NuSTAR J095551+6940.8: a highly magnetised neutron star with super-Eddington mass accretion
MMon. Not. R. Astron. Soc. , 000–000 (0000) Printed 23 September 2018 (MN L A TEX style file v2.2)
NuSTAR J095551+6940.8: a highly-magnetised neutronstar with super-Eddington mass accretion
Simone Dall’Osso , Rosalba Perna , Luigi Stella Theoretical Astrophysics, University of T¨ubingen, Auf der Morgenstelle 10, 72076, Germany Department of Physics and Astronomy, Stony Brook University, Stony Brook, NY, USA INAF - Osservatorio Astronomico di Roma, via di Frascati 33, 00044, Monteporzio Catone, Roma, Italy
23 September 2018
ABSTRACT
The identification of the Ultraluminous X-ray source (ULX) X-2 in M82 as an ac-creting pulsar has shed new light on the nature of a subset of ULXs, while risingnew questions on the nature of the super-Eddington accretion. Here, by numericallysolving the torque equation of the accreting pulsar within the framework of the mag-netically threaded-disk scenario, we show that three classes of solutions, correspondingto different values of the magnetic field, are mathematically allowed. We argue that thehighest magnetic field one, corresponding to B ∼ G, is favoured based on physicalconsiderations and the observed properties of the source. In particular, that is the onlysolution which can account for the observed variations in ˙ P (over four time intervals)without requiring major changes in ˙ M , which would be at odds with the approximatelyconstant X-ray emission of the source during the same time. For this solution, we findthat the source can only accomodate a moderate amount of beaming, 0 . (cid:46) b < L X < . × erg s − fromarchival observations, is consistent with a highly-magnetized neutron star being in thepropeller phase at that time. Key words: stars: neutron – pulsars: general – stars: magnetic fields.
Bright X-ray sources, not associated with galactic nuclei,were discovered during surveys in the late 1970s and early1980s with the
ROSAT and
Einstein telescopes. These sur-veys showed that the X-ray luminosity function of typ-ical galaxies extends up to luminosities on the order of ∼ erg s − (e.g., Fabbiano 1989 for a review). The con-tribution to the bright end of the luminosity distributionwas revealed to be made up of individual sources thanks tothe superior spatial resolution and sensitivity of the Chandra telescope (e.g. the case of the Antennae; Zesas et al. 2002).In addition, observations with
Chandra also revealed thatsome early type galaxies, while having no sign of nuclear ac-tivity, still displayed the presence of individual sources withluminosities in the ∼ − erg s − range (e.g. Horn-schemeier et al. 2004; Ptak & Colbert 2004).These point-like sources, emitting X-rays at values ex-ceeding the Eddington limit for a solar mass compact object,have been collectively dubbed ultraluminous X-ray sources(ULXs). Their physical nature has long been debated inthe literature, with a variety of models and ideas put for-ward to explain their properties. These include, among oth-ers, accreting intermediate-mass black holes with masses ∼ − M (cid:12) (e.g. Colbert & Mushotzky 1999; Milleret al. 2003), beamed X-ray binaries (King et al. 2001),young supernova remnants in extremely dense environments(Marston et al. 1995), super-Eddington accretors (e.g. Begel-man et al. 2006), young, rotation-powered pulsars (Perna &Stella 2004). It is possible that ULXs form a heterogeneousfamily consisting of different sub-classes (e.g. Gladstone etal. 2013).The recent detection of 1.37 s pulsations in the X-rayflux of an ULX in the M82 galaxy has allowed to identifythis source as an accreting magnetized NS (Bachetti et al.2014). The average period derivative during five of the NuS-tar observations indicated that the source is spinning up,as typical of a NS accreting matter from a companion star.The super-Eddington luminosity of the source could then beaccounted for by a combination of geometrical beaming andreduced opacities due to a strong magnetic field. However,Bachetti et al. (2014) highlighted a possible inconsistencybetween the value of the magnetic field inferred from mea-surements of the period derivative and that inferred fromthe X-ray luminosity, if due to accretion.In this paper we carry out a detailed modeling of thetiming properties of the source; we show that, by applyingthe Ghosh & Lamb (1979; GL in the following) model for c (cid:13) a r X i v : . [ a s t r o - ph . H E ] J a n S. Dall’Osso et al. magnetically-threaded disks in its general formulation, boththe period derivative and the X-ray luminosity of the sourcecan be self-consistently accounted for. Our work comple-ments and/or extends recent results which have appearedwhile we were working on our paper (Eksi et al. 2014; Lyu-tikov 2014; Kluzniak & Lasota 2014; Tong 2014).
The pulsations seen from the source and the measured spin-up indicate that an accretion disk exists around the NS, andthat it should be truncated at some distance from the NSwhere matter is forced to flow towards a localised region onthe stellar surface. The NS magnetic field is responsible fortruncating the disk at the magnetospheric radius, r m , andfor channeling the accretion flow along the B-field lines.We can write the X-ray luminosity, L X , in terms of theaccretion luminosity, L acc , asL X = η L acc = η GMR ns ˙ M , (1)where ˙ M is the mas accretion rate and η ∼ unity is theefficiency of conversion of released accretion energy to X-rays. The “isotropic” equivalent X-ray luminosity can thenbe expressed as L X, iso = L X /b = L acc ( η/b ), where b < X, iso = 10 erg s − (Bachetti etal. 2014) we obtain˙ m = ˙ M ˙ M E ≈ bη (cid:18) L X, iso erg s − (cid:19) , (2)where the Eddington luminosity is L E =4 πcGMm p /σ T (cid:39) . × ( M/ . M (cid:12) ) erg s − , and˙ M E = L E R ns / ( GM ) (cid:39) . × − ( R ns /
10 km) M (cid:12) yr − .Finally the disk luminosity, L disk = ( GM ˙ M ) / r m isL disc L E (cid:39) . (cid:18) ˙ m (cid:19) (cid:18) r m /R ns (cid:19) − . (3) When an accretion disk feeds a magnetised NS, the stellarmagnetic field is expected to thread the disk with field lines,which are twisted by the differential rotation in the (Keple-rian) disk. This interaction between magnetic field lines anddisk plasma allows a continuous exchange of angular mo-mentum between the NS and the disk, via Maxwell stressesassociated to twisted field lines. This couples the evolutionof the NS spin and of the mass accretion rate (GL).Maxwell stresses are strongest in the inner disk, andbecome the main mechanism for angular momentum re-moval inside the “magnetospheric radius”, which is definedas r m = ζ (cid:104) µ / (2 GM ˙ M ) (cid:105) / . The magnetic dipole momentof the NS is µ = B p R /
2, in terms of the magnetic fieldstrength at the magnetic pole, B p . We will adopt the normal-isation (in c.g.s. units) µ = 0 . p , , where Q x ≡ Q/ x . This expression holds for hydrogen. A factor
A/Z should beadded for a generic element of atomic weight A and number Z . At r m the disk is truncated and the plasma must flow alongmagnetic field lines. In the GL model ζ ≈ .
52 and, in gen-eral, 0 . < ζ < K ( r m ) > Ω spin . Defining the co-rotation radius r cor as the point where Ω K ( r cor ) = Ω spin , the condition for ac-cretion becomes r m < r cor . In this case the NS accretes thespecific angular momentum of matter at r m , thus receivinga spin-up torque N acc = ˙ M ( GMr m ) / , i.e. the accretion-induced spin-up torque in the absence of threading .The outer parts of the disk, at r > r cor , rotate slowerthan the NS thus extracting its angular momentum as theydrag the more rapidly-rotating magnetic field lines. This re-gion applies a spindown torque on the NS, with an amplitudethat depends on the microphysics of the interaction betweendisk and magnetic field.The total torque on the NS is thus the sum of these twoopposing torques. The resulting spin up is in general lessthan for a non-threaded disk, at a given accretion rate, dueto the negative contribution of regions where r cor < r < r out .We can formally write N tot = N acc ˆ n , (4)where magnetic threading is entirely encoded in the “torquefunction” ˆ n whose value is, in general, smaller than unity. As shown by GL, the torque function depends on the rele-vant radii, r m and r cor , only through their ratio. Definingthe fastness parameter ω s = Ω spin / Ω K ( r m ) = ( r m /r cor ) / ,one can write n = ˆ n ( ω s ).In general, n ( ω s ) should be a decreasing function of ω s ,reaching zero at some critical value ω c . If ω s < ω c then thetorque function is positive, while n ( ω s ) < ω c . The exact shape of n ( ω s ) depends on several details of the disk-magnetosphereinteraction and can vary among different models (e.g., Wang1995, 1997; Yi et al. 1997; Erkut et al. 2005). On quantitativegrounds, however, the main conclusions depend only weaklyon such details. We will therefore adopt the GL model fromhere on to keep the discussion focused, and discuss laterways to distinguish the alternatives.An analytic approximation for the torque function inthis model is (GL79b) n ( ω s ) ≈ .
39 1 − ω s (cid:2) . − ω s ) . − . (cid:3) − ω s , (5)which goes to zero for ω c ≈ .
35. When this condition ismet, for r m ≈ . r cor , the total torque is zero and the NSaccretes at constant spin. At a fixed magnetic field, a furtherdecrease in the mass accretion rate would increase r m , inturn increasing ω s and causing n ( ω s ) to become negative:the NS would spin down while accreting.We write the total torque as N tot = N n ( ω s ) ω / s , (6)where N = ˙ M ( GMr co ) / . The function n ( ω s ) ω / s is dis-played in Fig.1 vs. the dipole moment µ , for two valuesof the accretion rate. For each positive value of the function c (cid:13) , 000–000 ccreting NS as ULX m = m = P = μ - - ( μ , M )× ω s / Figure 1.
The product (see text) ω / s n ( ω s ) vs. the magneticdipole moment, µ , for two selected values of ˙ M in the GL model,adopting the spin period P=1.37 s of NuSTAR J095551+6940.8.When the function is positive, hence N tot >
0, two possible µ ’sexist for each value of ω / s n ( ω s ). For negative values of the func-tion the degeneracy is broken, and only one value of µ is possible. (and hence of N tot ), two solutions for µ are obtained. Theirseparation increases for increasing ˙ M . On the other hand,when the function (and hence N tot ) is negative, only onevalue of µ is allowed. In the GL model, the properties of standard, gas pressure-dominated, geometrically thin accretion disks are assumed(Shakura & Sunyaev 1979; SS hereafter). Hence a directapplication of the GL model to NuSTAR J095551+6940.8,with its super-critical mass accretion rate, needs to be dealtwith care.Formally, the standard SS model does not apply in thecase of super-Eddington accretion (Abramowicz 2004 andreferences therein); the so-called “slim disks” (Abramowiczet al 1988; Beloborodov 1998; Sadowski et al 2009, 2011)should be used instead, which generalise the SS model toa wider range of conditions . The crucial difference is thatslim disks become radiatively inefficient at some criticalradius, inside which a growing share of the locally-releasedgravitational energy is advected with the flow rather thanbeing radiated away. Beloborodov (1998) provides a simpleway to estimate this effect, and compares detailed solutionsof relativistic slim disks with radial profiles of standard SSdisks up to large radii . The largest deviations from the SSdisks occur at r ∼ (20 − r g , i.e. ∼ × cm for a NS,in the case ˙ m bl ∼ ∼ (5 − Abramowicz (2004) notes that even the slim disk can be re-garded as an higher-order approximation, in h/r , to the mostgeneral case of the “polish doughnut” (or fat torus). Radial profiles extend down to r ∼ r g in the recent work bySadowski et al. (2011), much smaller than needed for our discus-sion here (see sec. 3). Also note that, due to different notations,˙ m defined here is 4.8 smaller than ˙ m bl of Beloborodov (1998) and3.3 times larger than ˙ m sd of Sadowski et al. (2011). assuming a viscosity parameter α ∼ . Figure 2.
The locus of points in the µ − ˙ M plane satisfying Eq. (7)for the averaged value of the measured ˙ P . The equilibrium spincondition of the GL model is shown as the red, dashed line. Threemain regimes are identified: 1) low- B case, ∼ × − G,for mass accretion rates 10 (cid:46) ˙ m (cid:46)
37; 2) B -fields ∼ G, for10 (cid:46) ˙ m (cid:46)
60; 3) Intermediate values of B ∼ − G only fora narrow range of mass accretion rates around the minimum, 7 (cid:46) ˙ m (cid:46)
10. The inset highlights the region (above the green dashedline) for which r m > R ns and hence pulsations are possible. for the case ˙ m sd ∼
10. At r = r m ∼ cm (which isrelevant to our specific situation, see. Sec.4.3), the deviationfrom Keplerian rotation is tiny, as are the deviations of allrelevant quantities from the SS model. We conclude thatthe accretion disk is well described by the SS model if itsinner radius r m (cid:38) several × cm.In addition, the disk is likely to remain slightlyradiation-pressure-dominated out to large radii, r (cid:38) cm.Radiation pressure can introduce changes in the disk’s den-sity and height profile compared to standard GL-type mod-els, and in the exact dependence of r m on ˙ M and µ (Ghosh1996). While a quantitative analysis is beyond our scopehere, in sec. 3.4 we provide a qualitative discussion of theseeffects, along with other sources of uncertainty in models forthe disk-magnetic field coupling. Let us now focus our discussion on the ULX NuSTARJ095551+6940.8, whose pulsations at ∼ .
37 s have beenrecently discovered. In order to characterise the main fea-tures of the solution for this source, we first consider the˙ P ≈ − × − s s − reported by Bachetti et al. (2014) asrepresentative of the average over different measurements. In § P measured c (cid:13) , 000–000 S. Dall’Osso et al. by Bachetti et al. (2014) to further constrain the allowedrange of parameters.The total torque N tot is related to the instantaneousperiod derivative,N tot = n ( ω s ) ω / N = 2 πI ˙ ν , (7)where I is the NS moment of inertia and ˙ ν (cid:39) . × − s − . A numerical solution to this equation yields ω s ,which identifies a locus of points in the µ vs. ˙ M plane. Fig. 2illustrates the solution for the average ˙ P quoted above: foreach value of ˙ M two different values of the B field are al-lowed , reflecting the property of positive torques discussedin sec. 2.2 and Fig. 1.The inset in the left panel of Fig. 2 highlights the “for-bidden” area (below the green line) in which r m (cid:46) R ns andthe disk would reach the NS surface, thus quenching thepossibility to produce sizeable pulsations from the source. We define for convenience three main regions in the solution,identified by the vertical dashed line in Fig. 2 at ˙ m = 10.The narrow range 7 < ˙ m <
10 corresponds to a wide rangeof magnetic fields: 10 G < B p < × G. This is the“intermediate- B ” region, or low- ˙ m region.For larger values of ˙ m we have two clearly distinctbranches, a “low- B ” region where B p (cid:46) G and a “high- B ” region where 5 × G < B p (cid:46) × G.The low- B branch is far from spin equilibrium: r m < cm is rather close to the NS and spin-down torques arenegligible. The high- B branch is close to spin equilibrium,with the disk truncated very far from the NS ( r m /R ns ∼ − r cor andproduces a spin-down torque. The intermediate- B solutionis not much farther from spin equilibrium, at least for B p (cid:38) G, yet this difference will prove to be crucial.Further inspection of Fig. 2 shows that: (a) the low- B region exists only if B p (cid:38) × G, or elsethe disk reaches the NS surface, quenching the pulsations.(b) The µ − ˙ M curve turns around at ˙ M ∼ M E , corre-sponding to µ ∼
1, or B p ∼ × G. Eq. (2) thusimplies that b (cid:38) .
12, indicating that the source must bemildly beamed. Even in the most favorable case, its actualX-ray luminosity is L X (cid:38) erg s − , which has impor-tant implications ( § ˙ P fluctuations Table 3 of Bachetti et al. (2014) lists variations in the mea-sured ˙ P of up to one order of magnitude at different epochs(see caption of Fig. 3 for the individual values). A changein the sign of ˙ P (spin-down) at one particular epoch (obs.007) may have also been seen.We note that the variations of the pulse phase/shape Note that this is true in the threaded disk model, but not in amodel without magnetic threading of the disk. Note that the exact numerical value of this lower limit for B p is specific to the GL model. correlated with variations in the mass accretion rate maymimick the effects of spin period derivatives, potentially af-fecting the results of phase-coherent timing analyses. Thepossibility of such mimicking effects was discussed in rela-tion to accreting millisecond pulsars in transient low massX-ray binaries, where the pulsations could be studied over asource flux decrease of typically one order of magnitude ona timescale of tens of days (Patruno et al. 2009; Patruno2010). In order to interpret the measured period deriva-tives ˙ ν ∼ − s − in these systems, accretion-rate inducedchanges in the pulse profile would be required which cancause a cumulative phase shifts of ∼ . − . (cid:46) ∼ . − . ν ∼ afew × − s − would result, comparable to the largest measurement errors,e.g. those of obs. 007. We conclude that the measured ˙ P vari-ations observed in NuSTAR J095551+6940.8 are intrinsicto the source and amount to at least an order of magnitudefor only modest variations of the source flux. For the sake ofdefiniteness, we will therefore adopt in the following the val-ues reported by Bachetti et al. (2014), keeping in mind thatthe positive ˙ P (spin-down) measurement (obs. 007) mightnot be significant, in consideration of its error bar and thediscussion above.In the following, we apply condition (7) to each of thefour ˙ P values reported by Bachetti et al. (2014), and showthe corresponding solutions in Fig. 3. These will provideadditional constraints on the characteristics of the source. We discuss here the relative merits of the different solutionsand constrain the physical properties of the hyper-accretingNS in NuSTAR J095551+6940.8. × G < B p (cid:46) G) This solution corresponds to 10 < ˙ m (cid:46)
37, hence a highlysuper-Eddington luminosity L X > × erg s − , compat-ible with a modest amount of beaming b > .
2. The innerdisk radius ranges from r m (cid:38) R ns to r m (cid:39) ns .With the inner edge of the disk close to the NS, mag-netic threading is essentially irrelevant. The NS is only sub-ject to the material torque, N acc ≈ ˙ M ( GMr m ) / ∝ ˙ M / .Fluctuations of ˙ P by one order of magnitude can only beexplained by changes in ˙ M of comparable amplitude, instark contrast with the relatively stable X-ray emission ofthe source during the same period. Furthermore, the briefphase of torque reversal (or equivalently very low torque)revealed during obs. 007 would be impossible in these con-ditions (see the left panel of Fig. 3).Moroever, since B ∼ (cid:28) G, it would very hardto exceed L X = 2 × erg s − , even for an optically thickaccretion column (Basko & Sunyaev 1976a) c (cid:13) , 000–000 ccreting NS as ULX Δ M P = + ( ) × - P = - ( )× - P = - ( )× - P = - ( )× - P = s M M E μ Δ M P = + ( ) × - P = - ( )× - P = - ( )× - P = - ( )× - P = s M M E μ Figure 3.
Same as in Fig. 2, but using the four different values of ˙ P reported by Bachetti et al. (2014) in four distinct observations. Thered dashed line indicates the equilibrium condition N tot = ˙ ν = 0. Note that no solution exists in the low- B branch for the equilibrium caseor the positive ˙ P . The vertical dashed lines indicate a representative case for the variations in mass accretion rate needed to reproducethe observed ˙ P fluctuations. Left panel: the complete solution highlighting the high- B branch, with ∆ ˙ M/ ˙ M ∼
20 %;
Right panel: theintermediate- B region with ∆ ˙ M/ ˙ M ∼ G < B p < × G) This case corresponds to mass accretion rates in a narrowrange, ˙ m ∼ −
10 (Fig 2), which in turn implies a significantbeaming of the emitted radiation (eq. 2), 0 . (cid:46) b ≤ . X ∼ (1 − × erg s − ,marginally consistent with the theoretical maximum for anoptically thick accretion column (Basko & Sunyaev 1976a).Similarly to the low- B case, this regime cannot explainthe large ˙ P oscillations. Solutions for all four measured val-ues of ˙ P only exist for µ >
1, or B p > × G (Fig. 3,right panel). Even in that case, however, fluctuations of ˙ M by a factor of ∼ − M in Fig. 3, right panel),in contradiction with the fairly stable source emission.We note also that a beaming factor in b ∼ . − .
15 isnot easily achieved even in the optically thick environmentof sources accreting at (cid:38) ˙ M E (Basko & Sunyaev 1976b):NuSTAR J095551+6940.8 would thus represent an extremeexample of this issue. × G < B p (cid:46) × G) This solution corresponds to 10 < ˙ m <
60, the upper boundbeing the accretion rate that gives the luminosity L X , iso ∼ erg s − for a negligible beaming ( b ∼
1, Eq. 2). On theupper branch in Fig. 2 the implied B -field is in the range(0 . − × G. The inner disk radius is r m ∼ (80 − ns , the fastness parameter ω s ≈ (0 . − .
31) and the diskluminosity L disk ∼ (0 . − . E . At this distance fromthe NS the disk has a small thickness h ( r m ) /r m (cid:46) . M can explain the large variations in ˙ P . Re- quired ˙ M variation get progresively smaller for higher mag-netic fields: the left panel of Fig. 3 shows that, for µ =10, values of ˙ m ∼ (50 −
60) can reproduce all the observedtorque variations, including the torque reversal episode.From an independent argument, the isotropic X-ray lu-minosity L X , iso ∼ erg s − can be attributed to a reduc-tion in the electron scattering cross-section, caused by thestrong B -field (Basko & Sunyaev 1976a, Paczynski 1992).The scattering cross-section scales as ( E γ /E cyc ) , where E γ is the photon energy and E cyc (cid:39) . B p , keV is thecyclotron energy. For photons at E γ ∼
30 keV, a B -field ∼ G would be required to reduce the cross-section bya factor ∼
50. For the maximum luminosity from a plasmawhich is optically thick to electron scattering, and integrat-ing over photon energy and propagation angle, Paczynski(1992) derived, L E , mag = 2 L cr (cid:18) B p G (cid:19) / , (8)where L cr represents the limiting luminosity in the ab-sence of magnetic fields. Substituting L E and requiring thatL E , mag = L X , iso = 10 erg s − , gives B p ∼ . × G,in line with the value obtained from the torque analysis. Asdiscussed above, the geometry of an accretion column canincrease L cr a few times above L E , thus reducing the require-ment on B p . For a maximum L cr ∼ erg s − we wouldstill need B p ∼ × according to Eq.(8).The above arguments point independently to the samerange of values of the B -field, 5 × G (cid:46) B (cid:46) × G,and to only a modest degree of beaming (0 . (cid:46) b < The dependence on the direction of propagation is neglectedhere for simplicity.c (cid:13) , 000–000
S. Dall’Osso et al.
Therefore, the high- B solution provides a satisfactory in-tepretation of the source properties. Before concluding our discussion, we should note that al-ternative models for the disk-magnetic field interaction leadto slightly different locations of the inner disk, r m , and aslightly different dependence of the torque function n ( ω s )on the fastness parameter. Adopting e.g., the model by Yiet al. (1995; also discussed by Andersson et al. 2005), where ω c ≈ .
71, the low- B branch is unaffected by the changewhile the high- B branch - which is the most affected - isshifted upwards by (cid:46)
30 % at most.This is to be expected given that the low- B solutionis far from the equilibrium condition (N tot = 0), and thusit is essentially insensitive to magnetic threading, while thehigh- B solution is most sensitive to the details of torque bal-ance, being close to spin equilibrium. This also offers a keyto understand the impact of radiation-pressure in GL-typemodels. The low- B branch would still be mostly unaffectedbecause, as stated before, it is insensitive to the details ofmagnetic threading. For B p > G (including both theintermediate- B and the high- B case) the inner disk will beslightly radiation pressure dominated ( p rad /p gas ∼ . − m = (10 − To further clarify our discussion we briefly compare our re-sults with those recently presented in the literature. In thediscovery paper by Bachetti et al. (2014) the condition ofspin equilibrium was assumed and translated into the sim-ple equality r co = r m . Assuming a purely material torque,this was used in combination with the average value of ˙ P toconstrain the accretion rate. However, as noted by the sameauthors, the resulting value of ˙ M falls short of what requiredto explain the source luminosity. In order to alleviate thistension, Tong (2014) discussed possible caveats such as thecombination of a moderate beaming, a large NS mass and acomplex magnetic field structure.Eksi et al. (2014) and Lyutikov (2014), on the otherhand, considered a magnetically-threaded disk and, still as-suming the source to be close to spin equilibrium, estimatedthe NS field to be ∼ G and 10 G, respectively. Thedifference in these values comes largely from the differentsource luminosity that these authors adopted. This range of B -fields corresponds to our “high- B ” solution (sec. 3.3.3).Lasota & Kluzniak (2014) noted that the equilibriumcondition need not be verified, and discuss a solution inwhich the NS has a low magnetic field ( ∼ G) and ac-cretion rate ˙ M ∼
50 ˙ M E . This corresponds to our “low-B”solution (sec. 3.3.1).In our work we considered a magnetically-threaded diskwithout prior assumptions on how close the system is tospin equilibrium. By using the average value of ˙ P we haveshown that a continuum of solutions in the B vs. ˙ M planeis allowed, and the cases discussed above represent specificexamples, as already noted.The key addition in our study is that we took into ac-count the large fluctuations in the measured values of ˙ P , which are associated with small fluctuations of the sourceluminosity, as found by Bachetti et al. (2014). Unless varia-tions of the beaming factor are invoked to keep the luminos-ity stable in spite of large variations in ˙ M , this behaviour iscompatible only with the NS being close to the spin equilib-rium condition. This singles out the high- B solution, witha preferred value of ∼ G which is lower than the onederived by Eksi et al. (2014), and agrees with the estimateof Lyutikov (2014). Such a field is strong enough to affectradiative transfer in the emission region and allow a signif-icantly super-Eddington luminosity. Further implications ofthis scenario are briefly commented in the next section.
Our work has provided strong arguments in favor of the in-terpretation of NuSTAR J095551+6940.8 as a highly mag-netized NS with a magnetic field ∼ G, a mass accretionrate ∼ (20 −
50) ˙ M E corresponding to an accretion luminos-ity ∼ (0 . − × erg s − , and little beaming, if at all(0 . (cid:46) b < (cid:46) (80 −
90) R ns , and helps maintain-ing the large luminosity at the base of the accretion col-umn. For a simple dipole geometry, the field has decreasedby a factor ∼ A em ∼ cm , asrequired for thermal radiation at the observed luminosity topeak around/close to the spectral range in which NuSTARJ095551+6940.8 has been observed ( <
50 keV). Let us con-sider ˙ m = 20 and freely falling matter with v ff ∼ . c at afew kilometres from the NS surface. The particle density inthe flow is n ∼ ˙ M/ ( v ff A em m p ) ∼ ( ˙ m/ A − , cm − which implies τ = nσ T l ∼ m/ l /A em , ) where, tobe conservative, we allowed for an elongated shape of theflow ( l < √ A em , cf. Basko & Sunyaev 1976a), and assumeda relatively low value of ˙ m . Radiation from below will con-tinuously ablate the outer layers of the accretion columnto a scattering depth τ ∼ a few and, even if the particleswere flung away close to the speed of light, the resultingoutflow could only amount to a tiny fraction of the inflow ∼ τ − c/v ff ∼ . r m ∼ cm and h ( r m ) /r m ∼ . c (cid:13) , 000–000 ccreting NS as ULX reported by Feng & Karet (2007) and Kong et al. (2007).The source was found to alternate between very brightstates, with L X ∼ (0 . − . × erg s − , and dimundetected “off” states, below the sensitivity level of theChandra observations (this corresponds to a luminosity of ∼ . × erg s − ). Given that the system is close tothe condition for spin equilibrium ( r m (cid:39) . r cor in theGL model, and ∼ . − . r m (cid:38) r cor (cid:39) × cm. Eq. (3). We note that the expected disk luminosity L disk ∼ . L E ∼ × erg s − , would well below presentupper limits to luminosity during the source ”off” state.Since the magnetospheric radius scales as ˙ m − / , a decreaseof ˙ m by a factor ∼
10 with respect to the level at which thesource emits at L X ∼ erg s − would be needed in or-der to cause the transition to the propeller regime in the GLmodel. A decrease by only a factor (cid:38) r m = 0 . r cor , as envisaged inother disk-magnetosphere interaction models. The expectedtransition luminosity is ∼ erg s − and 5 × erg s − ,respectively, in the two cases.Note that, if the magnetic field is significantly lower, theinner edge of the disk would extend closer to the NS andlarger decreases in ˙ m would be required for the propellermechanism to set in. Correspondingly the source would vis-ible over a wider range, before the suddenly switchinh tomuch lower luminosities. Hence we argue that all flux levelsso far measured in NuSTAR J095551+6940.8are consistentwith the scenario involving of a ∼ G NS.The X-ray spectral characteristics of the source canprovide additional clues on the B-field strength of NuS-TAR J095551+6940.8. The power law-like X-ray spectrumof many accreting X-ray pulsars was found to bend steeplyabove a cutoff energy, E cut , which is empirically relatedto the energy of cyclotron resonance features throughE cyc (1 . − . cut (Makishima et al 1990). If the samerelationship holds for NuSTAR J095551+6940.8, its X-rayspectrum would be expected not to show any cutoff up toenergies of at least ∼
50 keV, if B (cid:38) G. On the otherhand, a solution with B (cid:46) × G would be characterisedby a spectral cutoff at energies (cid:46) (20 −
40) keV.We finally note that the almost circular orbit of thesystem ( e (cid:46) . For a constant mass accretion rate We caution however that for very high accretion ratescyclotron-related spectral features might shift to lower energies,as observed e.g. in the highest luminosity states of 4U0115+63and V0332+53 (Nakajima et al. 2006; Tsygankov et al. 2006).This might be due to an increased height of the post-shock re-gion of the accretion column. possible to use the low eccentricity alone to constrain theNS age. However, if it were possible to constrain the age ofthe system through, e.g. population synthesis studies, thenthis would offer the intriguing possibility to also test theoriesof magnetic field evolution in highly magnetised NSs.In fact, detailed magnetothermal simulations (Vigan´oet al. 2013) show that the lifetime of the NS dipole fieldstrongly depends on where the field is located, and on thepresence or not of a strong toroidal component in additionto the dipole. Crustal fields tend to decay faster than thoserooted in the NS core, due to the different electrical con-ductivities, and additional toroidal components increase therate of dissipation, making the dipolar field decay faster thanin cases with no toroidal field. Internal fields that are mostlyconcentrated in the (superconducting) core have the slow-est rate of dissipation. Therefore, the lifetime of a ∼ Gfield could be (cid:46) (cid:38) yrs if thefield were largely confined to the core. ACKNOWLEDGEMENTS
For this work S.D. was supported by the SFB/Transregio7, funded by the Deutsche Forschungsgemeinschaft (DFG).RP was partially supported by NSF grant No. AST 1414246and Chandra grants (awarded by SAO) G03-13068A andG04-15068X (RP). LS acknowledges discussions with P.G.Casella, G.L. Israel and A. Papitto. LS acknowledges partialsupport by PRIN INAF 2011.
REFERENCES
Abramowicz, M. A., Czerny, B., Lasota, J. P., Szuszkiewicz, E.1988, ApJ, 332, 646Abramowicz, M. A., 2004, arxiv:astro-ph/0411185v1Bachetti, M. et al. 2014, Nature, 514, 202Basko, M. M. & Sunyaev, R. A. 1976a, MNRAS, 175, 395Basko, M. M. & Sunyaev, R. A. 1976b, Sov. Astr., 20, 537Begelman, M. C. King, A. R. & Pringle, J. E. 2006 MNRAS, 370,399Belczynski, K., Kalogera, V., Rasio, F. A., Taam, R. E., Zezas,A., Bulik, T., Maccarone, T. J.; Ivanova, N. 2008, ApJS, 174,223Beloborodov, A. 1998, MNRAS, 297, 739Colbert, E. J. M., & Mushotzky, R. F. 1999, ApJ, 519, 89Eksi, K. Y., Anda, I. C., Cikintoglu, S., Gencali, A. A., Gungor,C., Oztekin, F. 2014, arXiv1410.5205Erkut, M. H., Alpar, M. A. 2005, ApJ, 620, 390Fabbiano, G. 1989, AR&A, 27, 87Feng, H., Kaaret, P. 2007, ApJ, 668, 941Gladstone, J. C., Copperwheat, C., Heinke, C. O., Roberts, T.P., Cartwright, T. F., Levan, A. J., Goad, M. R. 2013, ApJS,206, 14Frank, J., King, A., Raine, D. J. 2002, Accretion Power in Astro-physics, Cambridge University PressGhosh, P. 1996, ApJ, 459, 244Ghosh, P., Pethick, C. J., Lamb, F. K. 1977, ApJ, 217, 578Ghosh, P., Lamb, F. K. 1979, ApJ, 234, 296Hornschemeier, A. E. et al. 2004, ApJ, 600L, 147King, A. R., Davies, M. B., Ward, M. J., Fabbiano, G., Elvis, M.2001, ApJ, 552, L109Kluzniak, W., Lasota, J.-P. 2014, arXiv1411.100c (cid:13) , 000–000
S. Dall’Osso et al.
Kong, A. K. H., Yang, Y. J., Hsieh, P.-Y., Mak, D. S. Y., Pun,C. S. J. 2007, ApJ, 671, 349Lyutikov, M. 2014, arXiv1410.8745Makishima, K. et al. 1990, ApJ, 365L, 59Marston, A. P., Elmegreen, D., Elmegreen, B., Forman, W.,Jones, C. & Flanagan, K. 1995, ApJ, 483, 663Miller, J. M., Fabbiano, G., Miller, M. C. & Fabian, A. C. 2003,ApJ, 585, L37Nakajima, M., Mihara, T., Makishima, K., Niko, H. 2006, ApJ,646, 1125Paczynski, B. 1992, AcA, 42, 145Patruno, A., Wijnands, R., & van der Klis, M. 2009, ApJ, 698,L60Patruno, A. 2010, ApJ, 722, 909Perna, R., Stella, L. 2004, ApJ, 615, 222Psaltis, D. & Chakrabarty, D. 1999, ApJ, 521, 332Ptak, A., & Colbert, E. 2004, ApJ, 606, 291Sadowski, A., Abramowicz, M. A., Bursa, M., Kluzniak, W.,Rzanska, A., Straub, O. 2009, A&A, 502, 7Sadowski, A., Abramowicz, M., Bursa, M., Kluzniak, W., Lasota,J.-P., Rzanska, A 2011, A&. 527, 17Shakura, N. I., Sunyaev, R. A. 1973, A&A, 24, 337Tong, H. 2014, arXiv1411.3168Tsygankov, S. S., Lutovinov, A. A., Churazov, E. M., Sunyaev,R. A. 2006, MNRAS, 371, 19Verbunt, F., Phinney, E. S. 1995, A&A, 296, 709Vigan´o, D., Rea, N., Pons, J. A., Perna, R., Aguilera, D. N.,Miralles, J. A. 2013, MNRAS, 434, 123Wang, Y.-M. 1995, ApJ, 449L, 153Wang, Y.-M. 1997, ApJ, 487L, 85Yi, I. Wheeler, J. C., Vishniac, E. T. 1997, ApJ, 481L, 51Zesas, A., Fabbiano, G., Rots, A. H., Murray, S. S. 2002, ApJ,577, 710 c (cid:13)000