0805.1849/ms.tex
1: \documentclass[onecolumn]{emulateapj}
2: %\documentclass[onecolumn]{emulateapj}
3: 
4: %\usepackage{amssymb}
5: %\usepackage{amsmath}
6: %\usepackage{xspace} 
7: %\usepackage{graphicx} 
8: %\usepackage{natbib}
9: \def\IGR{IGR\,J17544-2619}
10: \def\xte{XTE\,J1739-302}
11: \def\igg{IGR\,J16479-4514}
12: \def\axj{AX\,J1841.0-0536}    
13: \def\igrjj{IGR\,J16465-4507} 
14: \def\igrj11{IGR\,J11215-5952}
15: \def\ax{AX\,J1749.1-2733}
16: \def\iigr{IGR\,J16418-4532}
17: \def\2s{2S\,0114+65}
18: \def\igrpatel{IGR\,J16358-4726}
19: \def\chan{{\em Chandra}}
20: \def\xmm{{\em XMM-Newton}}
21: \def\int{{\em INTEGRAL}}
22: \def\swift{{\em SWIFT}}
23: \shorttitle{Magnetars and SFXTs}
24: 
25: \begin{document}
26: 
27: \title{Are There Magnetars in High Mass X-ray Binaries? The Case of SuperGiant Fast X-Ray Transients}    
28: 
29: \author{E. Bozzo\altaffilmark{1,2}, 
30:         M. Falanga\altaffilmark{3},
31:         L. Stella\altaffilmark{1}
32:        }
33: 
34: \altaffiltext{1}{INAF - Osservatorio Astronomico di Roma, Via Frascati 33,
35: 00044 Rome, Italy} 
36: \altaffiltext{2}{Dipartimento di Fisica - Universit\`a di Roma ``Tor
37:   Vergata'', via della Ricerca Scientifica 1, 00133 Rome, Italy}
38: \altaffiltext{3}{CEA Saclay, DSM/DAPNIA/Service d'Astrophysique (CNRS FRE
39: 2591), F-91191, Gif sur Yvette, France}
40: 
41: \shorttitle{Magnetars and SFXTs}
42: \shortauthors{E. Bozzo, et al.}
43: 
44: \begin{abstract}
45: In this paper we survey the theory of wind accretion in high mass X-ray binaries 
46: hosting a magnetic neutron star and a supergiant companion.
47: We concentrate on the different types of interaction between the inflowing 
48: wind matter and the neutron star magnetosphere that are relevant when 
49: accretion of matter onto the neutron star surface is largely inhibited;  
50: these include the inhibition through the centrifugal and magnetic barriers.  
51: Expanding on earlier work, we calculate the expected luminosity for each regime and 
52: derive the conditions under which transition from one regime to another can take place. 
53: We show that very large luminosity swings ($\sim$10$^{4}$ or more
54: on time scales as short as hours) can result from transitions across different 
55: regimes. 
56: The activity displayed by supergiant fast X-ray transients, 
57: a recently discovered class of high mass X-ray binaries in our galaxy,
58: has often been interpreted in terms of direct accretion onto a neutron star
59: immersed in an extremely clumpy stellar wind. 
60: We show here that the transitions across the magnetic and/or centrifugal barriers 
61: can explain the variability properties of these sources as a results of 
62: relatively modest variations in the stellar wind velocity and/or density. 
63: According to this interpretation we expect that supergiant fast X-ray transients 
64: which display very large luminosity swings and host a slowly spinning neutron star 
65: are characterized by magnetar-like fields, irrespective of whether the magnetic 
66: or the centrifugal barrier applies. Supergiant fast X-ray transients might 
67: thus provide a new opportunity to detect and study magnetars in binary systems. 
68: \end{abstract} 
69: 
70: 
71: \keywords{accretion, accretion disks --- stars: neutron --- supergiant --- X-rays: binaries --- X-rays: stars}
72: 
73: 
74: 
75: \section{Introduction}
76: \label{sec:intro}
77: 
78: High mass X-ray binaries (HMXBs) consist of a collapsed object,
79: usually a magnetic neutron star (NS), that accretes matter from an OB companion star. 
80: Mass transfer takes place because of the intense stellar wind from the OB 
81: star, part of which is captured by the collapsed object \citep[e.g.][]{verbunt}. 
82: Only in some short orbital period systems, the early type star, often a 
83: supergiant, fills its Roche lobe and leads to mass transfer through 
84: Roche lobe overflow \citep{tauris}. Persistent HMXBs accrete all the time and in most 
85: cases display X-ray luminosities in the 10$^{35}$-10$^{38}$~erg~s$^{-1}$ range. 
86: Many HMXBs are transient systems that remain at low 
87: X-ray luminosity levels (10$^{32}$-10$^{33}$~erg~s$^{-1}$) most of the time 
88: and undergo outbursts lasting from weeks to months. 
89: During these outbursts they display nearly identical properties 
90: to those of persistent HMXBs. 
91: Transient systems usually comprise a Be star donor and relatively 
92: long, moderately eccentric orbits, such that the star sits deep in its 
93: Roche lobe and stellar wind capture is the only mechanism through which 
94: mass transfer takes place. The occurrence of the outbursts is likely 
95: associated to variations in the stellar wind of the Be star, such 
96: as shell ejection episodes, or build up of matter around the 
97: resonant orbits in the slow equatorial wind component \citep{heuvel}.   
98: However, there are characteristics of the outbursts that are difficult to 
99: interpret if accretion onto the neutron star surface takes place 
100: unimpeded also in quiescence; these are (a) the large 
101: outburst to quiescence X-ray luminosity swing (factor of $\sim$10$^3$ or 
102: larger) and (b) the presence in
103: a given source of low-luminosity (Type I) outbursts recurring close to 
104: periastron and, at different times, of high-luminosity (Type II) outbursts 
105: that last for several orbital cycles and display little (if any) X-ray flux 
106: variations associated to the orbital phase. These characteristics of 
107: Be transients can be explained if the accretion rate (and thus X-ray 
108: luminosity) variations that are produced by the stellar wind alone, could 
109: be amplified by some ``gating'' mechanism. Since most Be star HMXB transients
110: contain relatively fast spinning X-ray pulsars, such mechanism has been 
111: identified with the centrifugal barrier that 
112: results from the rotation of the neutron star magnetosphere \citep{stella86}. 
113: 
114: About 10 transient systems have been recently 
115: discovered, which display sporadic outbursts lasting from 
116: minutes to hours (i.e. much shorter than Be star transients') 
117: and reach peak luminosities of $\sim$10$^{36}$-10$^{37}$~erg~s$^{-1}$.   
118: These systems spend long time intervals in quiescence, with X-ray 
119: luminosities down to $\sim$10$^{-5}$ times lower than those in outburst; 
120: in spite of their association with OB supergiant companions, 
121: their behaviour is thus at variance with other persistent 
122: and transient HMXBs. They define a new class of HMXBs, collectively 
123: termed supergiant fast X-ray transients, SFXTs.
124: An overview of the properties of SFXTs is given in
125: \S~\ref{sec:properties}.
126: 
127: If accretion onto the collapsed object of SFXTs takes place both 
128: in quiescence and outburst, then the corresponding X-ray luminosity 
129: swing, typically a factor of $\sim$10$^4$-10$^5$, would require  
130: wind inhomogeneities
131: with a very large density and/or velocity contrast   
132: \citep[according to the standard wind accretion, the mass capture 
133: rate onto the NS scales like $\dot{M}_{\rm w}$v$_{\rm w}^{-4}$, 
134: with $\dot{M}_{\rm w}$, the mass loss rate
135: and v$_{\rm w}^{-4}$, the wind velocity of the supergiant star,][]{davidson}. 
136: Several authors \citep{zand05,leyder,walter07} 
137: suggested the presence of dense clumps in the wind of the OB 
138: companions in order to attain the luminosity variations of SFXTs. 
139: While some observations provide evidence for a clumpy wind, the 
140: characteristics of such inhomogeneities are still poorly known. 
141: Numerical simulations suggest that clumps may originate from small 
142: scale perturbations in the radiation-driven wind \citep{dessart03,prinja}. 
143: Models involving accretion of clumps are still being 
144: actively pursued \citep{negueruela08,walter07}.   
145: The requirement on the density and/or velocity contrasts 
146: in the wind can be eased if there is a barrier
147: that remains closed during quiescence, halting
148: most of the accretion flow, and opens up 
149: in outbursts, leading to direct accretion 
150: (this is similar to the case of Be star transients). 
151: This paper develops and discusses gated accretion models for SFXTs.  
152: 
153: In \S~\ref{sec:model} after reviewing the 
154: theory of wind accretion in HMXBs, we describe the different regimes 
155: of a rotating magnetic neutron star immersed the stellar wind 
156: from its companion. 
157: In \S~\ref{sec:trans} we 
158: discuss the conditions under which transitions across
159: regimes can take place in response to variations in the wind parameters. 
160: As compared to previous works addressing gating mechanisms
161: in transient accreting magnetic neutron stars \citep[notably][]{stella86}, 
162: we present here a more comprehensive treatment of the different physical 
163: processes that have been discussed
164: in this context. 
165: 
166: In \S~\ref{sec:results}, we present an application  
167: to the different states of two 
168: SFXTs and discuss in turn the possibility that the 
169: outbursts are driven by a 
170: centrifugal or a magnetic barrier. 
171: The latter mechanism involves a magnetosphere 
172: extending beyond the accretion radius \citep{bondi},  
173: which prevents most of the inflowing matter from accreting; 
174: it is discussed here for the first time in relation to the activity 
175: of transient binary sources. 
176: Unless the stellar wind were extremely clumpy (as envisaged in
177: other SFXT models), the onset of a barrier inhibiting direct accretion
178: would be required to explain the activity of SFXTs. In this case  
179: we conclude that if SFXTs host slowly rotating neutron stars 
180: (spin periods of several hundreds to thousands seconds), then they 
181: must possess magnetar-like fields 
182: ($\sim$10$^{14}$-10$^{15}$~G), independent of whether the centrifugal 
183: or magnetic barrier operates.  
184: We summarize our discussion and conclusions in \S~\ref{sec:discussion} 
185: and \S~\ref{sec:conclusions}. 
186: 
187: \begin{figure*}[t!]
188: \centering
189: \includegraphics[height=5.4 cm]{f1.eps}
190: \smallskip
191: \caption{Schematic view of a magnetized NS interacting with the 
192: inflowing matter from its supergiant companion. All the regimes of 
193: \S~\ref{sec:model} are shown, together with the relative 
194: position of the magnetospheric 
195: radius (solid line), the corotation radius (dashed line), and the accretion 
196: radius (dotted line). A wavy solid line is used when the magnetospheric 
197: boundary at R$_{\rm M}$ is Kelvin-Helmholtz unstable. In the supersonic and subsonic 
198: propeller regime convective motions at the base of the atmosphere are 
199: represented with small eddies.}
200: \label{fig:model} 
201: \end{figure*}
202: 
203: 
204: 
205: \section{The observed properties of SFXTs}
206: \label{sec:properties}
207: 
208: SFXTs are observed to exhibit sporadic outbursts, lasting from minutes to hours, 
209: with peak X-ray luminosities between $\sim$10$^{36}$ and 10$^{37}$~erg~s$^{-1}$ 
210: \citep[see e.g.,][]{gon04,sidoli05,grebenev05,Lutovinov05,sguera05,masetti06,
211: sguera06,gotz07,sguera07}. 
212: No firm orbital period measurement  
213: has been obtained yet\footnote{Only \objectname{IGR J11215-5952} and \objectname{AX J1749.1-2733}   
214: showed recurrent flaring activity, with periodicity of $\sim$165~d and $\sim$185~d, respectively. 
215: These are interpreted as outbursts from two systems with unusually long orbital 
216: periods \citep[$\gtrsim$100~d,][]{grebenev05,sidoli07,zurita07}. 
217: Thus \citet{walter07} excluded these sources from their SFXT list.}. 
218: A recent list of confirmed ($\sim$5) and candidate ($\sim$6)  
219: SFXTs is given by \citet{walter07}. 
220: 
221: Between outbursts, SFXTs remain in quiescence with 
222: luminosities in the range $\sim$10$^{31}$-10$^{33}$~erg~s$^{-1}$  
223: \citep{gon04,zand05,smith06,kennea06}. 
224: In some cases, very high peak-to-quiescence X-ray luminosity swings 
225: (factor of $\sim$10$^{4}$-10$^{5}$) were seen    
226: on timescales comparable to the outburst duration. 
227: Some SFXTs showed also flare-like
228: activity at intermediate luminosity levels \citep[e.g.,][]{gon04}.  
229: In the case of \objectname{IGR J17544-2619}, two states of intermediate luminosity 
230: were observed: one before the onset of the outburst and the other immediately after,  
231: with X-ray luminosities $\sim$3 and $\sim$1 decades below 
232: the value reached at the peak of the outburst, respectively \citep{zand05}. 
233: 
234: The sporadic character of SFXT outbursts, as observed with \int,\ suggested that the 
235: duty cycle of these sources (the fraction of time spent in a high luminosity state) 
236: is small \citep[$\sim$0.02-0.002,][]{walter07}. 
237: However, recent observations carried out with the very sensitive X-ray telescopes on board 
238: \xmm\ and \chan\ revealed that some SFXTs display flares around a luminosity of 
239: $\gtrsim$10$^{34}$-10$^{35}$~erg~s$^{-1}$ (i.e. well below the \int\ 
240: limiting sensitivity) for a large fraction of the time
241: \citep{gon04,zand05,tomsick06}.  
242: Therefore, the indication is that the active phase of SFXT sources (as opposed to true
243: quiescence) lasts longer than previously thought, and the duty cycles are of order 
244: $\sim$0.1 or higher.
245: 
246: Optical identifications of SFXTs show that these sources are
247: associated to OB supergiant companion stars \citep[see e.g.,][and 
248: reference therein]{walter07}. The SFXT 
249: OB companions have typically mass of   
250: M$_{*}$$\sim$30M$_{\rm \odot}$, luminosity of 
251: log (L$_{*}$/L$_{\odot}$)$\sim$5-6, mass loss 
252: rate of $\dot{M}_{\rm w}$=10$^{-7}$-10$^{-5}$~M$_{\odot}$~yr$^{-1}$,  
253: and wind velocity of 1000-2000~km~s$^{-1}$. 
254: Note, that isolated OB stars with log(L$_{*}$/L$_{\odot}$)$\sim$5-6 
255: are persistent soft X-ray sources with luminosity 
256: around $\sim$10$^{32}$~erg~s$^{-1}$ \citep{cassinelli81,berghofer97}.  
257: 
258: It is widely believed that SFXTs contain sporadically accreting 
259: neutron stars. Only little is known about their spin period.  
260: A coherent periodicity was detected at 4.7~s and  228~s 
261: in \objectname{AX J1841.0-0536} and \objectname{IGR J16465-4507},  
262: respectively \citep{bamba01,Lutovinov05}. However, \igrjj\  
263: showed only a factor of $\sim$100 luminosity swing 
264: between quiescence and outburst, so it is unclear whether the source 
265: should be considered a transient  
266: \citep[in fact it is classified as an ``intermediate system'',]
267: []{walter07}. The nature of the companion star in \axj\ is 
268: still debated \citep{halpern04,nespoli07}. 
269: Therefore, these two sources might not belong 
270: to the SFXT class. 
271: On the contrary, in the prototypical SFXTs 
272: \objectname{XTE J1739-302} \citep{sguera06} and \objectname{IGR J16479-4514} \citep{walter06} 
273: some evidence has been reported for periodicities in the 
274: $\sim$1000-2000~s range. We assume in the following that SFXTs host a  
275: rotating magnetic neutron star. 
276: 
277: 
278: \section{Stellar wind accretion}
279: \label{sec:model} 
280: 
281: We investigate here the conditions under which a magnetized 
282: neutron star can accrete matter from the 
283: wind of a massive companion. 
284: In the theory of wind accretion in HMXBs, the following radii are defined
285: \citep[see e.g.,][]{illarionov75, stella86}: 
286: \begin{itemize}
287: 
288: \item The accretion radius, R$_{\rm a}$ is the distance at which the 
289: inflowing matter is gravitationally focused toward the NS \citep{bondi}. 
290: It is usually expressed as   
291: \begin{equation}
292: R_{\rm a}=2GM_{\rm NS}/v_{\rm w}^2= 3.7\times10^{10} v_{8}^{-2} ~{\rm cm},
293:  \label{eq:ra}
294: \end{equation}
295: where v$_{8}$ is the wind velocity in units of 1000~km~s$^{-1}$ and we 
296: assumed that the orbital velocity of the star 
297: is negligible \citep{fkr}. Throughout the paper we fix the NS radius and 
298: mass at R$_{\rm NS}$=10$^6$~cm and M$_{\rm NS}$=1.4~M$_\odot$, respectively. 
299: The fraction $\dot{M}_{\rm capt}$/$\dot{M}_{\rm w}$ of the stellar wind mass loss rate 
300: ($\dot{M}_{\rm w}$) captured by the NS depends on R$_{\rm a}$ through 
301: \citep{fkr} 
302: \begin{equation}
303: \dot{M}_{\rm capt}/\dot{M}_{\rm w}\simeq R_{\rm a}^2/(4 a^2)=2\times10^{-5} v_{8}^{-4} 
304: a_{\rm 10d}^{-2}. 
305: \label{eq:dotmcapt}
306: \end{equation}
307: Here a=4.2$\times$10$^{12}$a$_{\rm 10d}$~cm is the orbital separation, 
308: a$_{\rm 10d}$=P$_{\rm 10d}^{2/3}$M$_{30}^{1/3}$, P$_{\rm 10d}$ is the binary orbital 
309: period in units of 10 days, and M$_{30}$ is the total 
310: mass in units of 30~M$_{\odot}$ (we assumed circular orbits). 
311:  
312: \item The magnetospheric radius, R$_{\rm M}$, at which the pressure of the NS magnetic 
313: field ($\mu^2$/(8$\pi$$R_{\rm NS}^6$), with $\mu$ the NS magnetic moment) 
314: balances the ram pressure of the inflowing matter 
315: ($\rho_{\rm w}$$v_{w}^{2}$). 
316: In the case in which R$_{\rm M}$$>$R$_{\rm a}$, 
317: the magnetospheric radius is given by \citep{pringle}  
318: \begin{equation}
319: R_{\rm M}=3.3\times10^{10} \dot{M}_{-6}^{-1/6} v_{8}^{-1/6} a_{\rm 10d}^{1/3} 
320: \mu_{33}^{1/3} ~ {\rm cm}.   
321: \label{eq:rm}
322: \end{equation} 
323: Here we assumed a non magnetized spherically symmetric wind \citep{elsner1977},  
324: with density\footnote{We approximated a-R$_{\rm M}$$\simeq$a ,  
325: which is satisfied for a very wide range of parameters.} 
326: $\rho_{\rm w}$(R$_{\rm M}$)$\sim$$\dot{M}_{\rm w}$/(4$\pi$a$^2$v$_{\rm w}$), 
327: a dipolar NS magnetic field with $\mu_{33}$=$\mu$/10$^{33}$~G~cm$^{3}$, and 
328: $\dot{M}_{-6}$=$\dot{M}_{\rm w}$/10$^{-6}$ M$_{\odot}$~yr$^{-1}$. 
329: In the following sections we discuss the range of applicability of Eq.~\ref{eq:rm}, 
330: and the regimes in which a different prescription for R$_{\rm M}$ should be 
331: used.  
332: 
333: \item The corotation radius, R$_{\rm co}$, at which the NS angular velocity 
334: equals the Keplerian angular velocity, i.e. 
335: \begin{equation}
336: R_{\rm co}=1.7\times10^{10} P_{\rm s3}^{2/3} ~{\rm cm}.    
337: \label{eq:rco}
338: \end{equation}
339: Here $P_{\rm s3}$ is the NS spin period in units of 10$^3$~s. 
340: \end{itemize}
341: 
342: Changes in the relative position of these radii
343: result into transitions across different regimes for the NS 
344: \citep{illarionov75, stella86}. Below we discuss these regimes 
345: singularly, and provide a schematic representation 
346: of each regime in Fig.~\ref{fig:model}. 
347: Being determined primarily by the spin 
348: of the neutron star, the corotation radius can change only over 
349: evolutionary timescales. \citet{illarionov75} summarises 
350: the different regimes experienced by a spinning down NS since
351: its birth, from the initial radio pulsar (i.e. rotation powered) stage, 
352: to the regime in which mass accretion onto the neutron star surface can 
353: take place. On the other hand the accretion radius and magnetospheric 
354: radius depend on the wind parameters (see Eqs.~\ref{eq:ra} and \ref{eq:rm}),
355: which can vary on a wide range of timescales (from hours to months).  
356: Therefore, variations in the wind parameters can cause the neutron 
357: star to undergo transitions across different regimes on comparably 
358: short timescales, thus opening the possibility to explain the 
359: properties of some classes of highly variable X-ray sources through them. 
360: In particular, the transition from 
361: the accretion regime to the propeller regime (and vice versa), across the 
362: so-called ``centrifugal barrier'', was identified as a likely 
363: mechanism responsible for the pronounced activity of Be X-ray pulsar 
364: transient systems \citep{stella86}.  
365: Below we summarise the different regimes of a 
366: magnetic rotating neutron star, subject to a varying 
367: stellar wind, with special attention to the condition under 
368: which a ``magnetic'' (as opposed to ``centrifugal'')
369: barrier inhibits accretion onto the neutron star
370: \citep{hard, ruth, mori, toropina, toropina2}. 
371: As it will be clear in the following, new motivation for 
372: investigating magnetic inhibition of accretion comes from 
373: the discovery of magnetars, neutrons stars with 
374: extremely high magnetic fields \citep[$\sim$10$^{14}$-10$^{15}$~G,][]{dun}. 
375: 
376: 
377: \subsection{Outside the accretion radius: the magnetic inhibition of accretion: 
378: R$_{\rm M}$$>$R$_{\rm a}$}
379: 
380: We consider here the case in which the magnetospheric radius is larger than 
381: the accretion radius\footnote{A similar case was considered  
382: also by \citet{lipunov02}, ``the georotator regime'', and by \citet{toropina2},  
383: ``the magnetic plow regime''.}.  
384: In systems with R$_{\rm M}$$>$R$_{\rm a}$ the mass flow from the companion star interacts 
385: directly with the NS magnetosphere without significant gravitational focusing, forming  
386: a bow shock at R$_{\rm M}$ \citep{hard, toropina2}. 
387: A region of shocked gas surrounds the NS magnetosphere with density 
388: $\rho_{\rm ps}$$\simeq$4$\rho_{\rm w}$ and velocity 
389: v$_{\rm ps}$$\simeq$v$_{\rm w}$/4 (the subscript ``ps'' stands for 
390: post-shock). These are only rough estimates because 
391: the shock is very close to the magnetopause and it does not satisfy the 
392: standard Rankine-Hugoniot conditions \citep{toropina2}. At least 
393: in the front part of the shock, i.e. in the region around the  
394: stagnation point, the whole kinetic energy of the inflowing matter is    
395: converted into thermal energy, and the expected temperature of the heated gas is  
396: T$\simeq$m$_{\rm p}$v$_{\rm w}^2$/(3k)$\simeq$4$\times$10$^7$v$_{8}^2$~K. 
397: Thus, the power released in this region is of order  
398: \begin{equation}
399: L_{\rm shock}\simeq\frac{\pi}{2} R_{\rm M}^2 \rho_{\rm w} v_{\rm w}^3 = 
400: 4.7\times10^{29} R_{\rm M10}^2 
401: v_8^2 a_{\rm 10d}^{-2} \dot{M}_{-6}~{\rm erg ~s^{-1}}  
402: \label{eq:lshock}
403: \end{equation}
404: (R$_{\rm M10}$ is the magnetospheric radius in units of 10$^{10}$~cm), 
405: and is mainly radiated in the X-ray band \citep{toropina}. 
406: Below we distinguish two different regimes of magnetic inhibition of accretion.  
407: 
408: \subsubsection{The superKeplerian magnetic inhibition regime: R$_{\rm M}$$>$R$_{\rm a}$, 
409: R$_{\rm co}$}
410: \label{sec:mprop} 
411:   
412: In this ``superKeplerian'' magnetic inhibition regime the magnetospheric radius is larger than 
413: both the accretion and corotation radii (R$_{\rm M}$$>$R$_{\rm a}$, R$_{\rm co}$).
414: Matter that is shocked and halted close to R$_{\rm M}$ cannot 
415: proceed further inward, due to the rotational drag of the NS 
416: magnetosphere which is locally superKeplerian. 
417: Since magnetospheric rotation is also supersonic\footnote{This can be easily 
418: seen by comparing v$_{\rm w}$ with $\Omega$R$_{\rm M}$, for the 
419: value of the parameters used in this section.}, the interaction 
420: between the NS magnetic field and matter at R$_{\rm M}$ results in 
421: rotational energy dissipation and thus, NS spin down. 
422: In order to derive an upper limit on the contribution of this dissipation to the overall 
423: luminosity, we assume that the above interaction is anelastic \citep[e.g.,][]{perna06}, 
424: i.e. that matter at R$_{\rm M}$ is forced to corotate. This process releases energy at 
425: a rate 
426: \begin{equation}
427: L_{\rm sd} \simeq \pi R_{\rm M}^2 \rho_{\rm w} v_{\rm w} (R_{\rm M}\Omega)^2\simeq  
428: 3.7\times10^{29} R_{\rm M10}^4 \dot{M}_{-6} a_{\rm 10d}^{-2} P_{\rm s3}^{-2}~{\rm erg ~s^{-1}}, 
429: \label{eq:lmagnsd}
430: \end{equation}
431: which adds to the shock luminosity (Eq.~\ref{eq:lshock}). 
432: 
433: 
434: \subsubsection{The subKeplerian magnetic inhibition regime: R$_{\rm a}$$<$R$_{\rm M}$$<$R$_{\rm co}$}
435: \label{sec:intermediate}
436: 
437: If R$_{\rm a}$$<$R$_{\rm M}$$<$R$_{\rm co}$ the magnetospheric drag is subKeplerian
438: and matter can penetrate the NS magnetosphere. 
439: In this ``subKeplerian'' magnetic inhibition regime, the boundary between  
440: the inflowing matter and the magnetosphere is subject to the Kelvin-Helmholtz 
441: instability \citep[KHI,][]{hard}. The mass inflow rate across R$_{\rm M}$ resulting
442: from the KHI is approximately \citep{bur} 
443: \begin{eqnarray}
444: \dot{M}_{\rm KH}\simeq2\pi R_{\rm M}^2 \rho_{\rm ps} v_{\rm conv} = 
445: 2 \pi R_{\rm M}^2 \rho_{\rm ps} v_{\rm sh} 
446:  \eta_{\rm KH} (\rho_{\rm i}/\rho_{\rm e})^{1/2}  
447: (1+\rho_{\rm i}/\rho_{\rm e})^{-1},  
448: \label{eq:dotmkh}
449: \end{eqnarray} 
450: where $\eta_{\rm KH}$$\sim$0.1 is an efficiency factor,  
451: v$_{\rm sh}$ is the shear velocity, $\rho_{\rm i}$ and $\rho_{\rm e}$ the density inside 
452: and outside the magnetospheric boundary at R$_{\rm M}$, respectively.  
453: Close to the stagnation point, virtually all the kinetic energy 
454: of the wind matter is converted into thermal energy (see also \S~\ref{sec:mprop}), and 
455: the shear velocity is thus dominated by the magnetosphere's rotation, 
456: such that v$_{\rm sh}$=v$_{\rm rot}$=2$\pi$P$_{\rm s}^{-1}$R$_{\rm M}$. 
457: Away from this region, the tangential component of the wind velocity with respect 
458: to the NS magnetic field lines increases up to v$_{\rm ps}$ and, in general, 
459: the shear velocity and rate at which plasma enters the magnetosphere 
460: due to the KHI, depends on both v$_{\rm ps}$ and v$_{\rm rot}$. 
461: For the aims of this paper, we adopt v$_{\rm sh}$=max(v$_{\rm ps}$, v$_{\rm rot}$), 
462: and use mass conservation across the KHI unstable layer to estimate the 
463: density ratio at R$_{\rm M}$ \citep{bur}. If matter crossing the unstable layer of height h$_{\rm t}$ 
464: is rapidly brought into corotation with the NS magnetosphere and free-falls onto 
465: the NS, mass conservation implies 
466: \begin{equation}
467: R_{\rm M}^2 \rho_{\rm e} v_{\rm conv}\simeq R_{\rm M} h_{\rm t} \rho_{\rm i} v_{\rm ff}(R_{\rm M}). 
468: \label{eq:masscons}
469: \end{equation} 
470: The height h$_{\rm t}$ of the unstable layer, where matter and magnetic field coexist, is mostly 
471: determined by the largest wavelength of the KHI unstable mode \citep{bur}. 
472: A detailed analysis of this instability is beyond the scope of this paper. 
473: In Eq.~\ref{eq:masscons} we use conservatively h$_{\rm t}$$\simeq$R$_{\rm M}$, 
474: and discuss in Appendix~\ref{app:ht} the effect of smaller values of h$_{\rm t}$.   
475: Therefore accretion of matter at a rate $\dot{M}_{\rm KH}$ onto the NS is  
476: expected to release a luminosity of  
477: \begin{eqnarray}
478: L_{\rm KH} = 3.5\times10^{34} \eta_{\rm KH} R_{\rm M10}^2 a_{\rm 10d}^{-2} \dot{M}_{-6} 
479: (\rho_{\rm i}/\rho_{\rm e})^{1/2} (1+\rho_{\rm i}/\rho_{\rm e})^{-1}~{\rm erg~s^{-1}}, 
480: \label{eq:lx_kh}
481: \end{eqnarray}
482: if v$_{\rm sh}$=v$_{\rm ps}$, or 
483: \begin{eqnarray}
484: L_{\rm KH} = 8.8\times10^{34} \eta_{\rm KH} P_{\rm s3}^{-1} R_{\rm M10}^3 
485: a_{\rm 10d}^{-2} v_8^{-1} \dot{M}_{-6} 
486: (\rho_{\rm i}/\rho_{\rm e})^{1/2} (1+\rho_{\rm i}/\rho_{\rm e})^{-1} ~{\rm erg~s^{-1}}, 
487: \label{eq:lx_kh2}
488: \end{eqnarray}
489: if v$_{\rm sh}$=v$_{\rm rot}$. 
490: The values of $\rho_{\rm i}$/$\rho_{\rm e}$ that we use in Eqs.~\ref{eq:lx_kh} and \ref{eq:lx_kh2} 
491: are derived numerically from Eq.~\ref{eq:masscons}. 
492: 
493: In the subKeplerian magnetic inhibition regime, plasma penetration inside the 
494: NS magnetosphere is sustained also by Bohm diffusion \citep{ikhsanov}. 
495: This diffusion, being dependent on the temperature of the plasma (rather than velocity), 
496: is highest in the region close to the stagnation point, where the shock slows down 
497: the inflowing plasma most efficiently (see also \S~\ref{sec:mprop}). 
