1: %%% 2nd Revision
2:
3: \documentclass[12pt,preprint]{aastex}
4:
5: % Abbreviations
6: \newcommand{\saxj}{SAX~J1808.4--3658}
7: \newcommand{\saxjb}{SAX~J1748.9--2021}
8: \newcommand{\igrj}{IGR~J00291+5934}
9: \newcommand{\xteja}{XTE~J0929--314}
10: \newcommand{\xtejb}{XTE~J1807--294}
11: \newcommand{\xtejc}{XTE~J1814--338}
12: \newcommand{\xtejd}{XTE~J1751--305}
13: \newcommand{\nudot}{\dot{\nu}}
14: \newcommand{\Mdot}{\dot{M}}
15:
16: % Right size parentheses
17: \newcommand{\pth}[1]{\left({#1}\right)}
18: \newcommand{\qpth}[1]{\left[{#1}\right]}
19: \newcommand{\cpth}[1]{\left\{{#1}\right\}}
20:
21: \newcommand{\Sin}[1]{\sin\pth{#1}}
22: \newcommand{\Cos}[1]{\cos\pth{#1}}
23:
24: \newcommand{\Tstar}{T_{\star}}
25: \newcommand{\Porb}{P_{orb}}
26:
27: \usepackage{subfigure, ulem}
28: \bibliographystyle{apj}
29: \bibpunct{(}{)}{;}{a}{}{,}
30:
31: \shorttitle{Timing of XTE~J1807--294}
32: \shortauthors{Riggio et al.}
33:
34: \begin{document}
35:
36: \title{Spin up and phase fluctuations in the timing of the accreting
37: millisecond pulsar XTE~J1807--294}
38:
39: \author{A. {Riggio}\altaffilmark{1}, T. {Di Salvo}\altaffilmark{2}, L.
40: {Burderi}\altaffilmark{1}, M. T. {Menna}\altaffilmark{3},
41: A. {Papitto}\altaffilmark{3}, R. {Iaria}\altaffilmark{2},
42: G. {Lavagetto}\altaffilmark{2}}
43:
44: \altaffiltext{1}{Dipartimento di Fisica, Universit\`a degli Studi di
45: Cagliari, Cittadella Universitaria, S.P. Monserrato - Sestu Km
46: 0,700 09042, Monserrato (CA), Italy; E-mail: riggio@dsf.unica.it}
47: \altaffiltext{2}{Dipartimento di Scienze Fisiche e Astronomiche,
48: Universit\`a~di Palermo, Via Archirafi 36, Palermo I-90123, Italy}
49: \altaffiltext{3}{Osservatorio Astronomico di Roma, Sede di
50: Monteporzio Catone, Via Frascati 33, Rome I-00040, Italy}
51:
52:
53: \begin{abstract}
54: We performed a timing analysis of the 2003 outburst of the accreting
55: X-ray millisecond pulsar \xtejb\ observed by RXTE. Using recently
56: refined orbital parameters we report for the first time a precise
57: estimate of the spin frequency and of the spin frequency
58: derivative. The phase delays of the pulse profile show a strong
59: erratic behavior superposed to what appears as a global spin-up
60: trend. The erratic behavior of the pulse phases is strongly related
61: to rapid variations of the light curve, making it very difficult to
62: fit these phase delays with a simple law. As in previous cases, we
63: have therefore analyzed separately the phase delays of the first
64: harmonic and of the second harmonic of the spin frequency, finding
65: that the phases of the second harmonic are far less affected by the
66: erratic behavior. In the hypothesis that the second harmonic pulse
67: phase delays are a good tracer of the spin frequency evolution we
68: give for the first time a estimation of the spin frequency
69: derivative in this source. The source shows a clear spin-up of $\dot
70: \nu = 2.5(7) \times 10^{-14}$ Hz sec$^{-1}$ (1 $\sigma$ confidence
71: level). The largest source of uncertainty in the value of the
72: spin-up rate is given by the uncertainties on the source position in
73: the sky. We discuss this systematics on the spin frequency and its
74: derivative.
75: \end{abstract}
76: \keywords{stars: neutron -- stars: magnetic fields -- pulsars: general --
77: pulsars: individual: \xtejb~ -- X-ray: binaries.}
78:
79: \maketitle
80:
81: \section{Introduction}
82: Binary systems in which one of the two stars is a neutron star (NS
83: hereafter) are among the most powerful X-ray sources of our
84: Galaxy. The emission of X-rays is due to the matter transferred from
85: the companion star and accreted onto the NS, and to the release of the
86: immense gravitational energy during the fall or in the impact with the
87: NS surface. A sub-category of such systems is called Low Mass X-ray
88: Binaries (LMXB). LMXBs are characterized by low NS superficial
89: magnetic fields ($< 10^9~ \mathrm{Gauss}$) and by the low-mass ($< 1
90: \, M_\odot$) of the companion star. The so-called recycling scenario
91: \citep[see for a review][]{Bhatta_91} sees in the millisecond radio
92: pulsars the last evolutionary step of LMXBs, where the torques due to
93: the accretion of matter and angular momentum, together with the
94: relatively weak magnetic fields, are able to spin-up the NS to periods
95: of the order of one millisecond. When the accretion phase terminates
96: and the companion star stops transferring matter, the NS can switch on
97: as a millisecond radio pulsar, although no example has been reported
98: yet.
99:
100: The recycling scenario received the long awaited confirmation only in
101: 1998 with the discovery of the first millisecond X-ray pulsar in a
102: transient LMXB; the first LMXB observed to show coherent pulsations at
103: a frequency of $\sim 400$ Hz was \saxj\ \citep{Wijnands_98}, in which
104: the NS is orbiting its companion star with a period of $\sim 2.5$ hr
105: \citep{Chakra_98}. Why millisecond X-ray pulsars were so elusive is
106: an argument still debated in literature. A possible reason can be due
107: to the relatively low magnetic fields of these sources which has
108: therefore less capability to channel the accreting matter onto the
109: polar caps, then the chance to see a pulsed emission from a LMXB is
110: quite low \citep[see e.g.][]{Vaughan_94}, especially at high accretion
111: rates. However, to date 10 LMXBs have been discovered to be accreting
112: millisecond pulsars (see \citealp{Wijnands_05}~for a review on the
113: first 6 discovered, for the last four see \citealp{Kaaret_06,
114: Krimm_07, Casella_07, Altamirano_07}), and all of them are in
115: transient systems. They spend most of the time in a quiescent state,
116: with very low luminosities (of the order of $10^{31} - 10^{32}$
117: ergs/s) and rarely they go into an X-ray outburst with luminosities in
118: the range $10^{36} - 10^{37}$ ergs/s. Although the recycling scenario
119: seems to be confirmed by these discoveries, from timing analysis of
120: accreting millisecond pulsars we now know that some of these sources
121: show spin-down while accreting~ \citep{Galloway_02, Papitto_07}. This
122: means that it is of fundamental importance to study the far from being
123: understood details of the mechanisms regulating the exchange of
124: angular momentum between the NS and the accreting matter, and chiefly
125: the role of the magnetic field in this exchange. The main way to do
126: this is the study of the pulse phase shifts and their relations with
127: other physical observable parameters of the NS.
128:
129: The pulse phase shifts are frequently affected by intrinsic long-term
130: variations and/or fluctuations (with which we mean an erratic behavior
131: of the phase delays possibly caused by variations in the instantaneous
132: accretion torques or movements of the accretion footprints on the NS
133: surface, see \citealp{DiSalvo_07}\ for a review). Examples of this
134: complex behavior of the pulse phase shifts in accreting millisecond
135: pulsars were already reported in literature. \cite{Burderi_06}, who
136: analyzed the 2002 outburst of the accreting millisecond pulsar \saxj\
137: found a jump of 0.2 in the pulse phases of the first harmonic which is
138: not present in the second harmonic phases, which show a much more
139: regular behavior. This change is in correspondence of a change in the
140: slope in the exponential decay of the X-ray light curve (see also
141: \cite{Hartman_07} for a discussion of the complex phase behavior in
142: other outbursts of \saxj). \cite{Papitto_07} found that the second
143: harmonic of XTE~J1814--338 follows the first harmonic giving
144: approximately the same spin frequency derivative. A clear model which
145: can explain this behavior is still missing, but these observational
146: evidences seem to suggest that perhaps the second harmonic has a more
147: fundamental physical meaning. For instance it may be related to the
148: emission of both the polar caps while the first harmonic may be
149: dominated by the most intense but less stable polar cap. Another
150: possible explanation comes from possible shape and/or size variations
151: of the accretion footprints related to variations of the accretion
152: rate. \cite{Romanova_03} found a such behavior in their numerical
153: simulations.
154:
155: In this paper we report the results of a timing analysis performed on
156: \xtejb, making use of an improved orbital solution \citep{Riggio_07}.
157: As in the cases mentioned above, \xtejb\ shows erratic fluctuations of
158: the phase delays of the first harmonic and a much more regular
159: behavior of the phase delays derived from the second harmonic. In the
160: hypothesis that the second harmonic pulse phase delays are a good
161: tracer of the spin frequency evolution we can derive a spin-up rate of
162: $2.5(7) \times 10^{-14}$ Hz/s (1 $\sigma$ confidence level).
163:
164: \section{Observation and Data Analysis}
165: The millisecond X-ray pulsar \xtejb~was discovered by RXTE on February
166: 21$^{st}$, 2003 \citep{Markwardt_atel_03a}. The source was observed
167: with the Proportional Counter Array (PCA, $2 - 60$ keV energy range)
168: and the High Energy X-ray Timing Experiment (HEXTE, $20 - 200$ keV)
169: on-board of the RXTE satellite \citep{Jahoda_96}, during a long X-ray
170: outburst which lasted from February 28 to June 22, 2003. \xtejb~was
171: also observed with other satellites such as XMM-Newton
172: \citep{Campana_03, Kirsch_04, Falanga_05}, Chandra
173: \citep{Markwardt_iauc_03b} and INTEGRAL \citep{Campana_03}. No optical
174: or radio counterpart have been reported to date. \cite{Linares_05}
175: have reported the presence of twin kHz QPOs analyzing RXTE
176: observations.
177:
178: Here we analyze all the archival RXTE observations of this source. In
179: particular, we use high-time resolution data from the PCA. We use
180: data collected in GoodXenon packing mode for the timing analysis,
181: which permits maximum time and energy resolution (respectively $1 \mu
182: s$ and 256 energy channels). In order to improve the signal to noise
183: ratio we select photon events from PCUs top layer and in the energy
184: range 3-13 keV \citep{Galloway_02}. Indeed, we have verified that this
185: is the range where we have the highest S/N ratio. In fact, using all
186: the energy range the pulsations at days 104 and 106, respectively,
187: after the beginning of the outburst are much less statistically
188: significant.
189:
190: Using the \textit{faxbary}\footnote{faxbary is a tool of the HEAsoft
191: Software Packages. It can be found at:
192: http://heasarc.nasa.gov/docs/software/lheasoft/} tool (DE-405 solar
193: system ephemeris) we corrected the photon arrival times for the motion
194: of the earth-spacecraft system and reported them to barycentric
195: dynamical times at the Solar System barycenter. We use the source
196: position reported by \cite{Markwardt_iauc_03b} using a
197: \textit{Chandra} observation during the same outburst.
198:
199: To obtain the X-ray light curve during the outburst we used the PCA
200: data collected in Standard2 mode (256 energy channel and 16s binned
201: data) and corrected for the background using the faint background
202: model suitable for the source count rate \citep[see][]{Jahoda_06}. No
203: energy selection was applied in this case since we are interested in a
204: good tracer of the bolometric luminosity. We did not apply any
205: correction for dead time since the maximum count rate was quite low
206: ($< 100$ cts sec$^{-1}$ PCU$^{-1}$, background included); in fact the
207: mean time between two event is at least two orders of magnitude higher
208: than the expected dead time ($10 \mu s$) for this count rate
209: \citep{Jahoda_06}. We selected all the data using both internal GTI
210: and applying criteria regards pointing offset, South Atlantic Anomaly
211: passage, electronic contamination and Sun offset\footnote{according to
212: the prescription given in
213: http://heasarc.gsfc.nasa.gov/docs/xte/abc/screening.html we adopted as
214: selection criteria the following: time since SAA greater than 30
215: minutes, elevation angle with respect the Earth greater than 10
216: degree, electron contamination lower than 0.1, and pointing offset
217: lower than 0.25 degree.}.
218:
219: The resulting light curve is shown in Fig.\ref{fig1} (pentagon
220: symbols). The flux shows an exponential decay with superimposed six
221: evident flares. To derive the characteristic time of the exponential
222: decay we fitted the light curve with an exponential law. In order to
223: remove the time intervals affected by X-ray flares we excluded from
224: the fit all the points whose flux was greater than the best fit
225: exponential model by at least a 15\%. The choice of this threshold is
226: arbitrary, but a different choice, like 10\% or 20\% include or
227: exclude very few points. We repeated the exponential fit on the flares
228: subtracted light curve. In this last fit the $\chi^2/\mathrm{d.o.f.} =
229: 23096/214$ which is awfully high. Such a large $\chi^2$ is due to
230: deviations of the X-ray light curve from a pure exponential decay (see
231: e.g.\ all the points after 100 days from the beginning of the
232: outburst).
233:
234: Although these deviations may be very small, they can be large if
235: compared with the statistical error on a single point. However, in
236: order to obtain a reliable estimate of the parameters of the fit, and
237: in particular a reliable estimate of the errors, we need to obtain a
238: reduced $\chi^2$ of the order of unity. Therefore, we have multiplied
239: the errors on each point by a factor 10. In this way we obtain a
240: characteristic decay time of $\tau = 17.50(25)$ days.
241:
242: It should be noted that a constant term must be added to the model to
243: obtain a good description of the light curve, although background
244: subtraction was performed. This residual results to be $\sim 10.8(2)$
245: cts sec$^{-1}$ PCU$^{-1}$ and may be due to a contaminating source in
246: the PCA field of view. It is unlikely it is due to quiescence emission
247: since the source was observed in quiescence by XMM Newton and was not
248: detected \citep{Campana_05}. In both cases this residual flux does not
249: affect the inferred decaying time of the light curve or any other
250: results of this paper.
251:
252: In order to minimize the time delays induced by the orbital motion, we
253: correct the photon arrival times with the formula \citep[see
254: e.g.]{Deeter_81}:
255: \begin{equation}
256: \label{eq:time_orb_doppler}
257: t_{em} \simeq t_{arr} - A \qpth{\Sin{m(t_{arr}) + \omega} +
258: \frac{\varepsilon}{2} \Sin{2m(t_{arr}) + \omega} -
259: \frac{3\varepsilon}{2}\Sin{\omega}},
260: \end{equation}
261: where $t_{em}$ is the photon emission time, $t_{arr}$ is the photon
262: arrival time, $A$ the projected semi-major axis in light seconds,
263: $m(t_{arr}) = 2\pi (t_{arr} - \Tstar)/\Porb$ is the mean anomaly,
264: $\Porb$ the orbital period, $\Tstar$ is the time of ascending node
265: passage, $\omega$ is the periastron angle and $\varepsilon$ the
266: eccentricity. In order to remove completely from the pulse phase
267: delays any effect due to the orbital motion it is of fundamental
268: importance to correct the arrival times of the events with very
269: precise orbital parameters. To accomplish this task we used the
270: orbital solution recently published by \cite{Riggio_07}, who, using
271: the total outburst time available (about 120 days), obtain a solution
272: that is about two orders of magnitude more precise than previously
273: reported orbital solutions.
274:
275: We divided the whole observation in time intervals of length
276: approximately equal to the orbital period\footnote{This is to minimize
277: possible residuals due to uncertainties in the orbital parameters,
278: since we expect these residuals to be periodic at the orbital period
279: of the system.} and epoch-folded each of these data intervals with
280: respect to the spin period we reported in Tab.~\ref{table1}. In this
281: way we are able to significantly detect the X-ray pulsations up to day
282: 106 from the beginning of the outburst, making this as the longest
283: time span in which timing analysis of an accreting millisecond pulsar
284: has been performed.
285:
286: The pulse phase delays are obtained fitting each pulse profile with
287: two sinusoidal components (with period fixed to 1 and 0.5 of the spin
288: period, respectively), since the second harmonic was significantly
289: detected in the folded light curve. In Fig.~\ref{fig1} and~\ref{fig2}
290: we show the pulse phase delays of the first harmonic and second
291: harmonic, respectively. We have plotted only the pulse phase delays
292: corresponding to the folded light curves for which the statistical
293: significance for the presence of the X-ray pulsations is $>
294: 3\sigma$. Moreover, we consider the second harmonic significantly
295: detected (and plotted its phases) only when the ratio between the best
296: fit amplitude of the second sinusoid and its error was greater than 3
297: ($A/\delta A > 3$). We propagated on each phase point the errors on
298: the orbital parameters with the formulas derived by
299: \citealp{Burderi_07}. We note that the propagated errors in this case,
300: for which the orbital parameters are known with great precision,
301: result to be much smaller than the statistical errors derived from the
302: sinusoidal fit.
303:
304: As it is evident from Fig.~\ref{fig1} and~\ref{fig2}, the phase delays
305: of the first harmonic show a noisy behavior with shifts up to 0.3 in
306: phase. The noise affecting the phases results strongly
307: anti-correlated to the source flux, as already noted for another
308: source of this class \citep{Papitto_07}. On the other hand, the phase
309: delays derived from the second harmonic are much more regular, a
310: behavior that is similar to the one shown by \saxj\
311: \citep{Burderi_06}. Although a few points (corresponding to rapid
312: flares in the light curve) appear to be significantly below the
313: general trend, the phase delays of the second harmonic clearly show a
314: parabolic decrease, as it is expected in case of a spin-up of the NS.
315:
316: \section{Timing Results}
317: Since the phase delays of the second harmonic are much less noisy than
318: the phases derived from the first harmonic, and assuming that the
319: pulse phase delays derived from second harmonic are a good tracer of
320: the spin frequency evolution, we decided to fit the second harmonic in
321: order to find information on the spin frequency behavior. To fit the
322: phase delays we start from the simplest assumption of a constant spin
323: frequency derivative. We hence fit the second harmonic phase delays
324: with the model:
325: \begin{equation}
326: \phi(t) = \phi_0 - \Delta\nu\, (t - T_0) - \frac{\nudot}{2}\, (t-T_0)^2,
327: \label{eq:model}
328: \end{equation}
329: where $T_0$ is the date of the beginning of the observation,
330: $\Delta\nu$ is a correction to the spin frequency and $\nudot$ the
331: spin frequency derivative. Using all the data points we obtained a
332: spin frequency derivative $\nudot = 2.05(28) \times 10^{-14}$ Hz/s
333: with a quite large $\chi^2/\mathrm{d.o.f.} = 1560.57/142$. From a
334: visual inspection of the phase residuals with respect to this model
335: (see Fig.~\ref{fig2}), we can see that the largest contribution to the
336: $\chi^2$ is given by a group of 3 points at MJD 52713.0 (about 14.5
337: days from the beginning of the outburst). These points (indicated with
338: triangles in Fig.~\ref{fig2}) correspond to the largest flare visible
339: in the light curve and to a strong decrease of the phases of the first
340: harmonic as well (cf.\ Fig.~\ref{fig1}). We therefore believe that
341: this is a phase shift induced by the rapid change of the X-ray flux
342: similar to the phase shifts observed in the first harmonic. If we
343: remove from our data set these three points (case B) and re-perform
344: the fit, we obtain a frequency derivative $\nudot = 2.26(15) \times
345: 10^{-14}$ Hz/s, perfectly compatible with the value previously found,
346: demonstrating that the three points we have eliminated do not affect
347: the spin frequency derivative obtained by the fit. In this case of
348: course the statistical quality of the fit increases, giving a
349: $\chi^2/\mathrm{d.o.f.} = 452.4/139$. However, this $\chi^2$ is still
350: unacceptable; again the post-fit residuals indicate that the major
351: contribution to the $\chi^2$ is given by all the points in
352: correspondence of the X-ray flares. We have therefore decided to
353: remove all the points (indicated with circles in Fig.~\ref{fig2}) that
354: fall in time intervals for which the flux results to be larger by 15\%
355: with respect to the exponential best fit function derived above. In
356: this way total of 21 points were excluded from the fit (case C). With
357: this last data set we obtained a value of spin frequency derivative
358: $\nudot = 2.46(15) \times 10^{-14}$ Hz/s (again compatible with the
359: results obtained with the complete data set) and a
360: $\chi^2/\mathrm{d.o.f.} = 257.6/121$. In this case, a value of $\nu =
361: 190.623507018(6)$ Hz for spin frequency at the beginning of the
362: outburst was obtained.
363:
364: We have also tried to fit this (reduced) data set with a spin-up model
365: which takes into account the decrease of the X-ray flux (supposed to
366: trace the mass accretion rate) during the outburst (see
367: \citealp{Burderi_06} for a more detailed discussion). In principle
368: this correction should be important for this source, given the
369: particularly long duration of the outburst (about 120 days). Fitting
370: the phase delays of the second harmonic with eq.~1 of
371: \cite{Burderi_06}, in which we adopted an exponential decay time of
372: the X-ray flux of 17.50(25) days, as derived from the X-ray light
373: curve, we obtain a significant improvement of the fit with a
374: $\chi^2/\mathrm{d.o.f.} = 225.5/121$ (a $\Delta \chi^2 = 32$ for the
375: same number of degrees of freedom). In this case, we obtain a spin
376: frequency derivative at the beginning of the outburst of $\dot \nu_0 =
377: 1.25(7)\times 10^{-13}$ Hz/s, corresponding to a mass accretion rate
378: at the beginning of the outburst of $\dot M_0 = 4.03(23) \times
379: 10^{-10}$ M$_\odot$/yr, and a best fit spin frequency of $\nu_0 =
380: 190.623506939(7)$ Hz. In Fig.\ref{fig3} we report, among the last
381: reduced data set, both parabolic and exponential best fit models and
382: the residuals of the exponential model (bottom panel).
383:
384: Unfortunately, the results above are affected by large systematic
385: uncertainties, given by the large uncertainty on the source
386: coordinates (which is about $0''.4 (1\sigma \; \textrm{confidence
387: level})$ from a Chandra observation\footnote{The Chandra observation
388: of \xtejb\ in outburst was performed with the instrument HRC-S. As
389: reported in http://asc.harvard.edu/cal/ASPECT/celmon/, the confidence
390: levels are given at 68\% ($0''.4$), 90\% ($0''.6$) and 99\%
391: ($0''.8$).}), that we are now going to discuss in detail.
392:
393: The uncertainties in the phase delays caused by the uncertainties on
394: the estimates of the source position in the sky, will produce a
395: sinusoidal oscillation at the Earth orbital period. For observation
396: times shorter than one year, as it is the case for most transient
397: accreting millisecond pulsars, this can cause systematic errors on the
398: determination of the NS spin period and its derivative, since a series
399: expansion of a sinusoid contains a linear and a quadratic term. In the
400: case of \xtejb, due to the low positional precision
401: \citep{Markwardt_atel_03b} and the long time span in which the
402: pulsation is visible (up to 106 days from the beginning of the
403: outburst), we obtain, from the expression given by \cite{Burderi_07},
404: the following systematic uncertainties in the spin frequency and the
405: spin frequency derivative, respectively: $\sigma_{\nu\, {\rm pos}}
406: \sim 4.1 \times 10^{-8}$ Hz and $\sigma_{\dot \nu\, {\rm pos}} \sim
407: 0.8 \times 10^{-14}$ Hz/s. Since this error is of the same order of
408: magnitude of our best fit estimate of $\nudot$, we need to evaluate
409: these effects in a more careful manner.
410:
411: Let us consider the expression of the phase delays induced by the
412: Earth motion for a small displacement, $\delta \lambda$ and $\delta
413: \beta$, in the position of the source in ecliptic coordinates,
414: $\lambda$ and $\beta$ \citep[see e.g.][]{Lyne_90}:
415: \begin{equation}
416: \label{eq:posdel}
417: \Delta \phi_{\rm pos}(t) = \nu_0\, y \,\left[\,\sin (M_0 + \epsilon)
418: \cos \beta\, \delta \lambda - \cos (M_0 + \epsilon) \sin \beta
419: \,\delta \beta \,\right]
420: \end{equation}
421: where $y = r_{\rm E}/c$ is the distance of the Earth with respect to
422: the Solar system barycenter in light seconds, and $M_0 = 2 \pi (T_0 -
423: T_\gamma)/P_\oplus - \lambda$, where $T_0$ is the begin of the
424: observation, $P_\oplus$ the Earth orbital period, $T_\gamma$ the time
425: of passage through the Vernal point, and $\epsilon = 2 \pi (t -
426: T_0)/P_\oplus$. As already done by \cite{Burderi_07},
427: Eq.~\ref{eq:posdel} can be rewritten as:
428: \begin{equation}
429: \label{eq:posdel2}
430: \Delta \phi_{\rm pos} = \nu_0\, y \,\sigma_\gamma \, u \,\Sin{M_0 +
431: \epsilon - \theta^*}
432: \end{equation}
433: where $\sigma_\gamma$ is the positional error circle, $\theta^* =
434: \arctan (\tan \beta \; \delta \beta / \delta \lambda)$, and $u =
435: [(\cos \beta\; \delta \lambda)^2 + (\sin \beta\; \delta \beta)^2\;
436: ]^{1/2} / \sigma_\gamma$. We can safely pose $u = 1$ as an upper
437: limit.
438:
439: In order to take into account the effects of an incorrect source
440: position, we fitted the reduced data set (case C) with a model which
441: takes also into account the modulation caused by the incorrect source
442: coordinates and given by Eq.~\ref{eq:posdel2}:
443: \begin{equation}
444: \phi(t) = \phi_0 - \Delta\nu\, (t - T_0) - \frac{\nudot}{2}\,
445: (t - T_0)^2 + \Delta \phi_{\rm pos}(t)
446: \label{eq:model2}
447: \end{equation}
448: We have repeated the fit changing $\sigma_\gamma$ and $\theta^*$ in
449: such a manner to cover the Chandra error box up to a 90 \% confidence
450: level, that is sky region within an angular distance of $0.''6$ from
451: the reported source position. The obtained values of the spin
452: frequency and its derivative for each possible position of the source
453: within the Chandra error box are shown in Fig.~\ref{fig4}. The values
454: of $\dot \nu$, at $1\sigma$ confidence level, range in the interval
455: $(1.8 - 3.2) \times 10^{-14}$ Hz s$^{-1}$, while the best fit value of
456: the frequency derivative for the source nominal position is $2.46(15)
457: \times 10^{-14}$ Hz s$^{-1}$. It is evident that the effect of the
458: poor source position knowledge is much larger than the statistical
459: error on the parabolic fit. Still the spin-up behavior of the source
460: remains significant even considering the large uncertainties caused by
461: the position uncertainties.
462:
463: A similar discussion must be done for the spin frequency. The best fit
464: value for the nominal position is $\nu =190.623507018(4)$ Hz, while
465: the variation of the linear term in the fit at different positions of
466: the source inside the Chandra error box are in the range $\Delta\nu =
467: \pm 4 \times 10^{-8}$ Hz, one order of magnitude greater than the
468: single fit statistical error. Finally, the reduced $\chi^2$ for these
469: fits varies in the range (2.1 - 2.4).
470:
471: Summarizing, using the pulse phase delays derived from the second
472: harmonic, we inferred the spin frequency derivative in \xtejb. In the
473: hypothesis of constant spin frequency derivative we obtain a value of
474: $\nudot = 2.46(15) \times 10^{-14}$ Hz s$^{-1}$. In the hypothesis of
475: an exponential decay of the accretion rate we obtained a value of the
476: spin frequency derivative at the beginning of the outburst of $\dot
477: \nu_0 = 1.25(7)\times 10^{-13}$ Hz s$^{-1}$. These results do not
478: include the systematic errors induced by the poorly constrained source
479: position. Taking into account the errors on the source position we
480: obtained, for the constant and exponential decay models, respectively,
481: the values of $2.5(7) \times 10^{-14}$ Hz s$^{-1}$ and $ 1.25(33)
482: \times 10^{-13}$ Hz s$^{-1}$.
483:
484: \section{Discussion and conclusion}
485: We have analyzed a long RXTE observation of the accreting millisecond
486: pulsar \xtejb\ and reported the results of an accurate timing analysis
487: performed on a time span of about $120$ days, the longest outburst of
488: an accreting millisecond pulsar for which a timing analysis has been
489: performed to date. We find that the phase delays derived from the
490: first harmonic show an erratic behavior around a global parabolic
491: spin-up trend. This behavior is similar to that previously shown by
492: two accreting millisecond pulsar, \saxj\ \citep{Burderi_06} and
493: \xtejc\ \citep{Papitto_07}. In the case of the 2002 outburst of \saxj,
494: the phase delays of the first harmonic show a shift by about 0.2 in
495: phase at day 14 from the beginning of the outburst, when the X-ray
496: flux abruptly changed the slope of the exponential decay. On the
497: other hand, the phase delays of the second harmonic in \saxj\ showed
498: no sign of the phase shift of the first harmonic, and could be fitted
499: by a spin-up in the first part of the outburst plus a barely
500: significant spin-down at the end of the outburst. In the case of
501: \xtejc, the fluctuations in the phase delays were visible both in the
502: first harmonic and in the second harmonic, superposed to a global
503: parabolic spin-down trend. \cite{Papitto_07} have shown that the
504: post-fit phase residuals were strongly anti-correlated to variations
505: of the X-ray light curve. These fluctuations were interpreted as due
506: to movements of the accretion footprints (or accretion column) induced
507: by variations of the X-ray flux.
508:
509: In the case of \xtejb, the fluctuations in the phase delays affect
510: mostly the first harmonic, which shows a trend that is very difficult
511: to reproduce with a simple model. As in the case of \xtejc, the
512: post-fit phase residuals are clearly anti-correlated with variations
513: observed in the X-ray light curve; from Figure~\ref{fig1} we see that
514: the phases decrease when the X-ray flux shows rapid increases. It is
515: important to note that the anti-correlation visible between the
516: post-fit phase delays and the X-ray flux is independent of the
517: spin-down or spin-up behavior of the source, since it is observed in
518: \xtejc, which shows spin-down, and in \xtejb, which shows spin-up.
519: The correlation between the phase delays and the X-ray flux affects
520: the second harmonic only marginally. Indeed, there are a few points
521: in the phase delays of the second harmonic that are significantly
522: below the global trend observed in the phase delays, and all of them
523: correspond to flares in the X-ray light curve. Excluding these points
524: marginally affects the values we obtain for the spin frequency and its
525: derivative, but gives a significant improvement of the $\chi^2$ of the
526: fit.
527:
528: We find that the phase delays of the harmonic are fitted by a
529: parabolic spin-up model. We have also showed that the quality of the
530: fit is much improved if we use a more physical model in which the
531: spin-up rate decreases exponentially with time following the decrease
532: of the X-ray flux (and hence of the inferred mass accretion rate). In
533: fact, if the spin-up of the source is related to the mass accretion
534: rate, then it should not be constant with time, but, in first
535: approximation, should decrease proportionally with the mass accretion
536: rate onto the NS. For instance, assuming that accretion of matter and
537: angular momentum occurs at the corotation radius $R_{co}$, the
538: relation between the spin frequency derivative and the mass accretion
539: rate is, from the angular momentum conservation law, $\nudot = \Mdot
540: \, \sqrt{G M R_{co}}/ 2 \pi I$, where G is the gravitational constant,
541: M the NS mass and I is the NS moment of inertia; this gives a lower
542: limit on the mass accretion rate since the specific angular momentum
543: at the corotation radius is the maximum that can be transferred to the
544: NS. In the case of \xtejb, the duration of the outburst is
545: particularly long (about 120 days), and the effect of the global
546: decrease of the mass accretion rate during the outburst should be
547: particularly relevant for this source. Indeed in this case the fit we
548: obtain using an exponentially decreasing spin-up rate is significantly
549: better than using a constant spin-up rate.
550:
551: From the fit of the phase delays of the second harmonic of \xtejb\
552: with the model discussed above we find a mass accretion rate at the
553: beginning of the outburst of $4(1) \times 10^{-10}$ M$_\odot$
554: yr$^{-1}$.\footnote{For this estimation we adopted the value of
555: $I=10^{45}$ g cm$^2$, M = 1.4 M$_\odot$ and NS radius $R_{NS} = 10^6$
556: cm.} We can compare this mass accretion rate with the X-ray flux of
557: the source at the beginning of the outburst that was $2 \times
558: 10^{-9}$ ergs cm$^{-2}$ s$^{-1}$ \citep{Falanga_05} from which we
559: derive an X-ray luminosity of $4.7 \times 10^{36}$ ergs s$^{-1}$ and a
560: distance to the source of 4.4(6) kpc. Clearly this is only a crude
561: estimation of the distance on the basis of our timing results and
562: future independent estimates are needed to confirm or disprove our
563: hypothesis.
564:
565: \acknowledgements This work was supported by the Ministero della
566: Istruzione, della Universit\`a e della Ricerca (MIUR), national
567: program PRIN2005 2005024090$\_$004.
568:
569: % Stile bibliografico e bibliografia
570: \bibliography{ms}
571:
572: \begin{deluxetable}{lc}
573: \tablecaption{Orbital and Spin Parameters for XTE J1807-294.}
574: \label{table1}
575: \tablehead{ \colhead{Parameter} & \colhead{Value}}
576: \startdata
577: RA (J2000) & \(18^h\, 06^m\, 59.8^s \)\tablenotemark{a} \\
578: Dec (J2000) & \(-29^o\, 24'\, 30"\)\tablenotemark{a} \\
579: Orbital period, $P_{orb}$ (s) & 2404.41665(40)\tablenotemark{b} \\
580: Projected semi-major axis, $a_x \sin i$ (lt-ms) & 4.819(4)\tablenotemark{b} \\
581: Ascending node passage, T$^\star$~\tablenotemark{a} (MJD) & 52720.675603(6)\tablenotemark{b} \\
582: Eccentricity, e & $<$ 0.0036\tablenotemark{b} \\
583: Reference epoch, T$_0$~\tablenotemark{c} (MJD) & 52698.5 \\
584: \textit{Parabolic fit results} & \\
585: Spin frequency, $\nu_0$ (Hz) & 190.62350702(4) \\
586: Spin frequency derivative, $\nudot$ (Hz s$^{-1}$) & $ 2.5(7) \times 10^{-14}$ \\
587: \textit{Exponential fit results} & \\
588: Spin frequency, $\nu_0$ (Hz) & 190.62350694(5) \\
589: Spin frequency derivative, $\nudot$ (Hz s$^{-1}$) & $ 1.25(33) \times 10^{-13}$
590: \enddata
591:
592: \tablecomments{Errors on orbital parameters are intended to be at
593: $1\sigma$ confidence level (c. l.), upper limits are
594: given at 95\% c.l. Best fit spin parameters are
595: derived in both hypothesis of a constant spin-up and
596: flux dependent spin-up, and the uncertainties include
597: systematics due to the uncertainties in the source
598: position (see text).}
599:
600: \tablenotetext{a}{\cite{Markwardt_iauc_03b}.}
601: \tablenotetext{b}{\cite{Riggio_07}.}
602: \tablenotetext{c}{This is the Epoch at which are referred the
603: reported values of $\nu$ and $\nudot$.}
604: \end{deluxetable}
605:
606:
607: \begin{figure}
608: \plotone{f1.eps}
609: \caption{Light curve of \xtejb\ during the 2003 outburst (pentagon)
610: and phase delays of the first harmonic as a function of time
611: (small dot). The dashed vertical lines indicate the times of six
612: clearly visible flares of the X-ray flux superimposed to a global
613: exponential decay. The dotted curve represent the exponential fit
614: of the light curve, obtained after having previously excluded from
615: the data the six flares. The dashed curve represent the parabolic
616: best fit obtained fitting the second harmonic phase delays and
617: considering the nominal source position. Strong fluctuations of
618: the phase delays are apparent and are strongly anti-correlated to
619: the flares present in the X-ray light curve.\label{fig1}}
620: \end{figure}
621:
622:
623: \begin{figure}
624: \plotone{f2.eps}
625: \caption{ Plot of \xtejb\ second harmonic pulse phase delays. The
626: four curves represent the parabolic best fit for the nominal
627: source position, respectively, using all the data points (case A),
628: excluding the three point at MJD 52713.0 (case B, where the points
629: excluded are identified by triangles), and excluding all the data
630: points for which the flux exceeds the best-fit exponential decay
631: for more than 15\% (Case C, where the points excluded are
632: identified by circles), and the best fit obtained using an
633: exponentially decreasing mass accretion rate (see text). The
634: exponential fit was performed on the data sub-set corresponding to
635: case C.\label{fig2}}
636: \end{figure}
637:
638: \begin{figure}
639: \plotone{f3.eps}
640: \caption{second harmonic pulse phase delays together with the
641: parabolic and exponential best fit (top panel), and residuals in
642: units of $\sigma$ with respect to the exponential best fit model
643: (bottom panel) considering only the sub-set of case C (see
644: Fig.\ref{fig2}).\label{fig3}}
645: \end{figure}
646:
647: \begin{figure}
648: % \epsscale{.80}
649: \plottwo{f4a.eps}{f4b.eps}
650: \caption{Diagrams of the best fit values of $\Delta \nu$ (left
651: panel) and $\Mdot$ (right panel) obtained fitting the first
652: harmonic pulse phase delays with the expression \ref{eq:model2},
653: as function of the parameters $\sigma_\gamma$ and $\theta^*$ (see
654: text).\label{fig4}}
655: \end{figure}
656:
657: \end{document}
658:
659:
660:
661: text).\label{fig4}}
662: \end{figure}
663:
664: \end{document}
665:
666:
667:
668: