⁻¹ of PSR J1119−6127 (Gonzalez et al. 2005; Ng et al. 2012; Blumer et al. 2021) and could mean that Swift J1555.2−5402 is one of the cooler young magnetars (see, e.g., Potekhin et al.2015).

The persistent X-ray flux of most transient magnetars remains in a bright plateau state for a few weeks immediately after the onset of outbursts and then starts to fade over the next several months (Coti Zelati et al. 2018). Typically, the plateau duration and decaying slope are τ₀ = 11–43 days and p ∼ 0.7–2 (Enoto et al. 2017), respectively, when the X-ray flux F_x is fitted with an empirical formula ${F}_{{\rm{x}}}{(t)={F}_{0}/(1+t/{\tau }_{0})}^{p}$, where t and F₀ are the elapsed time and plateau flux, respectively. A peculiarity of Swift J1555.2−5402 is its long-lasting outburst. The absorbed X-ray flux stayed nearly stable at around 4 × 10⁻¹¹ erg s⁻¹ cm⁻² with only a slow decline over a month. There was no apparent rollover of the flux trend as of July 1. Such long-lasting outbursts are rare for magnetars, with only a few exceptions ever recorded, e.g., the radio-loud magnetar 1E 1547.0−5408.

Another peculiarity of the Swift J1555.2−5402 outburst is the higher temperature (kT ∼ 1.1 keV) of the blackbody component than the typical value of known magnetars, kT ∼ 0.3–0.7 keV (Enoto et al. 2017), despite the surface emission radius R ∼ 2d₁₀ km of Swift J1555.2−5402 being well within the typical range for a neutron star. During the 1 month observation, the slow flux decline originated from the decreasing emission radius rather than the temperature decline. This fact suggests a situation where the hot spot on the neutron star surface responsible for the X-ray emission was shrinking, which is consistent with the twisted magnetosphere model (Kaspi & Beloborodov 2017).

We detected 45 and 5 short bursts with NICER and Swift/BAT, respectively, during the initial 2 weeks since discovery of Swift J1555.2−5402 (Figure 1(j)). The burst-active periods of persistently bright magnetars (e.g., SGR 1806−20) are known to be longer (≳100 days) than those (≲10 days) of low-burst rate transient ones (e.g., SGR 0501+4516; Göǧüş 2014). The burst-active period of Swift J1555.2−5402 is close to the latter case. We conjecture that repeated bursts as observed in Swift J1555.2−5402 would provide impulsive heating of the surface to sustain the long-lasting decay. To investigate the potential relationship between the bursts and persistent emission on the pulse profile, we plot in Figure 2(e) the phase distribution of the observed bursts. An Anderson–Darling (AD) test suggests that the burst phase distribution differs from a uniform distribution with an AD statistic of 0.42 and corresponding p-value of 0.82. Thus, there is neither a statistical difference between the phase distribution of the bursts and uniform distribution nor a statistically significant correlation between the burst occurrence and the pulse profile so far.

We did not detect any radio emission from this source. It has been long established theoretically that as far as ordinary pulsars are concerned, the occurrence of rotation-powered polar-cap radio emission requires a sufficiently large potential drop ΔΦ to generate electron-positron pairs near polar caps (Goldreich & Julian 1969; Sturrock 1971; Ruderman & Sutherland 1975; Arons & Scharlemann 1979); note that the conventional definition of ${\rm{\Delta }}{\rm{\Phi }}\propto {L}_{\mathrm{sd}}^{1/2}$ implies that radio emission is equivalently related to L_sd (Ho 2013). The observed radio luminosity of rotation-powered pulsars indeed follows the relation (e.g., Arzoumanian et al. 2002). In the absence of magnetically induced nonpotential fields, magnetars that satisfy this condition at their polar caps before crossing the death line should, in principle, be capable of producing coherent radio emission. However, as shown in Figures 4(a)–(c) and summarized in Appendix Table C1, pulsed radio emission has only been reported from six radio-loud magnetars and a high-B pulsar that exhibited a magnetar-like outburst. It is an open question under what conditions a magnetar becomes radio-loud. The new magnetar Swift J1555.2−5402 is located in the P–$\dot{P}$ parameter space close to radio-loud magnetars (Figure 4(a)). The DSN upper limits on the radio luminosity ($\nu {L}_{\nu }=1.1\times {10}^{28}{d}_{10}^{2}$ erg s⁻¹ for the S band and $2.6\times {10}^{28}{d}_{10}^{2}$ erg s⁻¹ for the X band) would be much lower than those of the known radio-loud magnetars, ν L_ν ∼ (0.08–1) × 10³⁰ and ∼(0.01–3) × 10³¹ erg s⁻¹ at the S and X bands, respectively (Camilo et al. 2007a, 2007b, 2008; Levin et al. 2010, 2012; Shannon & Johnston 2013; Pennucci et al. 2015; Huang et al. 2021). Therefore, the lack of radio emission suggests the existence of some other physical factors than simply P and $\dot{P}$ that govern radio emission.

One crucial factor is the geometry among the pulsar rotation axis, magnetic axis, line of sight to the observer, and width of any putative radio beam. For example, an anisotropic radio beam aligned to the magnetic axis must cross the line of sight to be detected, and longer-period rotation-powered pulsars tend to have narrower beams (e.g., Lyne & Manchester 1988; Rankin 1993). The radio nondetection of Swift J1555.2−5402 might suggest that this magnetar may not have a favorable geometry for detection (e.g., Lazarus et al. 2012). Some radio-loud magnetars have observational signatures that suggest aligned rotators (i.e., the angle between the magnetic axis and rotation axis is ≲30°; Camilo et al. 2008; Levin et al. 2012; Lower et al. 2020). As shown in Figure 4(c), the PF both in quiescence and during X-ray outbursts of radio-loud magnetars is lower than that of the other magnetars. A low PF could be due to radio-loud magnetars being observed as near-aligned rotators. The same interpretation can be inferred from the systematic Fourier-decomposition study of X-ray profiles of magnetars in quiescence conducted by Hu et al. (2019); radio-loud magnetars have a more sinusoidal pulse profile with a pronounced first Fourier component accompanied by a weaker second component. Another factor for the radio emission is the effects of higher-order (and possibly not curl-free) magnetic field components other than the dipolar field inferred from P–$\dot{P}$, which has already been taken into account. Quantum electrodynamic effects affect the conditions for pair cascades and radio emission when a magnetic field is sufficiently strong, above 10¹³ G. For example, if photon splitting occurs in a region above the critical magnetic field, the way the electron/positron cascade occurs is modified and perhaps quenched, and radio pulsation may be suppressed (e.g., Baring & Harding 2001). Finally, magnetar magnetospheres are considered to be dynamic, and the radio flux, pulse profile, and polarization swing pattern can change significantly within a few days (e.g., Lower et al. 2021). Future monitoring of this new magnetar, Swift J1555.2−5402, will be important for understanding the conditions for the magnetar radio emission.