498: In accordance with \citet{ikhsanov2001}, the maximum inflow rate  
499: allowed by Bohm diffusion is 
500: \begin{eqnarray}
501: \dot{M}_{\rm diff} \simeq 2\pi R_{\rm M}^2 \rho_{\rm ps} V_{\rm m} = 
502: 4.5\times10^{9} \zeta^{1/2} \dot{M}_{-6} \mu_{33}^{-1/2} R_{\rm M10}^{11/4}  
503: a_{\rm 10d}^{-2} ~{\rm g ~s^{-1}}, 
504: \label{eq:dotmdiff} 
505: \end{eqnarray}
506: where V$_{\rm m}$=$\sqrt{D_{\rm eff}/t_{\rm ff}}$ is the diffusion velocity, 
507: D$_{\rm eff}$=($\zeta$ckT$_{\rm i}(R_{\rm M})$)/(16eB(R$_{\rm M}$)) the diffusion coefficient, 
508: T$_{\rm i}$(R$_{\rm M}$)$\simeq$m$_{\rm p}$v$_{\rm w}^2$/(3k) the post-shock ion temperature, 
509: t$_{\rm ff}$=$\sqrt{R_{\rm M}^3/(2GM)}$ the free-fall time, $\zeta$$\simeq$0.1  
510: an efficiency factor and m$_{\rm p}$ the proton mass. In the above equation we also 
511: approximated the density outside R$_{\rm M}$, around the stagnation point, with 
512: $\rho_{\rm ps}$ \citep[though this might be underestimate by a factor of a few, see][]{toropina2}.  
513: Over the whole range of parameters relevant to this work, 
514: the diffusion-induced mass accretion rate
515: is orders of magnitude smaller than that due to the KHI. 
516: Similarly, the contribution to the total luminosity resulting from
517: the shock and anelastic drag at the magnetospheric boundary can be 
518: neglected in this regime. 
519: 
520: 
521: \subsection{Inside the accretion radius: R$_{\rm M}$$<$R$_{\rm a}$}
522: 
523: \subsubsection{The supersonic propeller regime: R$_{\rm co}$$<$R$_{\rm M}$$<$R$_{\rm a}$}
524: \label{sec:supersonic}
525: 
526: Once R$_{\rm M}$ is inside the accretion radius, matter flowing from the 
527: companion star is shocked adiabatically at R$_{\rm a}$ and halted at the NS 
528: magnetosphere. In the region between R$_{\rm a}$ and R$_{\rm M}$  
529: this matter redistributes itself into an approximately spherical
530: configuration  
531: (resembling an ``atmosphere''), whose shape and properties 
532: are determined by the interaction between matter and 
533: NS magnetic field at R$_{\rm M}$. 
534: This scenario was considered previously by \citet{davies} 
535: and \citet{pringle}, and we follow here their treatment. 
536: These authors demonstrated that hydrostatic equilibrium ensues 
537: when radiative losses inside 
538: R$_{\rm a}$ are negligible (we discuss this approximation 
539: in Appendix~\ref{app:supersonic} and \ref{app:subsonic}) and the atmosphere is 
540: stationary on dynamical time-scales.  
541: Assuming a polytropic law of the form p$\propto$$\rho^{1+1/n}$, 
542: the pressure and density of this atmosphere are:
543: \begin{eqnarray}
544: p(R)=\rho_{\rm ps} v_{\rm ps}^2 \left[1+(1/(1+n))8R_{\rm a}/R\right]^{n+1} \\
545: \rho(R)=\rho_{\rm ps}\left[1+(1/(1+n))8R_{\rm a}/R\right]^{n}. ~~~~~~ 
546: \label{eq:hydro} 
547: \end{eqnarray} 
548: The value of the polytropic index $n$ depends on the conditions at 
549: the inner boundary of the atmosphere, and in particular on the rate 
550: at which energy is deposited there.  
551: 
552: When the rotational velocity of the NS magnetosphere at R$_{\rm M}$ 
553: is supersonic (see also \S~\ref{sec:mprop}), 
554: the interaction with matter in the atmosphere leads  
555: to dissipation of some of the star's rotational energy and thus spin-down. 
556: In the supersonic propeller regime, \citet{pringle} showed that 
557: turbulent motions are generated at R$_{\rm M}$ which convect this 
558: energy up through the atmosphere, until it is lost 
559: at its outer boundary. In this case $n$=1/2. 
560: Accordingly, taking into account the structure of the surrounding atmosphere,
561: the magnetospheric radius is given by  
562: \begin{equation}
563: R_{\rm M}^{-6}(1+16 R_{\rm a}/ (3 R_{\rm M}))^{-3/2}=\frac{1}{2}\dot{M}_{\rm w} a^{-2} 
564: \mu^{-2} v_{\rm w}. 
565: \label{eq:rmsuper} 
566: \end{equation} 
567: This can be approximated by
568: \begin{equation}
569: R_{\rm M}\simeq2.3\times10^{10} a_{\rm 10d}^{4/9} \dot{M}_{-6}^{-2/9} 
570: v_8^{4/9} \mu_{33}^{4/9} ~{\rm cm} .
571: \label{eq:rmsuperapprox} 
572: \end{equation}
573: Matter that is shocked at $\sim$R$_{\rm a}$, reaches the magnetospheric boundary at 
574: R$_{\rm M}$ where the interaction with the 
575: NS magnetic field draws energy from NS rotation 
576: (see also \S~\ref{sec:mprop}). According to \citet{pringle}, this 
577: contributes     
578: \begin{equation}
579: L_{\rm sd} = 2\pi R_{\rm M}^2 \rho(R_{\rm M}) c_{\rm s}^3
580: (R_{\rm M})\simeq5.4\times10^{31} \dot{M}_{-6}
581: a_{\rm 10d}^{-2} v_8^{-1} R_{\rm M10}^{1/2} 
582: (1+16 R_{\rm a10}/(3 R_{\rm M10}))^{1/2} ~{\rm erg ~s}^{-1} 
583: \label{eq:lspsuper}
584: \end{equation}
585: to the total luminosity.
586: In the above equation R$_{\rm a10}$=10$^{-10}$R$_{\rm a}$
587: and c$_{\rm s}$(R$_{\rm M}$)=v$_{\rm ff}$ 
588: (R$_{\rm M}$)=(2GM$_{\rm NS}$/R$_{\rm M}$)$^{1/2}$ \citep{pringle}. 
589: In the supersonic propeller regime the energy released through 
590: the shock at R$_{\rm a}$, 
591: \begin{equation}
592: L_{\rm shock} =\frac{9}{32}\pi R_{\rm a}^2 \rho_{\rm w} v_{\rm w}^3  
593: \simeq 2.6\times10^{29} R_{\rm a10}^2 
594: v_8^2 a_{\rm 10d}^{-2} \dot{M}_{-6} ~{\rm erg ~s^{-1}},   
595: \label{eq:lshock2}
596: \end{equation}  
597: is negligible. 
598: 
599: 
600: \subsubsection{The subsonic propeller regime: R$_{\rm M}$$<$R$_{\rm a}$, R$_{\rm co}$, 
601:               $\dot{M}_{\rm w}$$<$$\dot{M}_{\rm lim}$}
602: \label{sec:subsonic}
603: 
604: The break down of the supersonic propeller regime occurs when 
605: R$_{\rm M}$$<$R$_{\rm co}$, i.e., when the magnetosphere rotation 
606: is no longer supersonic with respect to the 
607: surrounding material. The structure of the atmosphere changes and   
608: the transition to the subsonic propeller regime takes place. 
609: Since the rotation of the magnetosphere is subsonic, the atmosphere is roughly 
610: adiabatic ($n$=3/2), and the magnetospheric radius is given by 
611: \citep{pringle}:
612: \begin{equation}
613: R_{\rm M}^{-6}(1+16 R_{\rm a}/(5 R_{\rm M}))^{-5/2}=
614: \frac{1}{2}\dot{M}_{\rm w} a^{-2} \mu^{-2} v_{\rm w}. 
615: \label{eq:rmsub}
616: \end{equation} 
617: This can be approximated by
618: \begin{equation}
619: R_{\rm M}\simeq2\times10^{10} a_{\rm 10d}^{4/7} 
620: \dot{M}_{-6}^{-2/7} v_8^{8/7} \mu_{33}^{4/7} ~{\rm cm}. 
621: \label{eq:rmsubapprox} 
622: \end{equation}
623: In the subsonic propeller regime, the centrifugal barrier 
624: does not operate because R$_{\rm M}$$<$R$_{\rm co}$,  
625: but the energy input at the base of the 
626: atmosphere (due to NS rotational 
627: energy dissipation) is still too high for matter 
628: to penetrate the magnetosphere at a rate $\dot{M}_{\rm capt}$ 
629: \citep{pringle}. 
630: Nevertheless a fraction of the 
631: matter inflow at R$_{\rm a}$  is expected to accrete onto the 
632: NS, due to the KHI and Bohm diffusion\footnote{To our knowledge this is the first 
633: application of the KHI to the subsonic propeller regime.}.  
634: 
635: Based on the discussion 
636: in \S~\ref{sec:intermediate}, we estimate the accretion luminosity 
637: of this matter by using Eqs.~\ref{eq:dotmkh} and 
638: \ref{eq:dotmdiff} (we approximate here the surface of interaction  
639: between matter and magnetic field with 4$\pi$R$_{\rm M}^2$). This gives   
640: \begin{equation}
641: L_{\rm diff} \simeq G M_{\rm NS} \dot{M}_{\rm diff}/R_{\rm NS} 
642: =4.5\times10^{30} \dot{M}_{-6} a_{\rm 10d}^{-2}  
643: R_{\rm M10}^{9/4} \mu_{33}^{-1/2} \zeta^{1/2} v_8^{-1} 
644: (1+16 R_{\rm a10}/(5 R_{\rm M10}))^{3/2} ~{\rm erg ~s}^{-1},  
645: \label{eq:ldiffsub}
646: \end{equation}
647: and   
648: \begin{eqnarray}
649: & & L_{\rm KH} \simeq G M_{\rm NS} \dot{M}_{\rm KH}/R_{\rm NS}=\nonumber \\ 
650: & & 1.8\times10^{35} \eta_{\rm KH} P_{\rm s3}^{-1} R_{\rm M10}^3
651: \dot{M}_{-6} a_{\rm 10d}^{-2} v_8^{-1} 
652: (1+16 R_{\rm a10}/(5 R_{\rm M10}))^{3/2} 
653: (\rho_{\rm i}/\rho_{\rm e})^{1/2}
654: (1+\rho_{\rm i}/\rho_{\rm e})^{-1} ~{\rm erg ~s}^{-1} ,\nonumber \\
655: \label{eq:lkhsub} 
656: \end{eqnarray}
657: for the accretion luminosity arising from matter entering the  
658: magnetosphere through Bohm diffusion and KHI, respectively. 
659: For the range of parameters of interest here, 
660: Eqs.~\ref{eq:lshock}, \ref{eq:ldiffsub}, and \ref{eq:lkhsub} show that 
661: $L_{\rm KH}$ dominates. 
662: The rotational energy dissipation at R$_{\rm M}$ 
663: (see \S~\ref{sec:supersonic}) gives a small contribution with respect 
664: to Eq.~\ref{eq:lkhsub} \citep{pringle}:    
665: \begin{equation}
666: L_{\rm sd} =  2\pi R_{\rm M}^5 \rho(R_{\rm M}) \Omega^3 
667: = 2.2\times10^{30}  P_{\rm s3}^{-3} 
668: R_{\rm M10}^5 \dot{M}_{-6} v_8^{-1} a_{\rm 10d}^{-2} 
669: (1+16 R_{\rm a10}/(5 R_{\rm M10}))^{3/2} ~{\rm erg ~s}^{-1}.  
670: \label{eq:lspsub} 
671: \end{equation}
672: 
673: The subsonic propeller regime applies until the critical accretion rate 
674: \begin{equation}
675: \dot{M}_{\rm lim_{-6}} = 2.8\times10^2 P_{\rm s3}^{-3} 
676: a_{\rm 10d}^{2} v_8 R_{\rm M10}^{5/2} (1+16 R_{\rm a10}/(5 R_{\rm M10}))^{-3/2}
677: \label{eq:dotmlimsub} 
678: \end{equation}
679: is reached, at which the gas radiative cooling (bremsstralhung) completely damps  
680: convective motions inside the atmosphere (see Appendix~\ref{app:subsonic}). 
681: If this cooling takes place, direct accretion at a rate $\dot{M}_{\rm capt}$ 
682: onto the NS surface is permitted.   
683:  
684: 
685: 
686: \subsubsection{The direct accretion regime: R$_{\rm M}$$<$R$_{\rm a}$, R$_{\rm co}$, 
687:                $\dot{M}_{\rm w}$$>$$\dot{M}_{\rm lim}$}
688: \label{sec:accretor}
689: If R$_{\rm M}$$<$R$_{\rm co}$ and matter outside the 
690: magnetosphere cools efficiently, 
691: accretion onto the NS takes place at the same rate  
692: $\dot{M}_{\rm capt}$ (see Eq.~\ref{eq:dotmcapt}) at which it 
693: flows towards the magnetosphere. 
694: The corresponding luminosity is 
695: \begin{equation}
696: L_{\rm acc}= G M_{\rm NS} \dot{M}_{\rm capt} / R_{\rm NS} = 
697: 2\times10^{35} \dot{M}_{-6} a_{\rm 10d}^{-2} 
698: v_8^{-4} ~{\rm erg ~s}^{-1}\simeq2\times10^{35}\dot{M}_{15}~{\rm erg ~s}^{-1}, 
699: \label{eq:lacc} 
700: \end{equation}
701: where $\dot{M}_{15}$=$\dot{M}_{\rm capt}$/10$^{15}$ g s$^{-1}$. 
702: This is the standard accretion regime; the system achieves the highest mass to 
703: luminosity conversion efficiency.  
704: 
705: \begin{figure}
706: \centering
707: \includegraphics[height=22.0cm]{f2.eps}
708: \caption{Transitions between regimes described in \S~\ref{sec:model} 
709: for selected values of the parameters $\mu_{33}$, P$_{\rm s3}$, and v$_{8}$  
710: (in each panel these parameters are indicated in the top right corner). 
711: In all cases we fixed a$_{\rm 10d}$=1 (see \S~\ref{sec:trans}). 
712: In panels (a) and (b) we fixed $\mu_{33}$ and v$_{8}$ and investigated 
713: the different regimes in the  P$_{\rm s3}$-$\dot{M}_{-6}$ plane. 
714: In panels (c), (d), (e), and (f), instead, fixed parameters are P$_{\rm s3}$ 
715: and $\mu_{33}$, and the relevant regimes are shown in the 
716: v$_{8}$-$\dot{M}_{-6}$ plane. 
717: In all cases Eqs.~\ref{eq:dotmrarm}, \ref{eq:pspinlim}, \ref{eq:pspinlimsuper}, 
718: \ref{eq:pspinlimsub}, and \ref{eq:consistencysuper} (see Appendix~\ref{app:supersonic})   
719: are represented with a solid line, a dotted line, a dot-dot-dot dashed line, 
720: a dot-dashed line, and a dashed line, respectively. 
721: Note that the line from Eq.~\ref{eq:consistencysuper} is present only in panel (e), i.e  
722: our treatment of the the supersonic propeller is self-consistent everywhere 
723: except for a small region away from the range of interest for 
724: SFXT sources (see also \S~\ref{sec:results}).}   
725: \label{fig:trans}
726: \end{figure}
727: 
728: \section{Transitions and paths across different regimes}
729: \label{sec:trans}
730: 
731: We explore here the conditions under which 
732: transitions across different regimes take place. 
733: As emphasised in \S~\ref{sec:model}, these transitions occur 
734: when the relative positions of R$_{\rm M}$, R$_{\rm a}$, and R$_{\rm co}$
735: change; we concentrate here on transitions that occur in response to 
736: variations in the stellar wind parameters. In the following, 
737: since R$_{\rm M}$ depends only weakly on the orbital period and the 
738: total mass of the system, we fix a$_{\rm 10d}$=1 
739: (we explain this choice in \S~\ref{sec:results}), and 
740: investigate variations in the other four parameters: 
741: $\mu_{33}$, P$_{\rm s3}$, v$_{8}$, and $\dot{M}_{-6}$. 
742: 
743: The equations that define the conditions for transitions between different 
744: regimes are\footnote{Here we used 
745: Eqs.~\ref{eq:rmsuperapprox} and \ref{eq:rmsubapprox} for the magnetospheric 
746: radius in the supersonic and subsonic propeller regime, respectively.} 
747: \begin{equation}
748: R_{\rm M}>R_{\rm a} \Rightarrow \dot{M}_{-6}\lesssim0.45 \mu_{33}^2 v_{8}^{11} a_{\rm 10d}^2  
749: \label{eq:dotmrarm}
750: \end{equation} 
751: (or equivalently $\dot{M}_{15}$$\lesssim$0.6$\mu_{33}^2$v$_{8}^{7}$, see Eqs.~\ref{eq:ra}, 
752: \ref{eq:rm} and \ref{eq:dotmcapt}), for the magnetic barrier; 
753: \begin{equation} 
754: R_{\rm M}>R_{\rm co} \Rightarrow P_{\rm s3}\lesssim2.6\dot{M}_{-6}^{-1/4} 
755: v_{8}^{-1/4} a_{\rm 10d}^{1/2} \mu_{33}^{1/2}\simeq2.8\dot{M}_{15}^{-1/4}v_{8}^{-5/4}\mu_{33}^{1/2}  
756: \label{eq:pspinlim}
757: \end{equation}
758: (see Eq.~\ref{eq:rm}) if $R_{\rm M}>R_{\rm a}$, or 
759: \begin{equation}
760: R_{\rm M}>R_{\rm co} \Rightarrow P_{\rm s3}\lesssim1.8 a_{\rm 10d}^{2/3} 
761: \dot{M}_{-6}^{-1/3} v_{8}^{2/3} \mu_{33}^{2/3}\simeq2\dot{M}_{15}^{-1/3}v_{8}^{-2/3}\mu_{33}^{2/3} 
762: \label{eq:pspinlimsuper}
763: \end{equation}
764: (see Eq.~\ref{eq:rmsuperapprox}) if R$_{\rm M}$$<$R$_{\rm a}$, for the centrifugal barrier. 
765:  
766: The equation that defines the transition from the subsonic propeller to the direct accretion regime is  
767: \begin{equation}
768: P_{\rm s3}\gtrsim4.5 \dot{M}_{-6}^{-15/21} a_{\rm 10d}^{30/21} v_{8}^{60/21} \mu_{33}^{16/21}
769: \simeq5.5\dot{M}_{15}^{-15/21}\mu_{33}^{16/21}. 
770: \label{eq:pspinlimsub}
771: \end{equation} 
772: 
773: In Fig.~\ref{fig:trans} the above equations are represented as lines separating different
774: regimes. In panels (a) and (b) we fixed $\mu_{33}$ and v$_{8}$ and investigated 
775: the different regimes in the  P$_{\rm s3}$-$\dot{M}_{-6}$ plane. In panels (c), (d), (e) and (f), instead, 
776: P$_{\rm s3}$ and $\mu_{33}$ were fixed and the relevant regimes shown in the 
777: v$_{8}$-$\dot{M}_{-6}$ plane. 
778: Below we summarise the different regimes that a system attains in response to
779: variations of $\dot{M}_{\rm w}$, in the different panels.
780: 
781: Panel (a) shows that, for a wind velocity of v$_{8}$=1.2, 
782: a strongly magnetized NS ($\mu_{33}$=0.8)
783: can undergo a transition between the superKeplerian and subKeplerian 
784: magnetic inhibition regimes in response to changes 
785: in the mass loss rate only for typical spin periods $\gtrsim$2000~s. 
786: When the mass loss rate reaches $\dot{M}_{-6}$$\sim$2.4, 
787: the direct accretion or subsonic propeller regime sets in, depending on whether 
788: the spin period is longer or shorter than $\sim$3700~s. 
789: For spin periods $\lesssim$420~s the direct accretion
790: regime is not attained for the interval of mass loss rates considered in 
791: Fig.~\ref{fig:trans} and only transitions between the superKeplerian magnetic 
792: inhibition and supersonic propeller regime are expected.  
793: 
794: For lower magnetic fields ($\mu_{33}$=0.01), panel (b) shows that only the supersonic 
795: propeller, subsonic propeller, and direct accretion regime can be attained. 
796: For spin periods in the range $\sim$60-230~s transitions can occur between all these three regimes, 
797: whereas systems with spin periods longer than $\sim$1300~s and shorter than $\sim$20~s are expected 
798: to be in the direct accretion and the supersonic propeller regime, respectively. 
799: For $\mu_{33}$=0.01 transitions to the superKeplerian and subKeplerian magnetic 
800: inhibition regimes cannot take place because the magnetospheric radius is too small to  
801: exceed the accretion radius. 
802: 
803: In panel (c) ($\mu_{33}$=0.8 and P$_{\rm s3}$=1), transitions can occur virtually 
804: between all the regimes described in \S~\ref{sec:model}. 
805: In particular, for v$_{8}$ in the range 0.9-1.5 transitions are expected to take place between 
806: the superKeplerian magnetic inhibition, the supersonic and subsonic 
807: propeller, and the direct accretion regime, as the mass loss rate increases from 
808: $\dot{M}_{-6}$$\sim$0.1 to $\dot{M}_{-6}$$\sim$100. 
809: For velocities v$_{8}$$<$0.9 transitions can occur only between the supersonic propeller, 
810: the subsonic propeller and the direct accretion regime, whereas transitions 
811: to the superKeplerian and subKeplerian magnetic inhibition regimes  
812: are impeded by the fact that the accretion radius cannot be overtaken by $R_{\rm M}$.  
813: On the contrary, for v$_{8}$$>$1.5, the magnetospheric radius is located beyond  
814: the accretion radius for any considered value of the mass loss rate, and thus transitions can 
815: take place only between the superKeplerian and subKeplerian magnetic inhibition regimes. 
816: Similar considerations apply to panels (d), (e), and (f). 
817:  
818: For a system with $\mu_{33}$=0.01 and P$_{\rm s3}$=1 (panel (d)), the superKeplerian magnetic 
819: inhibition regime never occurs, because the magnetic field is too low and 
820: R$_{\rm M}$$<$R$_{\rm co}$ for any 0.1$<$$\dot{M}_{-6}$$<$100. 
821: Instead, the subKeplerian magnetic 
822: inhibition regime can be attained for high wind velocities (v$_{8}$=2), 
823: because R$_{\rm a}$$\propto$v$_{\rm w}^{-2}$. 
824: 
825: In panel (e), for $\mu_{33}$=0.8 and P$_{\rm s3}$=0.1, the magnetospheric radius is 
826: larger than the corotation radius for the entire range spanned by $\dot{M}_{-6}$.  
827: Thus accretion onto the NS does not take place \citep[note that in the region 
828: below the dashed line accretion can occur even if 
829: R$_{\rm M}$ $\gtrsim$R$_{\rm co}$,][]{pringle}\footnote{This is because, 
830: in this region, the mass flow rate is so high that Eq.~\ref{eq:consistencysuper} 
831: is violated (the supersonic propeller is no longer self-consistent), 
832: and convective motions are damped by radiative cooling.}.  
833: 
834: Finally, in panel (f) we show the transitions for a system with 
835: $\mu_{33}$=0.01 and P$_{\rm s3}$=0.1~s. 
836: In this case all the regimes described in \S~\ref{sec:model} are 
837: present in the figure, similar to the case of panel (c). 
838: However, the region corresponding to the subsonic propeller regime is larger,
839: such that there is only a modest range of (fixed) velocities for which mass loss rate 
840: variations in our chosen range (0.1$<$$\dot{M}_{-6}$$<$100) 
841: can cause transitions through all regimes, from the 
842: superKeplerian magnetic inhibition to the direct accretion regime. 
843: As we discuss below, this has important consequences for the 
844: expected luminosity variations. \\ 
845: 
846: \begin{figure}[t!]
847: \centering
848: \includegraphics[height=6.5 cm]{f3.eps}
849: \smallskip
850: \caption{(A) \textit{Upper panel}: Variation of the luminosity 
851: through different regimes, as a function of the mass loss rate from 
852: the companion star. 
853: In this case the parameters of the model are fixed at $\mu_{33}$=0.8, 
854: P$_{\rm s3}$=1, and v$_{8}$=1.2. 
855: \textit{Lower panel}: Relative position of the magnetospheric radius, 
856: R$_{\rm M}$ (solid line), with respect to the accretion radius R$_{\rm a}$ 
857: (dotted line), and the corotation radius R$_{\rm co}$ (dashed line), 
858: as a function of the mass loss rate from the companion star. \newline
859: (B) \textit{Upper panel}: Variation of the luminosity 
860: through different regimes, as a function of the mass loss rate 
861: from the companion star. 
862: In this case the parameters of the model are fixed at 
863: $\mu_{33}$=0.001, P$_{\rm s3}$=1, and v$_{8}$=1.2.  
864: \textit{Lower panel}: Relative position 
865: of the magnetospheric radius, R$_{\rm M}$ (solid line), 
866: with respect to the accretion radius R$_{\rm a}$ (dotted line), 
867: and corotation radius R$_{\rm co}$ (dashed line), 
868: as a function of the mass loss rate from the companion star.}  
869: \label{fig:magn_e_non} 
870: \end{figure}
871: 
872: We now compute the luminosity swings for some of the 
873: examples discussed above in a fashion similar 
874: to what was done in the context of 
875: centrifugally inhibited accretion in NS  
876: X-ray transients \citep{corbet96,campana98}.   
877: Fig.~\ref{fig:magn_e_non}A applies to a system with  
878: $\mu_{33}$=0.8, P$_{\rm s3}$=1, and v$_{8}$=1.2  
879: (this corresponds to the case P$_{\rm s3}$=1 of Fig.~\ref{fig:trans} panel (a)). 
880: The lower panel of this figure shows that, for 0.1$<$$\dot{M}_{-6}$$<$100, 
881: the magnetospheric radius crosses both the centrifugal (R$_{\rm co}$) and magnetic 
882: (R$_{\rm a}$) barriers. Correspondingly, the system moves from  
883: the superKeplerian magnetic inhibition regime, to the supersonic and 
884: subsonic propeller regime, and, finally, to the direct accretion regime, 
885: giving rise to a six-decade luminosity swing from $\sim$10$^{31}$ to 
886: $\sim$10$^{37}$~erg~s$^{-1}$. 
887: We note that a large part of this swing (about five decades) is attained  
888: across the transitions from the superKeplerian magnetic inhibition to
889: the direct accretion regimes, 
890: which take a mere factor of $\sim$5 variation of $\dot{M}_{\rm w}$. 
891: 
892: In the presence of a standard NS magnetic field ($10^{12}$~G), 
893: Fig.~\ref{fig:magn_e_non}B  shows that such abrupt luminosity jumps 
894: are not expected for a very slowly rotating (1000~s) NS 
895: (the other system parameters are the same 
896: as those of Fig.~\ref{fig:magn_e_non}A), since the magnetospheric 
897: radius is smaller than both R$_{\rm a}$ and R$_{\rm co}$, for any 
898: reasonable value of $\dot{M}_{\rm w}$. Therefore, 
899: the direct accretion regime applies, with the  
900: the luminosity proportional to $\dot{M}_{\rm w}$. 
901: 
902: In Fig.~\ref{fig:lshortspin} we show the 
903: transitions for a system with $\mu_{33}$=0.01 and P$_{\rm s3}$=0.1.  
904: The wind velocity is v$_{8}$=1.2 in Fig.~\ref{fig:lshortspin}A, and v$_{8}$=2.2 
905: in Fig.~\ref{fig:lshortspin}B (see also panel (f) of Fig.~\ref{fig:trans}).  
906: These two figures show that, for sub-magnetar fields, a 100~s
907: spinning NS can undergo a transition across the magnetic barrier 
908: (besides the centrifugal barrier), for suitable parameters 
909: (a high wind velocity in the case at hand). Such transitions
910: take place over a more extended interval of mass loss rates. 
911: For instance Fig.~\ref{fig:lshortspin}B shows that an increase by  
912: a factor $\sim$100 in the mass loss rate is required, in this case,  
913: to achieve a factor $\sim$10$^5$ luminosity swing 
914: comparable with the magnetar case of Fig.~\ref{fig:magn_e_non}A. \\ 
915: 
916: In order to illustrate further the role of the magnetic field, spin
917: period and wind velocity, 
918: we show in Fig.~\ref{fig:vwind} the way in which the transitions across 
919: regimes take place, by holding two of the above variables
920: fixed and stepping the third variable. 
921: Figure~\ref{fig:vwind}A shows the effect of   
922: increasing the value of v$_{8}$ from 1 to 1.8
923: (in turn resulting in a decrease of the accretion radius), in a system with 
924: $\mu_{33}$=0.8 and P$_{\rm s3}$=1.  
925:  
926: The behaviour of the luminosity changes mainly because a different set 
927: of regimes is involved in each case. 
928: For v$_{8}$=1 (solid line) the system passes through the superKeplerian 
929: magnetic inhibition regime, the supersonic, and subsonic propeller regime,  
930: finally reaching the direct accretion regime at $\dot{M}_{-6}$$\simeq$6. 
931: In the case v$_{8}$=1.2 (dotted line), 
932: the transition to the supersonic propeller shifts towards higher mass 
933: loss rates, 
934: such that superKeplerian magnetic inhibition applies up to $\dot{M}_{-6}$$<$2. 
935: Further increasing the wind velocity to v$_{8}$=1.4 (dashed line), 
936: the system first undergoes a transition to the subsonic propeller 
937: at $\dot{M}_{-6}$$\simeq$10, bypassing the supersonic propeller    
938: (this is because for $\dot{M}_{-6}$$\simeq$10 the magnetospheric radius 
939: defined by Eq.~\ref{eq:rmsuper} is smaller than the corotation radius).  
940: As the mass loss rate increases further, the direct accretion regime sets in  
941: for $\dot{M}_{-6}$$\simeq$20. 
942: In the case v$_{8}$=1.8 (dot-dashed line), the corotation radius exceeds the 
943: accretion radius. The system is thus in the superKeplerian magnetic inhibition 
944: regime for $\dot{M}_{-6}$$<$20, while for higher mass 
945: loss rates the subKeplerian magnetic inhibition regime applies and 
946: the luminosity is dominated by accretion through the KHI. 
947: A transition to the direct accretion regime is expected 
948: for mass loss rates $\dot{M}_{-6}$$>$100. 
949: 
950: Figure~\ref{fig:vwind}B shows the transitions across different regimes 
951: for selected values of the magnetic field, in a system with 
952: P$_{\rm s3}$=1 and v$_{8}$=1.2. For the lowest magnetic field in this 
953: figure ($\mu_{33}$=0.01, solid line), the magnetospheric radius is smaller than 
954: both the accretion and corotation radius for the whole range of mass loss rates 
955: spanned in the figure, and the system is always in the direct accretion regime. 
956: In the case $\mu_{33}$=0.1 (dotted line), the system is in the subsonic propeller for 
957: $\dot{M}_{-6}$$<$1, while for higher mass inflow rates the direct accretion regime applies. 
958: For $\mu_{33}$=0.5 (dashed line) the luminosity behaviour becomes more complex, as 
959: the system goes through the superKeplerian 
960: magnetic inhibition regime, the supersonic and subsonic propeller regime, and 
961: eventually reaches the direct accretion regime at $\dot{M}_{-6}$$\simeq$8. 
962: For $\mu_{33}$=0.8 the same sequence of transitions applies, with 
963: the entire luminosity swing taking place over a smaller interval 
964: of mass loss rates.
965: 
966: \begin{figure}[t!]
967: \centering
968: \includegraphics[height=6.3 cm]{f4.eps}
969: \smallskip
970: \caption{(A) \textit{Upper panel}: Variation of the luminosity through 
971: different regimes, as a function of the mass loss rate. 
972: In this case the parameters of the model are fixed at 
973: $\mu_{33}$=0.01, P$_{\rm s3}$=0.1, and v$_{8}$=1.2.  
974: \textit{Lower panel}: Relative position of the 
975: magnetospheric radius, R$_{\rm M}$ (solid line), with respect to the 
976: accretion radius R$_{\rm a}$ (dotted line), and corotation radius 
977: R$_{\rm co}$ (dashed line), as a function of the mass loss rate from 
978: the companion star. \newline
979: (B) Same as (A) but for $\mu_{33}$=0.01, P$_{\rm s3}$=0.1, and 
980: v$_{8}$=2.2.} 
981: \label{fig:lshortspin} 
982: \end{figure}
983: 
984: Finally, in Fig.~\ref{fig:vwind}C we show the effects of increasing 
985: the spin period in a system 
986: with $\mu_{33}$=0.1 and v$_{8}$=1.2. For the lowest spin period  
987: considered here (P$_{\rm s3}$=0.01, solid line), the system 
988: remains in the supersonic propeller regime 
989: for the whole range spanned by $\dot{M}_{-6}$. In the case  
990: $P_{\rm s3}$=0.5 (dotted line) transitions occur from the 
991: supersonic propeller regime ($\dot{M}_{-6}$$<$0.7), to the subsonic 
992: propeller (0.7$<$$\dot{M}_{-6}$$<$4) and then to the direct accretion regime 
993: ($\dot{M}_{-6}$$>$4). By further increasing $P_{\rm s3}$ 
994: to a value of 1 (dashed line), the system goes through the subsonic 
995: propeller and the 
996: direct accretion regime, while the supersonic propeller regime does not 
997: occur due to the longer spin period as compared to the previous case. 
998: For $P_{\rm s3}$=6 (dot-dashed line) the spin period is so 
999: high that the direct accretion regime applies for any reasonable value of $\dot{M}_{-6}$. 
1000: 
1001: The above results show that over a range of values of the 
1002: key parameters $\mu_{33}$, P$_{\rm s3}$, and v$_{8}$,  
1003: large luminosity swings can be achieved with comparatively modest 
1004: changes in the mass loss rate, as the neutron star undergoes 
1005: transitions from one regime to another. 
1006: More generally, these transitions result  
1007: from changes in the relative position
1008: of the accretion, magnetospheric, and corotation radii,
1009: reflecting short term variations of the wind 
1010: velocity and mass loss rate, the only parameters 
1011: that can vary on shorter timescales than secular. 
1012: Therefore, transitions between different regimes take place once the source 
1013: parameters are such that the NS straddles the centrifugal barrier (i.e.  
1014: R$_{\rm M}$$\simeq$R$_{\rm co}$, when R$_{\rm a}$$>$R$_{\rm co}$) 
1015: or the magnetic barrier (i.e. R$_{\rm M}$$\simeq$R$_{\rm a}$, 
1016: when R$_{\rm a}$$<$R$_{\rm co}$) or both 
1017: (i.e. R$_{\rm M}$$\simeq$R$_{\rm co}$$\simeq$R$_{\rm a}$). 
1018: The centrifugal barrier applies to relatively short spin periods.  
1019: It is well known that the longer the spin period of 
1020: a transient neutron star, the higher its magnetic field must be
1021: for the centrifugal barrier to operate 
1022: (see Eqs.~\ref{eq:pspinlim} and~\ref{eq:pspinlimsuper}).
1023: In particular, for periods of hundreds seconds, or longer, magnetic field 
1024: strengths of $\gtrsim$10$^{13}$-10$^{14}$~G are required\footnote{This point 
1025: was already noted in \citet{stella86}.}. 
1026: On the other hand, we have shown that 
1027: neutron stars with even longer spin periods 
1028: and magnetar-like fields are expected to undergo transitions across 
1029: the magnetic barrier and thus are expected to have an 
1030: inherently different ``switch off'' mechanism than short spin period 
1031: systems. A necessary condition for this is that R$_{\rm co}$$>$R$_{\rm a}$, 
1032: which translates into P$_{\rm s3}$$\gtrsim$3v$_{8}^{-3}$. 
1033: The magnetic and centrifugal barriers set in (nearly) simultaneously 
1034: (i.e. R$_{\rm a}$$\simeq$R$_{\rm co}$$\simeq$R$_{\rm M}$) 
1035: for $\mu_{33}$$\simeq$0.3P$_{\rm s3}^{7/6}$$\dot{M}_{15}^{1/2}$.  
1036: 
1037: Taking into account of all the 
1038: examples discussed in this section, we conclude that: 
1039: \begin{enumerate}
1040: \item Long spin period systems (P$_{\rm s3}$$\gtrsim$1) require magnetar-like 
1041: B-fields ($\mu_{33}$$\gtrsim$0.1) in order for a large luminosity swing   
1042: ($\sim$10$^5$) to arise from modest variations in the wind parameters 
1043: (e.g. a factor $\sim$5 in $\dot{M}_{-6}$). These luminosity swings might  
1044: result from transition across different regimes through both the centrifugal 
1045: and magnetic barriers. 
1046: \item Shorter spin period systems (P$_{\rm s3}$$\ll$1) must posses lower magnetic 
1047: fields ($\mu_{33}$$\ll$0.1) for similar transitions to take place.  
1048: Somewhat larger variations in the wind parameters are required 
1049: in order to achieve similar luminosity swings to those of the long period case, and 
1050: transitions between different regimes occur in most cases through the centrifugal barrier. 
1051: \item Few or no transitions are expected for systems with either high magnetic 
1052: fields and short spin periods, or systems with lower magnetic fields and long spin periods. 
1053: In the first case the centrifugal barrier halts the 
1054: inflowing matter at R$_{\rm M}$ and accretion does not take place; such systems might 
1055: thus be observable only at very low (X-ray) luminosity levels 
1056: ($\simeq$10$^{32}$-10$^{33}$~erg~s$^{-1}$). 
1057: In the second case R$_{\rm M}$$<$R$_{\rm co}$ for a wide range of wind parameters,  
1058: accretion can take place, and a high persistent luminosity is released  
1059: ($\simeq$10$^{35}$-10$^{37}$~erg~s$^{-1}$). 
1060: \end{enumerate}
1061: 
1062: \begin{figure}[t!]
1063: \centering
1064: \includegraphics[height=6.0 cm, width=18.0 cm]{f5.eps}
1065: \smallskip
1066: \caption{Variations of the luminosity through different regimes, 
1067: as a function of the mass loss rate, for different sets of 
1068: parameters. \newline
1069: (A): In this case we fixed $\mu_{33}$=0.8 and P$_{\rm s3}$=1.  
1070: The solid line shows luminosity variations for v$_{8}$=1, the dotted line is for 
1071: v$_{8}$=1.2, and the cases v$_{8}$=1.4 and v$_{8}$=1.8 are represented with 
1072: a dashed and dot-dashed line, respectively. \newline
1073: (B): Here the fixed parameters are P$_{\rm s3}$=1 and v$_{8}$=1.2. 
1074: The different curves correspond to $\mu_{33}$=0.01 (solid line), 
1075: 0.1 (dotted line), 0.5 (dashed line), and 0.8 (dot-dashed line). \newline
1076: (C): In this case we fixed $\mu_{33}$=0.1 and v$_{8}$=1.2. Luminosity variations across 
1077: different regimes are shown for P$_{\rm s3}$=0.01 (solid line), 0.5 (dotted line), 
1078: 1 (dashed line), and 6 (dot-dashed line).} 
1079: \label{fig:vwind} 
1080: \end{figure}
1081: 
1082: \section{Application to SFXT sources}
1083: \label{sec:results}
1084:  
1085: In this section we propose that transitions across different regimes 
1086: caused by relatively mild variations of the wind parameters are responsible
1087: for the outbursts of SFXTs. In consideration of the wide range 
1088: of spin periods inferred for SFTXs, the outbursts of these sources are expected
1089: to result from the opening of the magnetic barrier in very long spin period systems and 
1090: the centrifugal barrier in all other systems \citep[see also][]{grebenev07,sidoli07}. 
1091: More crucially, we conclude that slowly spinning SFXTs should host magnetars,
1092: independent of which of the two mechanisms is responsible for their outbursts.
1093: 
1094: As a case study we consider \IGR\ \citep{sunyaev03}, a SFXT observed by \chan\ 
1095: during a complex transition to and from a $\sim$1~hour-long outburst,
1096: yielding the first detailed characterization of a SFXT light curve. 
1097: The spin period of \IGR\ is presently unknown. 
1098: \citet{zand05} showed that four different 
1099: stages, with very different luminosity levels, 
1100: could be singled out during the \chan\
1101: observation: (a) a quiescent state with 
1102: L$_{\rm X}$$\simeq$2$\times$10$^{32}$~erg~s$^{-1}$, (b) a rise stage with 
1103: L$_{\rm X}$$\simeq$1.5$\times$10$^{34}$~erg~s$^{-1}$, (c) the outburst peak 
1104: with L$_{\rm X}$$\simeq$4$\times$10$^{37}$~erg~s$^{-1}$, and (d) a post-outburst 
1105: stage (or``tail'') with L$_{\rm X}$$\simeq$2$\times$10$^{36}$~erg~s$^{-1}$ 
1106: \citep[see panel (a) of Fig.~\ref{fig:IGRJ17544}; these luminosities are for a source 
1107: distance of $\sim$3.6 kpc,][]{rahoui08}.   
1108: The maximum luminosity swing observed across these stages was 
1109: a factor of $\gtrsim$6.5$\times$10$^4$. 
1110: 
1111: Motivated by the evidence for  $>$1000~s periodicities  
1112: in \xte\ and \igg\ (see \S~\ref{sec:properties}), we discuss first 
1113: the possibility that \IGR\ contains a very slowly spinning neutron
1114: star. We use $\mu_{33}$=1, P$_{\rm s3}$=1.3, v$_{8}$=1.4, 
1115: and show in Fig.~\ref{fig:IGRJ17544}(b) the different regimes 
1116: experienced by such a neutron star as a function of the mass loss rate. 
1117: For $\dot{M}_{-6}$$<$20 the above values values give R$_{\rm M}$$>$R$_{\rm a}$ 
1118: and R$_{\rm M}$$>$R$_{\rm co}$, such that superKeplerian magnetic inhibition of accretion applies.  
1119: The expected luminosity 
1120: in this regime, $\sim$10$^{31}$~erg~s$^{-1}$, is likely outshined by the X-ray luminosity 
1121: of the supergiant star \citep[the companion star's luminosity is not shown in 
1122: Fig.~\ref{fig:IGRJ17544}, but it is typically of order $\sim$10$^{32}$~erg~s$^{-1}$,][]
1123: {cassinelli81,berghofer97}. We conclude that the lowest emission state 
1124: (quiescence) of \IGR\ can be explained in this way, with the companion 
1125: star dominating the high energy luminosity \citep{zand05}. 
1126: The rise stage is in good agreement with the subKeplerian magnetic inhibition regime, 
1127: where the luminosity ($\sim$10$^{34}$~erg~s$^{-1}$) is dominated by accretion of matter 
1128: onto the NS due to the KHI. The uncertainty in the value of $h$ translates into 
1129: an upper limit on the luminosity in this regime which is
1130: a factor of $\sim$10 higher than that given above (see \S~3 and Appendix~\ref{app:ht}). 
1131: 
1132: During the outburst peak the direct accretion regime must apply at a  
1133: mass loss rate of $\dot{M}_{-6}$$=$500. In this interpretation 
1134: direct accretion must also be at work in the outburst tail at 
1135: $\dot{M}_{-6}$$\sim$3, where  
1136: a slight decrease in $\dot{M}_{\rm w}$ would cause the magnetic barrier to close and the 
1137: source to return to quiescence. According to this interpretation, 
1138: if \IGR\ has a spin period of $>$1000~s, then it must host a magnetar. 
1139: 
1140: Panel (c) of Fig.~\ref{fig:IGRJ17544} shows an alternative interpretation of 
1141: the \IGR\ light curve, where we fixed $\mu_{33}$=0.08, P$_{\rm s3}$=0.4, and 
1142: v$_{8}$=1. 
1143: For this somewhat faster spin (and lower magnetic field), the luminosity variation is mainly 
1144: driven by a transition across the centrifugal barrier (as opposed to the magnetic barrier). 
1145: In this case, the quiescent state corresponds to the supersonic propeller regime 
1146: ($\dot{M}_{-6}$ $<$0.6), the rise stage to the subsonic propeller (0.6$<$$\dot{M}_{-6}$$<$2),  
1147: while both the peak of the outburst and the tail take place in the direct accretion regime 
1148: at $\dot{M}_{-6}$=200 and $\dot{M}_{-6}$=10, respectively. 
1149:  
1150: Assuming an even faster NS spin period for \IGR,\ a weaker magnetic field would be
1151: required.   
1152: In panel (d) of Fig.~\ref{fig:IGRJ17544}, we show the results obtained by adopting 
1153: $\mu_{33}$=0.001, P$_{\rm s3}$=0.01, and v$_{8}$=2. The $\sim$10$^{34}$~erg~s$^{-1}$ 
1154: luminosity in the subsonic propeller regime compares well with the luminosity in the 
1155: rise stage. However, the luminosity of the supersonic propeller regime 
1156: is now significantly higher than the quiescence luminosity 
1157: of $\sim$10$^{32}$~erg~s$^{-1}$ (this is consequence of the higher value of 
1158: $\dot{M}_{\rm w}$ for which the supersonic propeller regime is attained in this interpretation). 
1159: We note that, the whole luminosity swing takes place 
1160: for a wider range of mass loss rates, and the outburst peak luminosity requires 
1161: $\dot{M}_{-6}$$\simeq$3000, an extremely high values even for an OB supergiant. 
1162: 
1163: Interpreting the properties of \IGR\ in terms of a NS 
1164: with a spin periods $\ll$100~s is more difficult.   
1165: For instance, for the subsonic propeller regime to set in, 
1166: the mass loss rate corresponding to the transition across
1167: R$_{\rm M}$=R$_{\rm co}$ must be lower than the limit 
1168: fixed by Eq.~\ref{eq:dotmlimsub}. If instead the transition  
1169: takes place at higher mass loss rate, the system goes directly from the supersonic 
1170: propeller to the direct accretion regime (or vice versa), bypassing the subsonic propeller: 
1171: therefore, the rise stage would remain unexplained. 
1172: Since fast rotating NSs require lower magnetic fields for direct accretion to take place
1173: while in outburst, Eq.~\ref{eq:dotmlimsub} is satisfied 
1174: only for very high wind velocities (v$_{8}$$>$2-3).  
1175: On the other hand, an increase by a factor of $\sim$2 in the wind velocity 
1176: (with respect to the longer spin period solutions) would give a substantially lower 
1177: $\dot{M}_{\rm capt}$, such that the subsonic and the direct accretion regime 
1178: luminosities fall shortwards of the observed values (unless unrealistically high 
1179: mass loss rate are considered).  
1180: 
1181: Based on the above discussion, we conclude that \IGR\ 
1182: likely hosts a slowly rotating NS, with spin period $>$100~s. 
1183: Whether the magnetic barrier or the centrifugal barrier sets in, causing 
1184: inhibition of accretion away from the outbursts, will depend on whether 
1185: the spin period is longer or shorter then $\sim$1000~s. 
1186: We note that \objectname{IGR J16418-4532}, a $\sim$1240~s pulsating source with a 3.7d orbital 
1187: period, displayed short duration flares similar to those of SFXTs and thus 
1188: might be considered a candidate for hosting an accreting neutron star 
1189: with a magnetar-like field.   
1190: However there is no clear evidence yet that \iigr\ is a transient source, 
1191: since the very low state revealed with \swift\ might well be due to an eclipse
1192: \citep{tomsick04,walter06,corbet06}. 
1193: 
1194: As another example we discuss the case of \igrjj,\ a SFXTs with  
1195: a spin period of 228~s. The luminosity behaviour  
1196: of this source is still poorly known. An outburst
1197: at 5$\times$10$^{36}$~erg~s$^{-1}$ was observed with \int\  
1198: \citep[assuming a distance of 12.5 kpc,][]{Lutovinov05,smith04}, which did not 
1199: detect the source before the outburst down to a level 
1200: of 5$\times$10$^{35}$~erg~s$^{-1}$. About a week later,  
1201: \xmm\ revealed the source at 5$\times$10$^{34}$~erg~s$^{-1}$  
1202: and discovered the 228~s pulsations \citep{Lutovinov05, zurita04}. 
1203: If the direct accretion regime applied all the way to the 
1204: lowest luminosity level observed so far, then an upper limit 
1205: of $\mu_{33}$=0.004 would be obtained by imposing that 
1206: the neutron star did not enter  
1207: the subsonic propeller regime (see Eq.~\ref{eq:pspinlimsub}). 
1208: On the other hand, if the luminosity measured by 
1209: \xmm\ signalled that the source entered the subsonic propeller
1210: regime, while direct accretion occurred only during the outburst 
1211: detected by \int,\ then a considerably higher 
1212: magnetic field of $\mu_{33}$=0.07 would be required. 
1213: 
1214: The above discussion emphasizes the importance of determining, 
1215: through extended high sensitivity observations,  
1216: the luminosity at which transitions between different 
1217: source states occur, in particular the lowest luminosity 
1218: level for which direct accretion is still at work. 
1219: In combination with the neutron star spin, this can
1220: be used to infer the neutron star magnetic field.  
1221: Alternatively accretion might take place unimpeded 
1222: at all luminosity levels of SFXTs, a possibility
1223: which requires a very clumpy wind as envisaged in 
1224: other scenarios \citep{negueruela08,zurita07}. In this case 
1225: the neutron star magnetic field can be considerably lower 
1226: than discussed here. 
1227: More extensive studies of these sources (and, by analogy, 
1228: other SFXTs) are clearly required.
1229: 
1230: \begin{figure*}[t!]
1231: \centering 
1232: \includegraphics[width=18.0 cm]{f6.eps}
1233: \smallskip
1234: \caption{Application of the model described in \S~\ref{sec:model} to 
1235: the \IGR\ transition from quiescence to outburst. \newline
1236: {\it Panel (a)}: A schematic representation of the \chan\ light curve 
1237: of \IGR\ obtained by using the segments that were
1238: not affected by pile-up \citep[see both panels of figure 2 in][]{zand05}. 
1239: The different luminosity stages are clearly visible. 
1240: According to \citet{zand05}, the count rates on the y-axis correspond 
1241: to $2\times10^{32}$~erg~s$^{-1}$, 
1242: $1.5\times10^{34}$~erg~s$^{-1}$, $2\times10^{36}$~erg~s$^{-1}$, and 
1243: $4\times10^{37}$~erg~s$^{-1}$, in the quiescence state, the rise state, 
1244: the tail, and the peak of the outburst, respectively. \newline
1245: {\it Panel (b)}: Interpretation of the quiescence to outburst transition of 
1246: \IGR\ in terms of the magnetic barrier model.
1247: The dotted horizontal lines mark the luminosity that 
1248: divide the different regimes. The parameters of the model are fixed at 
1249: $P_{\rm s3}$=1.3, $v_{8}$=1.4, and $\mu_{33}$=1. \newline
1250: {\it Panel (c)}: Interpretation of the quiescence to outburst transition of 
1251: \IGR\ based on the centrifugal barrier model. The parameters of the model are fixed at 
1252: $\mu_{33}$=0.08, $P_{\rm s3}$=0.4, and $v_{8}$=1. \newline
1253: {\it Panel (d)}: Same as panel (c) but here the parameters of the model are fixed at 
1254: $\mu_{33}$=0.001, $P_{\rm s3}$=0.01, and $v_{8}$=2.} 
1255: \label{fig:IGRJ17544} 
1256: \end{figure*} 
1257: 
1258: 
1259: \section{Discussion}
1260: \label{sec:discussion}
1261: 
1262: If the centrifugal barrier operates in \IGR,\ then the activity of this source 
1263: (and by extension that of SFXTs with similar properties) should parallel
1264: that of long spin period X-ray pulsar transients with Be star companions
1265: \citep{stella86}. 
1266: One crucial difference between the two classes,
1267: namely the duration of the outbursts, might well result from the presence in Be systems 
1268: of an accretion disk mediating the flow of matter outside the NS magnetosphere. 
1269: In fact, while in Be systems the star's slow equatorial wind has enough angular
1270: momentum to form such a disk, the supergiant's wind in SFXTs is fast and possesses
1271: only little angular momentum relative to the NS. In the absence of a disk,  
1272: variations in the wind parameters take effect on dynamical timescales, 
1273: whereas in the presence of a disk they are smoothed out  
1274: over viscous timescales. 
1275: 
1276: If the spin period is sufficiently long in \IGR\ and other SFXTs, the onset of 
1277: the magnetic barrier will inhibit accretion. While steady 
1278: magnetic inhibition of accretion is familiar, e.g. from the earth magnetosphere - 
1279: solar wind interaction, transitions in and out of this regime have not yet been observed, 
1280: to the best of our knowledge. Therefore, very long spin period SFXTs might 
1281: provide the first opportunity to study transitions across such a magnetic barrier. 
1282: Irrespective of whether the centrifugal or magnetic barrier operates in \IGR,\ 
1283: a long spin period would imply a high magnetic field, comparable to 
1284: those inferred for magnetar candidates. 
1285: 
1286: Scenarios involving magnetars with spin periods well above 1000~s 
1287: have been considered in several studies \citep{ruth, mori, toropina2, liu, zhang04}.   
1288: Moreover a few known accreting X-ray pulsars with unusually long spin 
1289: periods have already been proposed as magnetar candidates 
1290: \citep[see e.g.,][and references therein]{ikhsanov07}. 
1291: However these sources display different properties from SFXTs: some are persistent 
1292: sources; others display week to month-long outbursts; the high spin up measured 
1293: in \igrpatel\ testifies that the accretion flow is likely mediated by a disk \citep{patel07}. 
1294: Therefore some of the features of the model discussed in this paper would 
1295: not be applicable to these sources. 
1296: 
1297: In the context of wind-fed HMXBs,  
1298: \citet{zhang04} pointed out that magnetars might be hosted in binary systems 
1299: with relatively short orbits ($\sim$1-100~d) 
1300: and long pulse period ($\gtrsim$10$^{3}$-10$^{5}$~s). 
1301: By using evolutionary calculations, these authors 
1302: showed that magnetars would be easily spun-down to such long spin periods by 
1303: the interaction with the wind of the companion star, in less than 10$^{6}$~yr. 
1304: The systems we considered in this paper would   
1305: likely result from a similar evolutionary path. 
1306: 
1307: Once a system approaches the spin period that is required for 
1308: direct accretion to occur, the short timescale ($\sim$~hours) erratic 
1309: variations that characterize a supergiant's wind  
1310: will cause transitions across different NS regimes.   
1311: Insofar as the average wind properties evolve only secularly, the 
1312: NS spin will then remain locked around such period, alternating spin-up 
1313: intervals during accretion and spin-down intervals during quiescence, 
1314: when accretion is inhibited. 
1315: 
1316: The relevant spin-up timescale during accretion intervals is 
1317: approximately \citep[see e.g.][]{fkr} 
1318: \begin{equation}
1319: \tau_{\rm su} =  -P_{\rm spin}/\dot{P}_{\rm spin}=\Omega I/(\dot{M}_{\rm capt} l)   
1320: \simeq 8\times10^2 I_{45} v_{8}^{4} P_{\rm 10d} P_{\rm s3}^{-1} \dot{M}_{17}^{-1} ~{\rm yr},     
1321: \label{eq:spinup}
1322: \end{equation}
1323: where l=2$\pi$R$_{\rm a}^2$/(4P$_{\rm orb}$) the specific 
1324: angular momentum of wind matter at the accretion radius, 
1325: I=10$^{45}$I$_{45}$~g~cm$^2$ the NS moment of inertia, and 
1326: $\dot{M}_{17}$=$\dot{M}_{\rm capt}$/10$^{17}$~g~s$^{-1}$ 
1327: (this corresponds to an outburst luminosity 
1328: of $\sim$10$^{37}$erg s$^{-1}$).   
1329: 
1330: A rough estimate of the spin-down timescale is 
1331: obtained by assuming that most of the quiescence luminosity 
1332: draws from the rotational energy of the NS \citep{pringle}; 
1333: this gives
1334: \begin{equation}  
1335: \tau_{\rm sd}=P_{\rm spin}/\dot{P}_{\rm spin}=I \Omega^2/L_{\rm X}=
1336: 1.3\times10^2 I_{45} L_{31}^{-1} P_{\rm s3}^{-2} ~{\rm yr},   
1337: \label{eq:spindown}
1338: \end{equation}
1339: where L$_{31}$ is the luminosity produced by the 
1340: interaction between the magnetosphere and the wind  
1341: in units of 10$^{31}$~erg~s$^{-1}$. 
1342: It is apparent that, for magnetar-like fields and long spin periods 
1343: the above timescales are much shorter than the lifetime of 
1344: the supergiant's strong wind phase. Therefore secular changes 
1345: in the wind parameters and/or the NS magnetic field 
1346: will be easily tracked by the NS spin. 
1347: Moreover since spin-up and spin-down take place on
1348: comparable timescales, it is to be expected that spin-up 
1349: intervals during outbursts are compensated for by 
1350: spin-down during quiescent intervals of 
1351: comparable duration. In other words accretion state of 
1352: long spin period SFXTs would be expected to have 
1353: a duty cycle of order $\sim$0.5 or higher. As we discussed in 
1354: \S~\ref{sec:properties}, evidence for high values of the duty cycle 
1355: is gradually emerging from high sensitivity observations 
1356: of SFXTs.
1357: 
1358: On the other hand spin-down may also occur during the accretion 
1359: intervals as a results of velocity and density gradients in the 
1360: supergiant's wind that lead to temporary reversals 
1361: of the angular momentum of the captured wind relative to 
1362: the neutron star \citep[note that persistent wind accreting 
1363: X-ray pulsars in HMXBs have long been known to alternate 
1364: spin-up and spin-down intervals, see e.g.][]{henrichs}.
1365: This would tend to favor spin-down, such that the NS spin might 
1366: gradually evolve longwards of the spin period that is required 
1367: to inhibit direct accretion and a transient source becomes a persistent
1368: one. Interestingly, this might apply to \2s,\ a persistent 
1369: X-ray pulsar with a luminosity of $\sim$10$^{36}$~erg~s$^{-1}$,  
1370: which displayed variations by a factor of $\lesssim$10. 
1371: Since its spin period is extremely long ($\sim 10^4$)
1372: $R_{\rm M}$ is smaller than the 
1373: corotation radius and the NS accretes continuously from the relatively 
1374: weak supergiant companion's wind \citep[see also][]{li99}. 
1375: 
1376: The SFXTs population might thus represent those 
1377: supergiant HMXBs systems that have not (yet) evolved 
1378: away from the spin period at which transitions across 
1379: different NS state can take place. 
1380: 
1381: We note that for SFXTs to host magnetars their 
1382: dipole magnetic field must retain values in the 
1383: $\mu_{33}$=0.1-1  range for a few 10$^6$~yr, i.e. the typical
1384: timescale from the formation of the NS to the onset 
1385: of the supergiant's strong wind. 
1386: Presently known magnetar candidates (soft gamma repeaters, SGR, 
1387: and anomalous X-ray pulsars, AXP) have estimated ages in the 10$^4$-10$^5$~yr 
1388: range \citep[for a review see e.g.][]{woods}. 
1389: Little is known on the long term evolution of the 
1390: magnetic field of magnetars and different models have 
1391: been proposed which lead to different predictions. 
1392: We note that if the irrotational mode of 
1393: ambipolar diffusion dominates the B-field decay, 
1394: $\mu_{33}$=0.1-0.3 can be expected for ages of a few 10$^6$~yr
1395: \citep{heyl98}.  
1396:  
1397: According to our proposed scenario the combination of a long spin 
1398: period and a very large luminosity swing is indicative 
1399: of the presence of a magnetar. This can be further corroborated 
1400: through other magnetars signatures, such as e.g. proton cyclotron 
1401: features in the X-ray spectrum \citep{zane01} 
1402: or sporadic subsecond bursting activity such 
1403: as that observed in AXPs and SGRs \citep{gavriil02}.  
1404: 
1405: Finally we remark on the orbital period of SFXTs: in all 
1406: regimes described in \S~\ref{sec:model}, the luminosity 
1407: scales with L$_{\rm X}$$\propto$$P_{\rm orb}^{-4/3}$v$_{\rm w}^{-4}$.  
1408: For our fiducial wind parameters, orbital periods of  
1409: tens of days are required for the transition 
1410: between low and high luminosity states to occur at 
1411: $\simeq$10$^{36}$~erg~s$^{-1}$, a typical  
1412: luminosity for the onset of SFXT outbursts \citep{walter07}. 
1413: This is why throughout this paper we scaled our equations by 
1414: P$_{\rm orb}$=10~d and used the same value in the examples of Figs.~\ref{fig:trans}-\ref{fig:IGRJ17544}. 
1415: \citet{negueruela08} showed that orbital periods around 
1416: $\sim$10~d are compatible with the NSs being embedded in the   
1417: clumpy wind from the supergiant companions, rather 
1418: than in a quasi-continuous wind. In this case SFXT outburst 
1419: durations might be associated with the transit time of a clump
1420: \begin{equation}
1421: \tau_{\rm out}\simeq a/v_{\rm w}=4.2\times10^{4} 
1422: a_{\rm 10d} v_{8}^{-1} ~ {\rm s}, 
1423: \end{equation} 
1424: in reasonable agreement with the observed durations of individual flares 
1425: (see \S~\ref{sec:properties}). 
1426: 
1427: We conclude that clumpiness of the stellar wind, an often-used 
1428: concept for interpreting SFXT activity, applies to the gating 
1429: scenarios described here as well. The main advantage of introducing 
1430: a gating mechanism rests with the possibility to model 
1431: the very large luminosity swings of 
1432: SFXTs with much milder density (or velocity) 
1433: contrasts in the wind. 
1434: 
1435:  
1436: 
1437: \section{Conclusions}
1438: \label{sec:conclusions}
1439: 
1440: In this paper we reviewed the theory of wind accretion in HMXBs hosting a magnetic
1441: neutron star with a supergiant companion, and considered in some detail the 
1442: interaction processes between the inflowing plasma and the magnetosphere
1443: that are expected to take place when direct accretion onto the 
1444: neutron star surface is inhibited. We then applied this theory to SFXTs   
1445: and showed that their large luminosity swings
1446: between quiescence and outburst (up to a factor of $\sim$10$^{5}$)
1447: can be attained in response to relatively modest variations of the 
1448: wind parameters, provided the system undergoes transitions across different 
1449: regimes. Expanding on earlier work, we found that such transitions can be 
1450: driven mainly by 
1451: either: (a) a centrifugal barrier mechanism, which halts direct accretion 
1452: when the neutron star rotation becomes superKeplerian at the magnetospheric 
1453: radius, a mechanism that has been discussed extensively in Be star X-ray transient 
1454: pulsars, or (b) a magnetic barrier mechanism, when the magnetosphere 
1455: extends beyond the accretion radius. Which mechanism  
1456: and wind interaction regime apply will depend  
1457: sensitively on the NS spin period and magnetic field, 
1458: besides the velocity and mass loss rate in the supergiant's wind. 
1459: In particular, the magnetic barrier mechanism requires 
1460: long spin periods ($\gtrsim$1000~s) coupled with magnetar-like 
1461: fields ($\gtrsim$10$^{14}$~G). 
1462: On the other hand, magnetar-like 
1463: fields would also be required if the centrifugal barrier set in  
1464: at relatively high luminosities ($\gtrsim 10^{36}$~erg~s$^{-1}$)  
1465: in neutron stars with spin periods of hundreds seconds. 
1466: 
1467: Evidence has been found that the spin periods of a few SFXTs 
1468: might be as long as 1000-2000~s. Motivated by this, we 
1469: presented an interpretation of the 
1470: activity of \IGR\ (whose spin period is unknown) in terms of the   
1471: magnetic barrier by a 1300~s spinning neutron star 
1472: and showed that the luminosity
1473: stages singled out in a \chan\ observation of this source 
1474: are well matched by the different regimes of wind-magnetosphere
1475: interaction expected in this case.
1476: We discussed also an interpretation  
1477: of this source based on the centrifugal barrier and a slightly  shorter 
1478: spin period (400~s), which reproduced the luminosity stages comparably well. 
1479: We emphasise that in both solutions the required magnetic field 
1480: strength ($\gtrsim$10$^{15}$~G and $\gtrsim$8$\times$10$^{13}$~G, respectively) 
1481: are in the magnetar range. 
1482: 
1483: While the possibility that magnetars are hosted in binary system with supergiant 
1484: companions has been investigated by several authors \citep[e.g.,][]{zhang04,liu},   
1485: clear observational evidence for such extremely high magnetic field neutron
1486: star in binary systems is still missing. According to the present study, 
1487: long spin period SFXTs might provide a new prospective for detecting and studying 
1488: magnetars in binary systems.  
1489: 
1490: \acknowledgements
1491: EB thanks the CEA Saclay, 
1492: DSM/DAPNIA/Service d'Astrophysique 
1493: for hospitality during part of this work. 
1494: MF acknowledges the French Space
1495: Agency  (CNES) for financial support.  
1496: We would like to thank the referee for useful comments.
1497: This work was partially supported through ASI 
1498: and MUR grants.
1499: 
1500: 
1501: \appendix
1502: 
1503: 
1504: \section{On the height of the KHI unstable layer}
1505: \label{app:ht}
1506: In this section we expand on the approximation 
1507: h$_{\rm t}$$\sim$R$_{\rm M}$, introduced previously in \S~\ref{sec:intermediate}. 
1508: As discussed by \citet{bur}, the height h$_{\rm t}$ of the layer where matter and magnetic field 
1509: coexist due to the KHI, is mostly determined by the largest wavelength unstable 
1510: mode of the KHI itself. These authors suggested that 
1511: \begin{equation}
1512: h_{\rm t}\simeq h H_{\rm s}, 
1513: \label{eq:ht}
1514: \end{equation} 
1515: where $h$ is a factor of order $\sim$1, and 
1516: H$_{\rm s}$=R$_{\rm M}^2$ k$_{\rm b}$ T(R$_{\rm M}$)/(GM m$_{\rm p}$)  
1517: is the scale height of the 
1518: magnetosheath (note that H$_{\rm s}$ is roughly of the same order as
1519: R$_{\rm M}$$\simeq$10$^{10}$~cm for T$\sim$10$^{8}$~K). 
1520: In the subsonic propeller regime, Eq.~\ref{eq:ht} might be a reasonable assumption 
1521: due to the presence of an extended atmosphere around the magnetospheric boundary; 
1522: on the contrary, it cannot be used in the subKeplerian magnetic 
1523: inhibition regime, where matter flowing toward the NS is shocked very close to the 
1524: magnetospheric boundary. Despite all these uncertainties, we show below that 
1525: the assumption h$_{\rm t}$=h R$_{\rm M}$, with h$\sim$1, gives a conservative 
1526: estimate of the mass accretion rate due to the KHI. 
1527: Using Eq.~\ref{eq:masscons} and the definition 
1528: v$_{\rm conv}$=$\eta_{\rm KH}$v$_{\rm sh}$($\rho_{\rm i}$/$\rho_{\rm e}$)$^{1/2}$
1529: (1+$\rho_{\rm i}$/$\rho_{\rm e}$)$^{-1}$, we derive the equations that define 
1530: the ratio of the densities inside ($\rho_{\rm i}$) and outside ($\rho_{\rm e}$) 
1531: the magnetosphere. These are 
1532: \begin{equation}
1533: (\rho_{\rm i}/\rho_{\rm e})^{1/2}(1+\rho_{\rm i}/\rho_{\rm e})=
1534: 0.3 \eta_{\rm KH} h^{-1} R_{\rm M10}^{3/2} P_{\rm s3}^{-1}, 
1535: \label{eq:rho1}
1536: \end{equation} 
1537: if v$_{\rm sh}$=v$_{\rm rot}$, and 
1538: \begin{equation}
1539: (\rho_{\rm i}/\rho_{\rm e})^{1/2}(1+\rho_{\rm i}/\rho_{\rm e})=
1540: 0.1 \eta_{\rm KH} h^{-1} R_{\rm M10}^{1/2} v_{8}, 
1541: \label{eq:rho2}
1542: \end{equation} 
1543: if v$_{\rm sh}$=v$_{\rm ps}$.  
1544: Equations~\ref{eq:rho1} and \ref{eq:rho2} show that $\rho_{\rm i}/\rho_{\rm e}$ is an   
1545: increasing function of h$^{-1}$ (for fixed values of v$_{8}$, P$_{\rm s3}$ and R$_{\rm M10}$). 
1546: Therefore, being 
1547: v$_{\rm conv}$$\propto$($\rho_{\rm i}$/$\rho_{\rm e}$)$^{1/2}$(1+$\rho_{\rm i}$/$\rho_{\rm e}$)$^{-1}$ 
1548: and $\dot{M}_{\rm KH}$$\propto$v$_{\rm conv}$, 
1549: the KHI rate of plasma entry inside the magnetosphere is 
1550: also an increasing function of  h$^{-1}$ (provided that $\rho_{\rm i}$$\lesssim$$\rho_{\rm e}$). 
1551: Since the KHI unstable layer is located inside the NS magnetosphere, the maximum height attainable is 
1552: h$_{\rm t}$=R$_{\rm M}$, and thus the approximation used in \S~\ref{sec:intermediate} and 
1553: \S~\ref{sec:subsonic} gives a  lower limit on the mass flow rate controlled by the KHI in both 
1554: the subKeplerian magnetic inhibition and the subsonic propeller regime. 
1555: 
1556: Note also that the above lower limit does not violate the stability condition   of the 
1557: quasi-static atmosphere in the subsonic propeller regime. 
1558: In fact, following \citet{ikhsanov01b}, this atmosphere can be considered quasi-static  
1559: if the relaxation time scale of the envelope
1560: is less than the drift time scale of the mass flow crossing the magnetospheric boundary.  
1561: In our case matter penetration inside the NS magnetosphere in the subsonic propeller 
1562: regime is mostly provided by the KHI, and the above condition translates into 
1563: $\dot{M}_{\rm KH}$$\lesssim$(R$_{\rm M}$/R$_{\rm co}$)$^{3/2}$$\dot{M}_{\rm capt}$, 
1564: which is satisfied for a wide range of parameters (see Figs.~\ref{fig:magn_e_non}, 
1565: \ref{fig:lshortspin}, and \ref{fig:vwind}).  
1566: 
1567: An upper limit to the KHI mass flow rate can be obtained by assuming 
1568: $\rho_{\rm i}$/$\rho_{\rm e}$=1 \citep[a solution 
1569: adopted, for example, by][]{ruth}. 
1570: According to \citet{bur}, the density $\rho_{\rm i}$ can be increased 
1571: until the thermal pressure inside the magnetosphere p$_{\rm i}$$\propto$$\rho_{\rm i}$ 
1572: is comparable to the magnetic pressure p$_{\rm m}$=B$^2$(R$_{\rm M}$)/(8$\pi$). When this 
1573: limit is reached, an instability occurs on the lower surface of the unstable 
1574: layer that increases h$_{\rm t}$ until p$_{\rm i}$$<$p$_{\rm m}$ is restored. 
1575: Using the post-shocked gas temperature at R$_{\rm M}$ (see \S~\ref{sec:model}), 
1576: it is shown that $\rho_{\rm i}$/$\rho_{\rm e}$$\sim$1 does not violate the condition 
1577: p$_{\rm i}$$<$p$_{\rm m}$ (at least in the subKeplerian magnetic inhibition regime). 
1578: Therefore, even though we restricted ourself to the lower limit h=1, 
1579: the upper limit on $\dot{M}_{\rm KH}$ might be attainable in some instances.  
1580: The KHI would then provide matter penetration inside the magnetosphere at a rate 
1581: $\sim$$\dot{M}_{\rm capt}$, thus allowing almost all the captured matter to 
1582: accrete onto the NS. A detailed calculation of KHI accretion 
1583: in the subKeplerian magnetic inhibition and subsonic propeller regimes will be reported 
1584: elsewhere. 
1585: 
1586: 
1587: \section{Radiative losses in the supersonic propeller}
1588: \label{app:supersonic}
1589: As shown by \citet{pringle}, the treatment of the supersonic propeller 
1590: regime (see \S~\ref{sec:supersonic}) is self consistent 
1591: only if the energy input at the base of the atmosphere, due to the 
1592: supersonic rotating NS magnetosphere, is larger than radiative losses 
1593: within the atmosphere itself. 
1594: The range of validity of this assumption can be determined by using the 
1595: convective efficiency parameter
1596: \begin{equation}
1597: \Gamma=\mathfrak{M}_{\rm t}^2 v_{\rm t}(R) t_{\rm br}(R) R^{-1}, 
1598: \end{equation}   
1599: where $\mathfrak{M}_{\rm t}$=v$_{\rm t}$(R)/c$_{\rm s}$(R) is the Mach number, 
1600: v$_{\rm t}$ and c$_{\rm s}$ are the turbulent and sound velocity, and 
1601: t$_{\rm br}$=2$\times$10$^{11}$ T$^{1/2}$m$_{\rm p}$$\rho^{-1}$(R)~s is the bremsstralhung 
1602: cooling time. 
1603: For most of the NS rotational energy dissipated at the magnetospheric 
1604: radius to be convected away through the atmosphere's outer boundary, 
1605: $\Gamma$ should be $\gtrsim$1 across the entire envelope. 
1606: Taking into account that in the supersonic propeller regime  
1607: v$_{\rm t}$$\simeq$c$_{\rm s}$, c$_{\rm s}$$\sim$v$_{\rm ff}$, T$\sim$T$_{\rm ff}$ and $\rho$  
1608: is given by Eq.~\ref{eq:hydro}, one finds $\Gamma$$\propto$R$^{-3/2}$.  
1609: Therefore the said requirement is satisfied when $\Gamma$(R$_{\rm a}$)$\gtrsim$1, i.e.  
1610: \begin{equation}
1611: \dot{M}_{-6}\lesssim 2.2\times10^2 v_{8}^5 a_{\rm 10d}^2.  
1612: \label{eq:consistencysuper}
1613: \end{equation}
1614: For mass loss rates larger than the above limit, radiative losses are not 
1615: negligible and the treatment used in \S~\ref{sec:supersonic} for the 
1616: supersonic propeller regime is no longer self-consistent. 
1617: We checked that the limit of Eq.~\ref{eq:consistencysuper} is never 
1618: exceeded in the cases of interest. We note that a similar value 
1619: was also derived by \citet{ikhsanov02}, but his 
1620: limit is a factor 10 larger than ours. This might be due to the fact that
1621: \citet{ikhsanov02} used $\rho_{\rm w}$ instead of $\rho_{\rm ps}$(1+16/3)$^{1/2}$ 
1622: in the expression for the matter density at R$_{\rm a}$. 
1623: 
1624: \section{Radiative losses in the subsonic propeller}
1625: \label{app:subsonic}
1626: A similar calculation to that in Appendix~\ref{app:supersonic} in the 
1627: subsonic propeller regime shows that 
1628: $\Gamma$$\propto$R$^{1/2}$, and radiative losses are negligible if    
1629: $\Gamma$(R$_{\rm M}$)$\gtrsim$1, i.e. 
1630: \begin{equation}
1631: \dot{M}_{-6}\lesssim2.8\times10^2 P_{\rm s3}^{-3} 
1632: a_{\rm 10d}^{2} v_8 R_{\rm M10}^{5/2} (1+16 R_{\rm a10}/(5 R_{\rm M10}))^{-3/2}, 
1633: \label{eq:consistencysub}
1634: \end{equation}
1635: that is the same value as that in Eq.~\ref{eq:dotmlimsub} (a somewhat different value was 
1636: obtained by \citet{ikhsanov01a} assuming a density $\simeq$$\rho_{\rm w}$ at R$_{\rm M}$ 
1637: instead of $\rho_{\rm ps}$(1+16R$_{\rm M}$/(5R$_{\rm a}$))$^{3/2}$).    
1638: 
1639: 
1640: \begin{thebibliography}{}
1641: 
1642: \bibitem[\protect\citeauthoryear{Bamba et al.}{2001}]{bamba01}
1643: Bamba, A., Yokogawa, J., Ueno, M., Koyama, K., \& Yamauchi, S. 
1644: 2001, PASJ, 53, 1179
1645: 
1646: \bibitem[\protect\citeauthoryear{Berghoefer et al.}{2001}]{berghofer97} 
1647: Berghoefer, T. W., Schmitt, J. H. M. M., Danner, R., \& Cassinelli, J. P. 
1648: 1997, A\&A, 322, 167
1649: \bibitem[\protect\citeauthoryear{Bondi}{1952}]{bondi} 
1650: Bondi, H. 1952, MNRAS, 112, 195
1651: 
1652: \bibitem[\protect\citeauthoryear{Burnard et al.}{1983}]{bur} 
1653: Burnard, D. J., Arons, J., \& Lea, S. M. 1983, ApJ, 266, 175
1654: 
1655: \bibitem[\protect\citeauthoryear{Campana et al.}{1998}]{campana98}
1656: Campana, S., Colpi, M., Mereghetti, S., Stella, L., Tavani, M. 1998, A\&ARv, 8, 279 
1657: 	
1658: \bibitem[\protect\citeauthoryear{Cassinelli et al.}{1981}]{cassinelli81} 
1659: Cassinelli, J. P., Waldron, W. L., Sanders, W. T., Harnden, F. R., Rosner, R., \& Vaiana, G. S. 
1660: 1981, ApJ, 250, 677
1661: 
1662: \bibitem[\protect\citeauthoryear{Corbet}{1996}]{corbet96}
1663: Corbet, R. H. D. 1996, ApJ, 457, L31 
1664: 
1665: \bibitem[\protect\citeauthoryear{Corbet et al.}{2006}]{corbet06} 
1666: Corbet, R., et al. 2006, Astr. Tel., 779  
1667: 
1668: \bibitem[\protect\citeauthoryear{Davidson \& Ostriker}{1979}]{davidson}
1669: Davidson, K. \& Ostriker, J. P. 1979, ApJ, 179, 585
1670: 
1671: \bibitem[\protect\citeauthoryear{Davies et al.}{1979}]{davies} 
1672: Davies, R. E., Fabian, A. C., \& Pringle, J. E. 1979, MNRAS, 186, 779 
1673: 
1674: \bibitem[\protect\citeauthoryear{Davies \& Pringle}{1981}]{pringle} 
1675: Davies, R. E. \& Pringle, J. E. 1981, MNRAS, 196, 209 
1676: 
1677: \bibitem[\protect\citeauthoryear{Dessart \& Owocki}{2003}]{dessart03}
1678: Dessart, L. \& Owocki, S.P. 2003, A\&A, 406, L1  
1679: 
1680: \bibitem[\protect\citeauthoryear{Duncan \& Thompson}{1992}]{dun} 
1681: Duncan, R. C. \& Thompson, C. 1992, ApJ, 392, L9
1682: 
1683: \bibitem[\protect\citeauthoryear{Elsner \& Lamb}{1977}]{elsner1977} 
1684: Elsner, R. F. \& Lamb, F. K. 1977, ApJ, 215, 897
1685: 
1686: \bibitem[\protect\citeauthoryear{Frank et al.}{2002}]{fkr} 
1687: Frank J., King A., \& Rayne D. 2002, Accretion Power in Astrophysics 
1688: (3th ed., Cambridge: Cambridge University Press)
1689: 
1690: \bibitem[\protect\citeauthoryear{Gavriil et al.}{2002}]{gavriil02}
1691: Gavriil, F. P., Kaspi, V. M., \& Woods, P. M. 2002, Nature, 419, 142 
1692: 	
1693: \bibitem[\protect\citeauthoryear{Gonz\'alez-Riestra et al.}{2004}]{gon04}
1694: Gonz\'alez-Riestra, R., Oosterbroek, T., Kuulkers, E., Orr, A., \& Parmar, A. N. 
1695: 2004, A\&A, 420, 589 
1696: 
1697: \bibitem[\protect\citeauthoryear{G\"otz et al.}{2007}]{gotz07}
1698: G\"otz, D., Falanga, M., Senziani, F., De Luca, A., Schanne, S., \& von Kienlin, A. 
1699: 2007, ApJ, 655, L101 
1700: 
1701: \bibitem[\protect\citeauthoryear{Grebenev \& Sunyaev}{2005}]{grebenev05}
1702: Grebenev S. A. \& Sunyaev, R. A., 2005, AstL, 31, 672
1703: 
1704: \bibitem[\protect\citeauthoryear{Grebenev \& Sunyaev}{2007}]{grebenev07}
1705: Grebenev S. A. \& Sunyaev, R. A., 2005, AstL, 33, 149
1706: 
1707: \bibitem[\protect\citeauthoryear{Halpern et al.}{2004}]{halpern04} 
1708: Halpern, J. P., Gotthelf, E. V., Helfand, D. J., Gezari, S., \& Wegner, G. A. 
1709: 2004, Astr. Tel., 289 
1710: 
1711: \bibitem[\protect\citeauthoryear{Harding et al.}{1992}]{hard} 
1712: Harding, A. K. \& Leventhal, M. 1992, Nature, 357, 388
1713: 
1714: \bibitem[\protect\citeauthoryear{Heyl \& Kulkarni}{1998}]{heyl98} 
1715: Heyl, J. S. \& Kulkarni, S. R. 1998, ApJ, L61 
1716: 
1717: \bibitem[\protect\citeauthoryear{Henrichs}{1998}]{henrichs}
1718: Henrichs, H. F., 1983, in Accretion-driven stellar X-ray sources, ed. W. H. G. Lewin \&  E. P. J. van den Heuvel   
1719: (Cambridge, UK: Cambridge University Press), 393 
1720: 
1721: \bibitem[\protect\citeauthoryear{Ikhsanov \& Pustil'nik}{1996}]{ikhsanov} 
1722: Ikhsanov, N. R. \& Pustil'nik, L. A. 1996, A\&A, 312, 338
1723: 
1724: \bibitem[\protect\citeauthoryear{Ikhsanov}{2001a}]{ikhsanov01a} 
1725: Ikhsanov, N. R. 2001a, A\&A, 368, L5
1726: 
1727: \bibitem[\protect\citeauthoryear{Ikhsanov et al.}{2001b}]{ikhsanov2001} 
1728: Ikhsanov, N. R., Larionov, V. M., \& Beskrovnaya, N. G. 2001b, A\&A, 372, 227
1729: 
1730: \bibitem[\protect\citeauthoryear{Ikhsanov}{2001c}]{ikhsanov01b} 
1731: Ikhsanov, N. R. 2001c, A\&A, 375, 944
1732: 
1733: \bibitem[\protect\citeauthoryear{Ikhsanov}{2002}]{ikhsanov02} 
1734: Ikhsanov, N. R. 2002, A\&A, 381, L61
1735: 
1736: \bibitem[\protect\citeauthoryear{Ikhsanov}{2007}]{ikhsanov07}
1737: Ikhsanov, N. R. 2007, MNRAS, 375, 698 
1738: 
1739: \bibitem[\protect\citeauthoryear{Illarionov \& Sunyaev}{1975}]{illarionov75} 
1740: Illarionov, A. F. \& Sunyaev, R. A. 1975, A\&A, 39, 185 
1741: 
1742: \bibitem[\protect\citeauthoryear{In't Zand}{2005}]{zand05}
1743: In 't Zand 2005, A\&A, 441, L1
1744: 
1745: \bibitem[\protect\citeauthoryear{Kennea \& Campana}{2006}]{kennea06}
1746: Kennea, J. A. \& Campana, S. 2006, Astr. Tel., 818 
1747: 
1748: \bibitem[\protect\citeauthoryear{Leyder et al.}{2007}]{leyder} 
1749: Leyder, J.-C., Walter, R., Lazos, M., Masetti, N., \& Produit, N. 2007, A\&A, 465, L35
1750: 
1751: \bibitem[\protect\citeauthoryear{Li \& van den Heuvel}{1999}]{li99} 
1752: Li, X.-D. \& van den Heuvel, E.P.J. 1999, ApJ, 513, L45 
1753: 
1754: \bibitem[\protect\citeauthoryear{Lipunov}{1992}]{lipunov02}
1755: Lipunov, V. M. 1992, Astrophysics of Neutron Stars (Springer-Verlag)
1756: 
1757: \bibitem[\protect\citeauthoryear{Liu \& Yan}{2006}]{liu}
1758: Liu, Q. Z. \& Yan, J. Z. 2006, AdSpR, 38, 2906 
1759: 
1760: \bibitem[\protect\citeauthoryear{Lutovinov et al.}{2005}]{Lutovinov05}
1761: Lutovinov, A., Revnivtsev, M., Gilfanov, M., Shtykovskiy, P., Molkov, S., \& Sunyaev, R. 
1762: 2005, A\&A, 444, 821
1763: 
1764: \bibitem[\protect\citeauthoryear{Masetti et al.}{2006}]{masetti06}
1765: Masetti, N., et al. 2006, A\&A, 449, 1139  
1766: 
1767: \bibitem[\protect\citeauthoryear{Mori \& Ruderman}{2003}]{mori} 
1768: Mori, K. \& Ruderman, M. A. 2003, ApJ, 592, L75
1769: 
1770: \bibitem[\protect\citeauthoryear{Negueruela et al.}{2008}]{negueruela08}
1771: Negueruela, I., Torrejon, J. M., Reig, P., Ribo, M., \& Smith, D. M. 2008, 
1772: preprint (astro-ph/0801.3863) 
1773: 
1774: \bibitem[\protect\citeauthoryear{Nespoli et al.}{2007}]{nespoli07} 
1775: Nespoli, E., Fabregat, J., \& Mennickent, R. 2007, Astr. Tel., 983
1776: 
1777: \bibitem[\protect\citeauthoryear{Patel et al.}{2007}]{patel07} 
1778: Patel, S.K., et al. 2007, ApJ, 657, 1003 
1779: 
1780: \bibitem[\protect\citeauthoryear{Perna et al.}{2006}]{perna06} 	
1781: Perna, R., Bozzo, E., \& Stella, L. 2006, ApJ, 639, 363
1782: 
1783: \bibitem[\protect\citeauthoryear{Prinja et al.}{2005}]{prinja} 
1784: Prinja, R. K., Massa, D., \& Searle, S. C. 2005, A\&A, 430, L41 
1785: 
1786: \bibitem[\protect\citeauthoryear{Rahoui et al.}{2008}]{rahoui08} 
1787: Rahoui, F., Chaty, S., Lagage, P.-O., Pantin, E. 2008, preprint (astro-ph/0802.1770)
1788: 
1789: \bibitem[\protect\citeauthoryear{Rutledge}{2001}]{ruth} 
1790: Rutledge, E. R. 2001, ApJ, 553, 796 
1791: 
1792: \bibitem[\protect\citeauthoryear{Sguera et al.}{2005}]{sguera05} 
1793: Sguera, V., et al. 2005, A\&A, 444, 221
1794: 
1795: \bibitem[\protect\citeauthoryear{Sguera et al.}{2006}]{sguera06} 
1796: Sguera, V., et al. 2006, ApJ, 646, 452
1797: 
1798: \bibitem[\protect\citeauthoryear{Sguera et al.}{2007}]{sguera07} 
1799: Sguera, V., et al. 2007, A\&A, 462, 695
1800: 
1801: \bibitem[\protect\citeauthoryear{Sidoli et al.}{2005}]{sidoli05} 
1802: Sidoli, L., Vercellone, S., Mereghetti, S., \& Tavani, M. 2005, A\&A, 429, L47
1803: 
1804: \bibitem[\protect\citeauthoryear{Sidoli et al.}{2007}]{sidoli07} 
1805: Sidoli, L., Romano, P., Mereghetti, S., Paizis, A., Vercellone, S., Mangano, V., \& Götz, D. 
1806: 2007, A\&A, 476, 1307  
1807: 
1808: \bibitem[\protect\citeauthoryear{Smith}{2004}]{smith04} 
1809: Smith, D. M. 2004, Astr. Tel., 338 
1810: 
1811: \bibitem[\protect\citeauthoryear{Smith et al.}{2006}]{smith06}
1812: Smith, D. M., Heindl, W. A., Markwardt, C. B., Swank, J. H., Negueruela, I., Harrison, T. E., \& Huss, L. 
1813: 2006, ApJ, 638, 974
1814: 
1815: \bibitem[\protect\citeauthoryear{Stella et al.}{1986}]{stella86} 
1816: Stella, L., White, N. E., \& Rosner, R. 1986, ApJ, 308, 669
1817: 
1818: \bibitem[\protect\citeauthoryear{Sunyaev et al.}{2003}]{sunyaev03}  
1819: Sunyaev, R. A., Grebenev, S. A., Lutovinov, A. A., Rodriguez, J., Mereghetti, S., Gotz, D., \& Courvoisier, T. 
1820: 2003, Astr. Tel., 190
1821: 
1822: \bibitem[\protect\citeauthoryear{Tauris \& van den Heuvel}{2006}]{tauris} 
1823: Tauris, T. M. \& van den Heuvel, E. P. J. 2006, in Formation and evolution 
1824: of compact stellar X-ray sources, ed. W. Lewin \& M. van der Klis (Cambridge, UK: CambridgeUniversity Press), 623
1825: 
1826: \bibitem[\protect\citeauthoryear{Tomsick et al.}{2004}]{tomsick04}
1827: Tomsick, J. A., Lingenfelter, R., Corbel, S., Goldwurm, A., \& Kaaret, P. 2004, Astr. Tel., 224 
1828: 
1829: \bibitem[\protect\citeauthoryear{Tomsick et al.}{2006}]{tomsick06} 
1830: Tomsick, J. A., Chaty, S., Rodriguez, J., Foschini, L., Walter, R., \& Kaaret, P. 
1831: 2006, ApJ, 647, 1309 
1832: 
1833: \bibitem[\protect\citeauthoryear{Toropina et al.}{2006}]{toropina} 
1834: Toropina, O. D., Romanova, M. M., \& Lovelace, R. V. E. 2006, MNRAS, 371, 569 
1835: 
1836: \bibitem[\protect\citeauthoryear{Toropina et al.}{2001}]{toropina2} 
1837: Toropina, O. D., Romanova, M. M., Toropin, Y. M., \& Lovelace, R. V. E. 2001, ApJ, 561, 964
1838: 
1839: \bibitem[\protect\citeauthoryear{van den Heuvel \& Rappaport}{2005}]{heuvel} 
1840: van den Heuvel, E. P. J. \& Rappaport, S. 1987, in Physics of Be stars (Cambridge, UK: Cambridge University Press), 291
1841: 
1842: \bibitem[\protect\citeauthoryear{Verbunt \& van den Heuvel}{1995}]{verbunt} 
1843: Verbunt, F. \& van den Heuvel, E. P. J. 1995, X-ray binaries, ed. W. H. G. Lewin, J. van Paradijs, E. P. J. van den Heuvel, 
1844: (Cambridge, UK: Cambridge University Press), 457 
1845: 
1846: \bibitem[\protect\citeauthoryear{Walter et al.}{2006}]{walter06}
1847: Walter, R., et al. 2006, A\&A, 453, 133
1848: 
1849: \bibitem[\protect\citeauthoryear{Walter \& Zurita Heras}{2007}]{walter07}
1850: Walter, R. \& Zurita Heras, J. A. 2007, A\&A, 476, 335
1851: 
1852: \bibitem[\protect\citeauthoryear{Woods \& Thompson}{2006}]{woods}
1853: Woods, P. M. \& Thompson, C. 2006, in Compact stellar X-ray sources, ed. W. Lewin \& M. van der Klis 
1854: (Cambridge, UK: Cambridge University Press), 547 
1855: 
1856: \bibitem[\protect\citeauthoryear{Zane et al.}{2001}]{zane01}
1857: Zane, S., Turolla, R., Stella, L., \& Treves, A. 2001, ApJ, 560, 384 
1858: 
1859: \bibitem[\protect\citeauthoryear{Zhang et al.}{2004}]{zhang04} 
1860: Zhang, F., Li, X.-D. \& Wang, Z.-R. 2004, ChjAA, 4, 320
1861: 
1862: \bibitem[\protect\citeauthoryear{Zurita Heras \& Walter}{2004}]{zurita04}
1863: Zurita Heras, J. A. \&  Walter, R. 2004, Astr. Tel., 336 
1864: 
1865: \bibitem[\protect\citeauthoryear{Zurita Heras et al.}{2007}]{zurita07}
1866: Zurita Heras, J. A., Chaty, S., \& Rodriguez, J. 2007, Astr. Tel., 1035
1867: 
1868: \end{thebibliography}{}
1869: 
1870: \end{document}
1871: 
1872: 
1873: 
1874: 
1875: 
1876: 
1877: