Discovery of the New X-Ray Transient MAXI J1807+132: A Candidate of a Neutron Star Low-mass X-Ray Binary

After the first detection on 2017 March 13, MAXI J1807+132 has been monitored with the MAXI/GSC. Figure 2 presents background-subtracted 2–4 keV and 4–10 keV light curves, created through the method of image fitting (Morii et al. 2016) to the MAXI/GSC event data version 1.8. The source intensity reached a maximum at around MJD 57827–MJD 57832 (2017 March 15–20), and then gradually decreased. The averaged 4–10 keV intensity was ≈9 mCrab in that period, which is comparable with the peak intensity of the X-ray flaring event from 2MAXIt J1807+132 recorded in Kawamuro et al. (2016).

Zoom In Zoom Out Reset image size

A series of follow-up pointed observations of MAXI J1807+132 were carried out with Swift in the decaying phase. The net exposure was 0.5–2 ks in each observation. A log of these observations is given in Table 1, and the 0.3–10 keV XRT light curve is shown in Figure 2. We analyzed the XRT spectra taken during two weeks from the first Swift observation (March 26), using XSPEC version 12.9.1 m (Arnaud 1996). The light curve, spectra, and responses were downloaded via the online tools provided by the UK Swift Science Data Centre (Evans et al. 2009).⁷ The spectra taken on March 27 and 29, which have the best statistics among the present XRT data sets, were binned so that at least 30 counts are contained in each bin, and they were analyzed on the basis of the ${\chi }^{2}$ statistics. The other data, which have lower statistics, were grouped so that each bin has at least one count, and the Cash-statistics (Cash 1979) were adopted in the spectral analysis. The data taken on April 4 and 5 were omitted, because the source was not detected significantly. The data on April 6 and 7 were merged together to improve statistics; so were those on April 8 and 9. In the following analysis, the TBabs model is employed as the interstellar absorption model, with the table of solar abundances provided by Wilms et al. (2000).

As shown in Figure 3, we first analyzed the XRT spectra on March 27 and 31, which were obtained when the source intensity was relatively high and low, respectively. The spectrum on March 27 can be fit with an absorbed power-law model with a photon index of ${\rm{\Gamma }}={2.4}_{-0.1}^{+0.2}$ and a column density of ${N}_{{\rm{H}}}=(2.5\pm 0.6)\times {10}^{21}$ cm⁻², in agreement with the first report by Kennea et al. (2017a, 2017b). This model is fully acceptable (Figure 3(b)), with ${\chi }^{2}/\mathrm{dof}=63/65$. The resultant parameters are listed in Table 2. However, the estimated column density is somewhat higher than the total Galactic column, ${N}_{{\rm{H}}}=1.0\times {10}^{21}$ cm⁻² derived using the tool nh in HEASOFT version 6.19, and that calculated from the optical spectrum of the source (${N}_{{\rm{H}}}\sim 1.6\times {10}^{21}$ cm⁻²; Muñoz-Darias et al. 2017). The fit quality became worse (${\chi }^{2}/\mathrm{dof}=84/66$), when ${N}_{{\rm{H}}}$ is fixed at $1.0\times {10}^{21}$ cm⁻² (Figure 3(c)).

Zoom In Zoom Out Reset image size

Table 2.  Best-fit Parameters of Various Spectral Models for the Swift/XRT Data on March 27 and 31

Parameter	NH	Γ or τ	Npla	${R}_{\mathrm{bb}(\mathrm{comp})}$ b	kTbb or kTin	Rbb or Rinc	FXd	${\chi }^{2}/\mathrm{dof}$ e
Unit	1021 cm−2			km	keV	km	erg s−1 cm−2
Model: TBabs∗powerlaw
Mar 27	2.5 ± 0.6	${2.4}_{-0.1}^{+0.2}$	12 ± 1	⋯	⋯	⋯	$5.7\times {10}^{-11}$	63/65
Mar 27	1.0 (fixed)	2.07 ± 0.07	8.4 ± 0.4	⋯	⋯	⋯	$4.6\times {10}^{-11}$	84/66
Mar 31	$\lt 1.3$	${3.2}_{-0.4}^{+1.0}$	${0.7}_{-0.1}^{+0.4}$	⋯	⋯	⋯	$3.9\times {10}^{-12}$	19/18
Mar 31	1.0 (fixed)	3.9 ± 0.5	1.0 ± 0.2	⋯	⋯	⋯	$8.2\times {10}^{-12}$	21/19
Model: TBabs∗(diskbb+powerlaw)
Mar 27	1.0 (fixed)	${1.6}_{-0.7}^{+0.3}$	3 ± 2	⋯	${0.50}_{-0.06}^{+0.09}$	${1.8}_{-0.7}^{+0.3}$	$4.1\times {10}^{-11}$	60/64
Mar 31	1.0 (fixed)	${1.4}_{-1.3}^{+1.0}$	${0.2}_{-0.1}^{+0.3}$	⋯	0.15 ± 0.03	${19}_{-9}^{+17}$	$7.8\times {10}^{-12}$	10/17
Model: TBabs∗(bbodyrad+powerlaw)
Mar 27	1.0 (fixed)	1.8 ± 0.2	5 ± 1	⋯	${0.32}_{-0.04}^{+0.06}$	${4.2}_{-1.5}^{+1.6}$	$4.1\times {10}^{-11}$	59/64
Mar 31	1.0 (fixed)	1.5 ± 1.0	${0.2}_{-0.1}^{+0.3}$	⋯	0.12 ± 0.02	${32}_{-12}^{+22}$	$7.2\times {10}^{-12}$	10/17
Model: TBabs∗(bbodyrad+compps(bbodyrad))
Mar 27f	1.0 (fixed)	${2.5}_{-0.4}^{+0.5}$ g	⋯	${18}_{-3}^{+1}$	0.22 ± 0.05	$\lt 10$	$3.7\times {10}^{-11}$	62/64
Mar 27h	1.0 (fixed)	${0.59}_{-0.07}^{+0.08}$	⋯	${12.9}_{-1.7}^{+0.2}$	0.26 ± 0.03	$\lt 6$	$3.8\times {10}^{-11}$	64/64
Mar 31f	1.0 (fixed)	${3.0}_{-1.2}^{+0.0}$ g	⋯	$7\pm 0.2$	0.11 ± 0.02	${33}_{-12}^{+21}$	$6.3\times {10}^{-12}$	10/17
Mar 31h	1.0 (fixed)	${1.5}_{-1.3}^{+1.5}$ g	⋯	$7\pm 0.2$	0.12 ± 0.02	${30}_{-9}^{+20}$	$6.4\times {10}^{-12}$	10/17
Model: TBabs∗(diskbb+compps(diskbb))
Mar 27f	1.0 (fixed)	${3.0}_{-0.8}^{+0.0}$ g	⋯	${1.1}_{-0.1}^{+0.3}$	${0.41}_{-0.12}^{+0.07}$	$\lt 2.9$	$3.9\times {10}^{-11}$	60/64
Mar 27h	1.0 (fixed)	${0.7}_{-0.3}^{+2.3}$ g	⋯	1.2 ± 0.2	${0.44}_{-0.08}^{+0.10}$	$\lt 2.5$	$4.0\times {10}^{-11}$	59/64
Mar 31f	1.0 (fixed)	${3.0}_{-1.1}^{+0.0}$ g	⋯	${2.0}_{-0.2}^{+0.7}$	0.14 ± 0.03	${20}_{-9}^{+12}$	$6.8\times {10}^{-12}$	11/17
Mar 31h	1.0 (fixed)	${2.0}_{-1.7}^{+1.0}$ g	⋯	${1.8}_{-0.4}^{+0.7}$	0.15 ± 0.03	${17}_{-8}^{+16}$	$6.9\times {10}^{-12}$	10/17

Notes.

^aNormalization of the power-law component, in units of 10⁻³ photons keV cm⁻². ^bRadius of the emission region of the seed photons for Compton scattering, calculated from the photon flux of the compps component in 0.8–100 keV, by assuming a distance of 5 kpc and a spherical corona. ^cThe radius of emission region for the bbodyrad component, or the inner disk radius for the diskbb component. The distance and the inclination angle are assumed as 5 kpc and 0°, respectively. ^dThe unabsorbed X-ray flux in the 0.3–10 keV band. ^eC-statistic/dof is instead presented for the March 31 results. ^f ${{kT}}_{{\rm{e}}}=20\,\mathrm{keV}$ is assumed. ^gThe upper limit is pegged. ^h ${{kT}}_{{\rm{e}}}=100\,\mathrm{keV}$ is assumed.

Download table as:  ASCIITypeset image

As observed in Figure 3(a), the XRT spectrum changed significantly in shape from March 27 to 31. The March 31 spectrum appears to consist of a softer component dominant in $\lt 2\,\mathrm{keV}$, and a harder power-law-like component with a photon index of ≤2 extending above 2 keV. Then, the spectral change from March 27 to 31 can be understood if the latter component decreased by an order of magnitude, with a relatively small change in the former. As a confirmation, we forced a single power-law model to the March 31 spectrum. The fit was formally acceptable (Table 2), but as shown in Figure 3(e), the data in >2 keV systematically exceeded the best-fit power-law model, which was required to have a very steep (Γ ∼ 3.2) slope. We hence regard the single power-law model as inappropriate, both on March 27 and 31.

Based on the above consideration, we next fitted the spectra with a model composed of a power-law component and a thermal emission component: a blackbody (bbodyrad in XSPEC terminology) or a multi-color disk blackbody (diskbb Mitsuda et al. 1984). Here, ${N}_{{\rm{H}}}$ is assumed as $1\times {10}^{21}$ cm⁻². Then, the spectrum on March 27 has been well described by the model, with the reduced chi-squared values of ${\chi }^{2}/\mathrm{dof}=60/64$ and ${\chi }^{2}/\mathrm{dof}=59/64$ in the case of bbodyrad and diskbb, respectively. These values are almost the same as that obtained with a single power-law model in which N_H is left as a free parameter, while significantly smaller than that of the same model with ${N}_{{\rm{H}}}=1\times {10}^{21}$ cm⁻². The XSPEC script simftest shows a null-hypothesis probability of $\lt {10}^{-6}$ in the latter case. Both models gave relatively small photon indices: ${\rm{\Gamma }}=1.8\pm 0.2$ and ${1.6}_{-0.7}^{+0.3}$, with the bbodyrad and diskbb models, respectively.

The combination of the thermal and power-law components has successfully described the spectrum on March 31 as well. The reduced chi-squared values were slightly reduced from those of the single power-law model with ${N}_{{\rm{H}}}=1\times {10}^{21}$ cm⁻² (${\rm{\Delta }}{\rm{C}}$-statistic $\approx 11$ for ${\rm{\Delta }}\nu =2$), and the residual structure above $\approx 2\,\mathrm{keV}$ disappeared (Figure 3(g)). The probability that the improvement is only due to a random fluctuation is 0.004, according to simftest. The temperature of the soft thermal component became lower by a factor of ≈3, and the apparent linear size of its emission region became larger by a factor of 8–10 than those on March 27 (see Table 2).

The soft X-ray component is most likely optically thick thermal emission from the surface of a compact object (if it is not a black hole) or an accretion disk. The hard component, like in the present case, is often considered to originate via Comptonization of these thermal photons by a cloud of hot electrons (e.g., Done et al. 2007; Lin et al. 2007). Assuming that the observed hard tail is produced by Comptonization, we replaced the phenomenological power-law model with the compps model (Poutanen & Svensson 1996).

The compps model calculates a Comptonized spectrum when we specify the electron temperature kT_e, the optical depth for scattering τ, the energy distribution of the seed photons, and the geometry of the Comptonization cloud (or corona). In the present study, following previous works (e.g., Sakurai et al. 2014), a spherical corona (geom = 4) was assumed (but see below for the cases of different geometries), with only thermal electrons. We first tested the case where the seed photons are provided by blackbody emission. The model is expressed as TBabs∗(compps+bbodyrad), where the seed photon temperature kT_seed of the compps component was linked to kT_bb of the bbodyrad component. Because the data did not allow us to simultaneously constrain τ and kT_e, we fixed kT_e at 20 and 100 keV, and left τ as a free parameter. The other free parameters in this model are kT_bb and the normalizations of the bbodyrad and compps components. We ignored the reflection component from the disk.

This model, TBabs∗(compps+bbodyrad), fitted the two spectra well, and yielded the best-fit parameters given in Table 2. Both spectra favored slightly smaller values of R_bb and τ, if we assume ${{kT}}_{{\rm{e}}}=100\,\mathrm{keV}$, compared with the case of ${{kT}}_{{\rm{e}}}=20\,\mathrm{keV}$.

We also tested the alternative possibility that the seed photons of Comptonization are provided by disk blackbody emission. Thus, the bbodyrad component in the TBabs∗(compps+bbodyrad) model was replaced by diskbb, and the inner disk temperature kT_in of diskbb was set to be the same as kT_seed of compps (${{kT}}_{\mathrm{seed}}=-{{kT}}_{\mathrm{in}}$, in XSPEC terminology). This TBabs∗(compps+diskbb) model was also found to fit the two spectra well, yielding comparable reduced ${\chi }^{2}$ values to those of TBabs∗(compps+bbodyrad).

Although we have assumed a spherical corona above, following previous works, we have confirmed that the choice of the coronal geometry does not affect the main conclusions from the TBabs∗(compps+bbodyrad) and TBabs∗(compps+diskbb) models. If a slab or cylindrical corona is assumed (i.e., geom = 1 or 2, respectively), τ changes by a factor of ≲2, but the other free parameters remain unchanged within their 90% error ranges. The chi-squared values were also found to depend little ($| {\rm{\Delta }}{\chi }^{2}| \lesssim 1$) on the assumed coronal geometries.

We next investigated the spectral variation over a longer period. Figure 4 presents the other XRT spectra, in addition to those of March 27 and 31, which were already analyzed. The spectral profile varied in a complex manner: the peak energy of the soft component shifted toward higher energies from March 31 to April 6–7 and then moved back to lower energies in April 8–9, even though the X-ray flux was comparable among these three epochs. These XRT spectra were also well reproduced individually by the TBabs∗(powerlaw+bbodyrad) model.

Zoom In Zoom Out Reset image size

Figure 5 shows the best-fit parameters of these spectra in chronological order. It also shows the 3σ upper limits of the unabsorbed 0.3–10 keV flux on April 4 and 5, calculated by assuming the best-fit model on April 6. As suggested by the spectral shape changes, the variations of kT_bb and R_bb are rather complex and cannot be described as a simple function of the X-ray flux.

Zoom In Zoom Out Reset image size

The spectra were also fit well with the TBabs∗(powerlaw+diskbb) model. The derived kT_in and R_in varied in a similar manner to kT_bb and R_bb of the TBabs∗(powerlaw+bbodyrad) model, respectively.

We studied the behavior of the new X-ray transient MAXI J1807+132 using the multi-wavelength data of the MAXI/GSC, Swift, and optical telescopes. The source is likely to be identified with 2MAXIt J1807+132, which is listed in the first MAXI/GSC transient source catalog (Kawamuro et al. 2016) based on a long-term X-ray brightening event in 2011 May. Although Kawamuro et al. (2016) primarily aimed at a search for tidal disruption events (TDEs) by extragalactic supermassive black holes, the 2011 May episode of 2MAXIt J1807+132 was not categorized therein as a TDE. Below, we consider various interpretations of the nature of this object.

4.1.1. A Tidal Disruption Event?

Kawamuro et al. (2016) concluded that 2MAXIt J1807+132 is unlikely to be a TDE, because it has exhibited multiple enhancements (though with lower significances, at ∼2σ levels) from 2009 to 2013. Identifying the present source MAXI J1807+132 with 2MAXIt J1807+132, the TDE interpretation becomes even less likely, because the interval of the two strongest flaring events is much shorter than those predicted for TDEs (typically ∼10⁴ to ∼10⁵ years; e.g., Kawamuro et al. 2016).

To search for other brightening episodes of MAXI J1807+132 (=2MAXIt J1807+132), we analyzed the entire MAXI/GSC data of this sky region using the on-demand process system.⁹ Figure 9 shows the obtained light curve, from 2009 August to 2017 April. However, we detected no X-ray enhancements with a significance of >3σ, other than the flares in 2011 and 2017. As noticed in Figure 9, the present flaring event in 2017 is the brightest one in the last 7.5 years.

Zoom In Zoom Out Reset image size

The Swift/XRT data allow us to further argue against the TDE interpretation of the present object. The XRT spectra of MAXI J1807+132 in 2017 March and April have been well described with a soft thermal component (blackbody or disk blackbody) visible below ∼2 keV, and a hard tail with a photon index of ∼2. The thermal component has a temperature higher than those of TDEs (typically $\lesssim 0.1\,\mathrm{keV}$; Esquej et al. 2008; Maksym et al. 2010). The hard X-ray tail is much stronger than those of non-jetted TDEs, although it could be compatible with the spectrum of the jetted TDE Swift J164449.3+573451 (Bloom et al. 2011; Burrows et al. 2011; Levan et al. 2011).

Typical TDEs, either with or without jets, are considered to decay with time t as $\propto {t}^{-5/3}$ (e.g., Rees 1988; Phinney 1989). In contrast, the X-ray emission of MAXI J1807+132 in the present flaring event decayed much more rapidly; fitting the XRT light curve (Figure 5(a)) with a power-law model, we obtained the best-fit decay curve as $\propto {t}^{-6\pm 1}$, where t is the time since 2017 March 13 when the source was first recognized with MAXI. This fast decay also shows that the source is unlikely to be a TDE.

4.1.2. A Galactic Magnetic Cataclysmic Variable (CV)?

4.1.3. A High Mass X-Ray Binary (HMXB)?

4.1.4. A Transient BHXB?

4.1.5. A Neutron Star LMXB?

Let us examine the X-ray properties of MAXI J1807+132, assuming that it is a dim neutron star LMXB. The soft X-ray component of such an object, at luminosities below $\sim {10}^{35}$ erg s⁻¹, is generally considered to be thermal emission from the surface of the neutron star. The small radius of the blackbody emission region (a few km) can be naturally explained if only a part of the surface radiates X-rays. In fact, the blackbody radius of Aql X-1 was found to decrease from ∼10 km at $\sim 1\times {10}^{36}$ erg s⁻¹ down to ∼3 km at $\sim 1\times {10}^{34}$ erg s⁻¹ (Figure 6 of Sakurai et al. 2014), presumably because of the appearance of weak magnetic fields that limit the accretion flows to the magnetic poles. By contrast, the origin of the hard power-law tail is not yet fully understood. It allows several different interpretations, such as Comptonization of the blackbody emission in a hot accretion flow (Sakurai et al. 2014), bremsstrahlung from the hot flow itself (Chakrabarty et al. 2014), and the jet emission (Fender et al. 2003). In Section 2.2, we applied the compps model to the XRT spectra on March 27 and 31, to examine the first interpretation in comparison with the Suzaku results of Aql X-1 (Sakurai et al. 2014), and found similarities between their best-fit parameters (see also Section 4.1.5).

The contribution by bremsstrahlung from the Comptonized corona was evaluated quantitatively by Ono et al. (2017), based on the observationally estimated accretion flow geometry in Aql X-1. When the source luminosity is $\gt 5\times {10}^{36}$ erg s⁻¹, they estimated the bremsstrahlung luminosity to be ${L}_{\mathrm{Br}}\sim 1\times {10}^{34}$ erg s⁻¹. Then, assuming that MAXI J1807+132 has a typical luminosity of $2\times {10}^{35}$ erg s⁻¹ and that L_Br decreases as the mass accretion rate gets lower, we conclude that L_Br is likely to be still lower than the luminosity of the power-law tail.

We have detected significant variations in the temperature and the radius of the emission region of the blackbody component, which are not determined by the X-ray luminosity alone (see Figures 4 and 5). Similar peculiar behavior has been observed in other LMXBs, like Aql X-1 (Rutledge et al. 2002) and XTE J1709−267 (Jonker et al. 2005), at luminosities of 10³³–10³⁵ erg s⁻¹. These variations were suggested to arise from residual accretion onto the polar cap region of the neutron star, in association with neutron star cooling (e.g., Cackett et al. 2010; Degenaar & Wijnands 2012). In the case of MAXI J1807+132, however, it is unclear whether the neutron star was at the stage of crustal cooling during the Swift observations, provided that the timescale of its flux decay was somewhat shorter than those in other neutron star LMXBs in quiescence (Homan et al. 2015).

The spectra of MAXI J1807+132 could be categorized into two types: (1) a flatter profile with a higher kT_bb and a smaller R_bb (e.g., the March 27 spectrum), and (2) a more complex profile in which the soft component, with a lower kT_bb and a larger R_bb, and the hard tail are distinctive (e.g., the March 31 spectrum). This bimodality cannot be attributed to luminosity changes, considering that the spectrum switched a few times between the two types as the overall X-ray luminosity decayed almost monotonically.

One possible explanation of the observed variations between the two types would be to assume that the seed blackbody spectrum suffers strong color hardening, by a factor of $\kappa \sim 3$, in type-(1) spectra, whereas such an effect is small in type-(2) spectra so that the seed spectrum is close to a "bare" blackbody. In fact, Takahashi et al. (2009) observed spontaneous fluctuations in κ in the LMXB 4U 1608−52 (even though the effect was only ∼20% and was observed in the disk emission of the soft state). To examine this interpretation, we investigated the trend in the total photon flux of the soft thermal component, including both the direct and Compton-scattered components, assuming isotropic emission and conservation of the number of photons in Comptonization. In this analysis, the TBabs∗(bbodyrad+compps) model was applied to the individual XRT spectra with ${{kT}}_{{\rm{e}}}=100$ keV. As shown in Figure 10, the photon flux decreased rather monotonically. Thus, the observed variations could be described by a change in the color hardening factor (for some unspecified reasons) during a monotonic decrease in the mass accretion rate.

Zoom In Zoom Out Reset image size

The optical flux was reduced from March 27 to 31 by ∼0.4 mag per day as the X-ray flux decreased. The optical decay can be expressed as ${F}_{\mathrm{OPT}}\propto \exp (-t/{\beta }_{\mathrm{OPT}})$, where F_OPT represents the optical flux and ${\beta }_{\mathrm{OPT}}\sim 2.7$ day. Fitting the XRT 0.3–10 keV light curve (Figure 5) in March 26–31 with an exponential function, ${F}_{{\rm{X}}}\propto \exp (-t/{\beta }_{{\rm{X}}})$, we obtain ${\beta }_{{\rm{X}}}=2.4\pm 0.5$. From these two functions, the X-ray versus optical flux relation is derived as ${F}_{\mathrm{OPT}}\propto {F}_{{\rm{X}}}^{\alpha }$, where α ∼ 0.7–1.1. This α value is comparable with those obtained for other neutron star LMXBs in their dim (${L}_{{\rm{X}}}\lesssim {10}^{36}$ erg s⁻¹) periods (van Paradijs & McClintock 1994; Russell et al. 2006, 2007), which have been explained by the reprocessing of X-ray irradiation in the outer accretion disk. We also find that the optical and X-ray luminosities on March 27, estimated by assuming a distance of 5 kpc, lie on the trend in L_X and L_OPT of neutron star LMXBs in Russell et al. (2007). These results further reinforce the LMXB interpretation of MAXI J1807+132.