1: % \documentclass[12pt,preprint]{aastex}
2: \documentclass{emulateapj}
3: \usepackage{natbib}
4:
5: \shorttitle{A Massive Neutron Star in M5}
6: \shortauthors{Freire et al.}
7:
8: %\received{2004 February 21}
9: \begin{document}
10:
11: \title{A Massive Neutron Star in the Globular Cluster M5}
12:
13: \author{Paulo C. C. Freire\altaffilmark{1},
14: Alex Wolszczan\altaffilmark{2},
15: Maureen van den Berg\altaffilmark{3},
16: Jason W. T. Hessels\altaffilmark{4}}
17: \altaffiltext{1}{N.A.I.C., Arecibo Observatory, HC 03 Box 53995, PR 00612,
18: U.S.A.; {\tt pfreire@naic.edu}}
19: \altaffiltext{2}{Department of Astronomy and
20: Astrophysics, Penn State University, University Park, PA 16802,
21: U.S.A.; {\tt alex@astro.psu.edu}}
22: \altaffiltext{3}{Harvard-Smithsonian Center for Astrophysics, 60
23: Garden Street, Cambridge, MA 02138, U.S.A., {\tt
24: maureen@head.cfa.harvard.edu}}
25: \altaffiltext{4}{Astronomical Institute ``Anton Pannekoek'',
26: University of Amsterdam, Kruislaan 403, 1098 SJ Amsterdam, The
27: Netherlands; {\tt jhessels@science.uva.nl}}
28:
29: \begin{abstract}
30: We report the results of 19 years of Arecibo timing for two pulsars in
31: the globular cluster NGC~5904 (M5), PSR~B1516+02A (M5A) and
32: PSR~B1516+02B (M5B). This has resulted in the measurement of the
33: proper motions of these pulsars and, by extension, that of the cluster
34: itself. M5B is a 7.95-ms
35: pulsar in a binary system with a $> 0.13\,M_{\sun}$ companion and
36: an orbital period of 6.86 days. In deep HST images, no optical
37: counterpart is detected within $\sim 2.5\sigma$ of the position of the
38: pulsar, implying that
39: the companion is either a white dwarf or a low-mass main-sequence star. The
40: eccentricity of the orbit ($e = 0.14$) has allowed a measurement of
41: the rate of advance of periastron:
42: $\dot{\omega}\,=\,(0.0142\,\pm\,0.0007)^\circ \rm yr^{-1}$.
43: We argue that it is very likely that this periastron advance is due to
44: the effects of general relativity, the total mass of the binary system
45: then being $(2.29\,\pm\,0.17)\,M_{\sun}$. The small measured mass
46: function implies, in a statistical sense, that a very large fraction
47: of this total mass is contained in the pulsar: $M_p \, = \,
48: (2.08 \pm 0.19)\, M_{\sun}$ (1 $\sigma$); there is a 5\%
49: probability that the mass of this object is $<\,1.72\,M_{\sun}$
50: and a 0.77\% probability that $1.2 \,M_{\sun} \, \leq M_p \, \leq\,
51: 1.44\,M_{\sun}$. Confirmation of the median mass for this neutron star
52: would exclude most ``soft'' equations of state for dense neutron
53: matter. Millisecond pulsars (MSPs) appear to
54: have a much wider mass distribution than is found in double neutron
55: star systems; about half of these objects are significantly
56: more massive than $1.44\,M_{\sun}$. A possible cause is the much
57: longer episode of mass accretion necessary to recycle a MSP, which in
58: some cases corresponds to a much larger mass transfer.
59: \end{abstract}
60:
61: \keywords{binaries: general --- pulsars: general --- pulsars:
62: individual (PSR~B1516+02A) --- pulsars: individual (PSR~B1516+02B)
63: --- neutron stars: general --- equation of state: general
64: }
65:
66: \section{Introduction}\label{sec:intro}
67:
68: Over the past 21 years, more than 130 pulsars have been discovered in
69: globular clusters (GCs)\footnote{For an updated list, see
70: \url{http://www2.naic.edu/$\sim$pfreire/GCpsr.html}.}. Among the
71: first discoveries were PSR~B1516+02A and PSR~B1516+02B
72: \cite{wakp89}. Both are located in the GC NGC~5904. This cluster is
73: also known as M5 and we refer to these pulsars below as M5A and M5B.
74: M5A is an isolated millisecond pulsar (MSP) with a spin period of
75: 5.55 ms. M5B is a 7.95-ms pulsar in a binary system with a
76: low-mass companion (see \S\ref{sec:masses}) and an orbital
77: period of 6.86 days. At the time of its discovery, this was the MSP
78: with the most eccentric orbit known ($e = 0.14$), this being $\sim\,
79: 10^{4}\,-\,10^{5}$ times larger than that of MSP-White Dwarf
80: (WD) systems in the Galactic disk with similar orbital periods.
81:
82: In the Galactic disk, 80\% of all known MSPs are found to be in binary
83: systems and, with the single exception of PSR~J1903+0327 (Champion et
84: al. 2008)\nocite{crl+08}, they are in low-eccentricity orbits with
85: WD companions. In GCs, gravitational interactions with neighboring
86: stars, or even exchange encounters, can produce binary systems with
87: eccentric orbits \cite{rh95}. These high eccentricities allow
88: the measurement of post-Keplerian parameters such as the rate of advance
89: of periastron ($\dot{\omega}$) or, in the future, the Einstein delay
90: ($\gamma$) that are not normally measurable in MSP binaries in the
91: Galactic disk. If these effects are relativistic, they allow estimates of
92: the the total binary and component masses.
93:
94: When Anderson~et~al.~(1997)\nocite{awkp97} published timing
95: solutions for M5A and B they used the eccentricity of M5B to detect
96: its periastron advance. However, the large relative uncertainty of the
97: measurement did not allow any astrophysically useful constraints on
98: the total mass of the binary. In this paper we report the results of
99: recent (2001 to 2008) 1.1-1.6~GHz (L-band) observations of M5. The
100: first 2001 observations were part of an Arecibo search for pulsars in
101: GCs, which found 11 new MSPs \cite{hrs+07}. Three of these
102: were found in M5, and subsequent observations of this GC were made
103: chiefly with the aim of timing these new discoveries (Stairs et
104: al. 2008, in preparation). M5A and B are in the same radio beam as
105: the new pulsars and they are clearly detectable in the L-band data,
106: permitting timing (see Figs. \ref{fig:profiles} and
107: \ref{fig:residuals}) of much better (M5A) or comparable (M5B) quality
108: to that obtained at 430 MHz by Anderson et al (1997). The whole dataset
109: now spans nearly 19 years and provides much improved timing parameters.
110:
111: \section{Observations, data reduction and timing}
112: \label{sec:timing}
113:
114: M5A and B were observed with the Arecibo 305-m radio telescope from the
115: time of their discovery in 1989 April until 1994 July using the 430-MHz
116: Carriage House line feed. For this, a 10-MHz band centered at 430~MHz
117: was used. The Arecibo correlation spectrometer made a 3-level
118: quantization of the signal and correlated this for a total of 128
119: lags. These data were then integrated for 506.58561 $\mu$s, and the
120: orthogonal polarizations added in quadrature before being written to
121: magnetic tape. The L-band observations began in 2001 June, using the
122: ``old'' Gregorian L-Wide receiver ($T_{sys}$ = 40~K at 1400 MHz). The
123: ``new'' L-Wide receiver ($T_{sys} = 25$K at 1400 MHz) has been used
124: since it was installed in the Gregorian dome in 2003 February. The
125: Wide-band Arecibo Pulsar Processors (WAPPs, Dowd, Sisk \& Hagen
126: 2000\nocite{dsh00}) make a 3-level digitization of the voltages over a
127: 100-MHz band for both (linear) polarizations, autocorrelating these
128: for a total of 256 lags. The data are then integrated for a total of
129: 64$\,\mu$s and the orthogonal polarizations added in quadrature and
130: written to disk. At first, only one WAPP was available, and centered
131: the observing band at 1170~MHz or 1425~MHz. From 2003, three more
132: WAPPs have been available, and we now use three of them to
133: observe simultaneously at 1170, 1410 and 1510~MHz, the cleanest bands
134: within the wide frequency coverage of the new L-Wide receiver.
135:
136: For all observations, the lags were Fourier transformed to generate
137: power spectra. For the L-band observations, the power spectra were
138: partially dedispersed at a dispersion measure (DM) of 29.5~cm$^{-3}$pc
139: and stored as a set of 16 sub-bands on the disks of the Borg computer
140: cluster at McGill University. At 1170~MHz, the partial
141: dedispersion introduces an extra smearing of 18 and 1.6$\mu$s for M5A
142: and B respectively. Adding these values in quadrature to the
143: dispersive smearing per channel (60.9 and 59.7 $\mu$s respectively),
144: we obtain a total dispersive smearing of 63.5 and 59.7 $\mu$s for M5A
145: and B respectively, i.e., the sub-banding introduces very little
146: extra smearing. The 430-MHz power spectra and L-band sub-bands were
147: dedispersed at the known DM of these pulsars and folded modulo their
148: spin periods. All the L-band data reported in this paper were
149: processed using the {\tt PRESTO} pulsar software
150: package\footnote{http://www.cv.nrao.edu/$\sim$sransom/presto}.
151:
152: \begin{figure}
153: \epsscale{1.2}
154: \plotone{fig1.eps}
155: \caption{\label{fig:profiles} Full-cycle pulse profiles for M5A (top)
156: and M5B (bottom) at 1170 MHz, obtained by averaging the
157: best detections of these pulsars at this frequency. The horizontal
158: error bars denote the total time resolution of the system at
159: 1170~MHz: 90.2 and 87.5~$\mu$s for M5A and B respectively.}
160: \end{figure}
161:
162: We added the best 1170 MHz detections of both pulsars to derive
163: ``standard'' pulse profiles, and these are displayed in
164: Figure~\ref{fig:profiles}. A minimal set of Gaussian curves were
165: fitted to these profiles to derive synthetic templates for each
166: pulsar, and these were then cross-correlated
167: with each observation's pulse profile in the Fourier domain \cite{tay92}
168: to obtain topocentric times of arrival (TOAs). Adding all the
169: 1170 MHz observations irrespective of their signal-to-noise ratio
170: (SNR, this varies from day to day because of diffractive interstellar
171: scintillation), we obtain a ``global'' pulse profile which has a lower SNR
172: than the ``standard'' pulse profile. We calculated the average flux
173: densities for both pulsars (see Table \ref{tab:parameters}) from the
174: off-pulse r.m.s. in their global profile, assuming a system equivalent
175: flux density of 3.5 Jy which is valid for 1170~ MHz at the high zenith
176: angles required to observe M5.
177:
178: \begin{deluxetable*}{ l c c}
179: \tabletypesize{\scriptsize}
180: \tablecolumns{3}
181: \tablewidth{0pc}
182: \tablecaption{Parameters for two pulsars in NGC~5904}
183: \tablehead{
184: \colhead{Observation and flux parameters} &
185: \colhead{PSR~B1516+02A} &
186: \colhead{PSR~B1516+02B}}
187: \startdata
188: Start of 430-MHz observations \dotfill & \multicolumn{2}{c}{47635} \\
189: End of 430-MHz observations \dotfill & \multicolumn{2}{c}{49432} \\
190: Number of TOAs @ 430 MHz \dotfill & 86 & 82 \\
191: Residual rms @ 430 MHz ($\mu$s) \dotfill & 49 & 114 \\
192: Start of L-band observations \dotfill & \multicolumn{2}{c}{52087} \\
193: End of L-band observations \dotfill & \multicolumn{2}{c}{54497} \\
194: Number of TOAs @ L-band \dotfill & 1278 & 162 \\
195: Uncertainty scale factor\tablenotemark{a} \dotfill & 1.50 & 1.05 \\
196: Residual rms @ L-band ($\mu$s) \dotfill & 9 & 72 \\
197: Average flux density @ 1170 MHz (mJy) \dotfill & 0.155 & 0.027 \\
198: \hline
199: \multicolumn{1}{c}{Ephemeris} & \multicolumn{2}{c}{} \\
200: \hline
201: Reference Epoch (MJD) \dotfill & 54000 & 54000 \\
202: Right Ascension, $\alpha$ (J2000) \dotfill &
203: $15^{\rm h}18^{\rm m}33\fs 32307(6)$\tablenotemark{b} &
204: $15^{\rm h}18^{\rm m}31\fs 4625(8)$ \\
205: Declination, $\delta$ (J2000)\dotfill &
206: $+02^\circ 05\arcmin 27\farcs 435(3)$ &
207: $+02^\circ 05\arcmin 15\farcs 30(3)$ \\
208: Proper motion in $\alpha$, $\mu_{\alpha}$ (mas yr$^{-1}$, J2000)
209: \dotfill & 4.6(4) & 3.4(1.2) \\
210: Proper motion in $\delta$, $\mu_{\delta}$ (mas yr$^{-1}$, J2000)
211: \dotfill & $-$8.9(1.0) & $-$11.8(2.8) \\
212: Spin frequency, $\nu$ (Hz) \dotfill &
213: 180.063624055103(3) &
214: 125.83458757935(6) \\
215: Time derivative of $\nu$, $\dot{\nu}$ (10$^{-15}$ Hz s$^{-1}$) \dotfill &
216: $-$1.33874(4) &
217: 0.05233(15) \\
218: Dispersion Measure, DM (pc\,cm$^{-3}$) \dotfill & 30.0545(10) & 29.46(3) \\
219: Orbital period, $P_b$ (days) \dotfill & \nodata & 6.8584538(3) \\
220: Projected size or orbit, $x$ (l-s) \dotfill & \nodata & 3.04857(2) \\
221: Orbital eccentricity, $e$ \dotfill & \nodata & 0.137845(10) \\
222: Time of passage through periastron, $T_0$ (MJD) \dotfill & \nodata & 54004.02042(15) \\
223:
224: Longitude of periastron, $\omega$ ($^\circ$) \dotfill & \nodata & 359.898(8) \\
225: Rate of advance of periastron, $\dot{\omega}$ ($^\circ$ yr$^{-1}$)
226: \dotfill & \nodata & 0.0142(7) \\
227: Second time derivative of $\nu$, $\ddot{\nu}$ ($10^{-27}$ Hz s$^{-2}$) \dotfill &
228: [$-0.6 \pm 0.5$]\tablenotemark{c} & [$4.4 \pm 5.4$] \\
229: Time derivative of $x$, $\dot{x}$ ($10^{-12}$ l-s/s) \dotfill &
230: \nodata & [$-\,0.12\,\pm\,0.09$] \\
231: Time derivative of $P_B$, $\dot{P_B}$ ($10^{-12}$) \dotfill &
232: \nodata & [$15\,\pm\,31$] \\
233: \hline
234: \multicolumn{1}{c}{Derived parameters} & \multicolumn{2}{c}{}\\
235: \hline
236: Spin period, $P$ (ms) \dotfill & 5.55359254401089(11) & 7.946940656275(4) \\
237: Time derivative of $P$, $\dot{P}$ (10$^{-21}$ s s$^{-1}$) \dotfill &
238: 41.2899(13) & $-$3.306(10) \\
239: Mass function, $f\,(M_{\sun})$ \dotfill & \nodata & 0.000646723(13) \\
240: Total system mass, $M$ $(M_{\sun})$ \dotfill & \nodata & 2.29(17) \\
241: Maximum pulsar mass, $M_{p, \rm max}\,(M_{\sun})$ \dotfill & \nodata & 2.52 \\
242: Minimum companion mass, $M_{c, \rm min}\,(M_{\sun})$ \dotfill & \nodata & 0.13 \\
243: \enddata
244: \tablenotetext{a}{This is the scale factor for TOA uncertainties
245: required to achieve a reduced $\chi^2$ of unity in the solution.}
246: \tablenotetext{b}{1-$\sigma$ uncertainties are presented in
247: parenthesis. These are twice the estimate obtained using the
248: Monte-Carlo bootstrap method.}
249: \tablenotetext{c}{Values in square brackets are not
250: considered to be significant. They were not fit when
251: determining the remaining timing parameters.
252: % I really think these should stay as they are!
253: }
254: \label{tab:parameters}
255: \end{deluxetable*}
256:
257: \begin{figure*}
258: \epsscale{0.7}
259: \plotone{fig2.ps}
260: \caption{\label{fig:residuals} Timing residuals for M5A (upper) and
261: M5B (lower), obtained with the ephemerides in
262: Table~\ref{tab:parameters}. The large gap in time coverage is mostly
263: due to the Arecibo upgrade.}
264: \end{figure*}
265:
266: The TOAs were analyzed with
267: {\tt TEMPO}\footnote{\url{http://www.atnf.csiro.au/research/pulsar/tempo/}},
268: using the DE~405 Solar System ephemeris \cite{sta98b} to model the
269: motion of the Arecibo radio telescope relative to the Solar System
270: Barycenter. The orbital parameters for M5B were modeled using the
271: Damour \& Deruelle orbital model \cite{dd85,dd86}. For most of the
272: early 430-MHz TOAs we have no reliable uncertainty
273: estimates. Therefore, in order to achieve a reduced $\chi^2$ of 1 for
274: both pulsars, we attributed a constant uncertainty to the 430-MHz TOAs
275: that is similar to their unweighted rms. These times are listed in
276: Table \ref{tab:parameters}. With these TOAs, we obtain
277: timing parameters that are virtually identical to those obtained by the
278: previous analysis (Anderson et al. 1997). For the new L-band data, we
279: find the TOA uncertainties to be under-estimated in the case of
280: M5A. Multiplying these by a factor of about 1.5, we achieve a reduced
281: $\chi^2$ of 1 for its L-band TOAs. In the case of M5B, this factor is
282: 1.05. These values are similar to those derived for other MSPs
283: timed with the same software \cite{frb+08}.
284:
285: The resulting timing parameters and their 1-$\sigma$ uncertainties are
286: presented in Table \ref{tab:parameters}. We estimate these
287: uncertainties to be twice the 1-$\sigma$ Monte-Carlo bootstrap
288: \cite{et93,ptvf92} uncertainties. We discuss the validity of this
289: choice for the particular case of the periastron advance of
290: M5B in \S\ref{sec:omega-dot}. All the parameters that vary in time
291: ($\alpha$, $\delta$, $\nu$ and $\omega$) are estimated for the
292: arbitrary epoch, MJD = 54000 (2006 September 22); $T_0$ was the first
293: periastron passage to occur after that date. The post-fit timing
294: residuals are essentially featureless at the present timing accuracy
295: (see Fig. \ref{fig:residuals}). The large gap without measurements
296: between the early 430-MHz data and later L-band data is in
297: part due to the Arecibo upgrade of the late 1990's. We included an
298: arbitrary time step between these two datasets in the fit. We have tested
299: the timing solution by introducing extra pulsar rotations between the
300: two datasets, but these are always absorbed by this arbitrary time
301: step, with no other changes in the fitted timing parameters.
302:
303: The positions, periods and period derivatives that we have obtained are
304: consistent with those of Anderson et al (1997). In what follows, we
305: analyze solely the newly measured parameters: the proper motions and
306: the rate of advance of periastron of M5B.
307:
308: \section{Proper motions}
309:
310: In the reference frame of a GC, the rms of the velocities of its
311: pulsars along the orthogonal axes perpendicular to the line of sight
312: should be the same as the rms of their velocities along the line of
313: sight. The stellar rms velocity along the line of sight at the center
314: of M5 is 7.15~km~s$^{-1}$ \cite{web85}; the rms of the pulsar
315: velocities should be smaller given the larger masses of the neutron
316: stars (NSs). At the distance of M5, 7.5~kpc \cite{har96}, the stellar rms
317: velocity represents a proper motion rms of only
318: 0.2~mas~yr$^{-1}$. Given the present measurement precision, the
319: estimated pulsar proper motions should be
320: mutually consistent and reflect only the proper motion of the GC.
321: The proper motion measurements in Table~\ref{tab:parameters} are
322: indeed $\sim 1\,\sigma$ consistent with each other; from their weighted average
323: we derive a proper motion for M5 of $\mu_{\alpha}\,=\,(4.3\,\pm\,0.4)$
324: mas yr$^{-1}$ and $\mu_{\delta}\,=\, (-9.6\,\pm\,1.0)$ mas yr$^{-1}$.
325:
326: M5 is one of the four GCs in the Galaxy for which optical proper
327: motion measurements have not provided consistent (i.e., agreeing within
328: the formal uncertainty estimates) results
329: (Dinescu, Girard \& van Altena 1999)\nocite{dga99}. Our M5
330: proper motion measurement is in marginal agreement with the values derived by
331: Scholz et al. [1996, $\mu_{\alpha}\,=\,(6.7\,\pm\,0.5)$ mas yr$^{-1}$
332: and $\mu_{\delta}\,=\, (-7.8\,\pm\,0.4)$ mas
333: yr$^{-1}$]\nocite{soh+96}, but is in good
334: agreement with the values derived from Hipparcos [Odenkirchen et
335: al. 1997, $\mu_{\alpha}\,=\,(3.3\,\pm\,1.0)$ mas yr$^{-1}$ and
336: $\mu_{\delta}\,=\, (-10.1\,\pm\,1.0)$ mas yr$^{-1}$]\nocite{obgt97}.
337:
338: \begin{figure}
339: \epsscale{1.15}
340: \plotone{fig3.eps}
341: \caption{\label{fig:optical_M5B} A small portion of a GO-10120
342: {\em HST} image of M5, centered on the position of M5B, in
343: negative. The picture was taken through the F435W filter of the Advanced
344: Camera for Surveys/Wide Field Camera. The dark circle indicates the
345: position of M5B and the radius (0.2\arcsec) corresponds to the
346: 2$\sigma$ uncertainty in the absolute astrometry of the image. The
347: nearest stars (white circles) lie at 2.7 and 3.0$\sigma$ from the
348: radio position; these are indicated as A and B and are discussed in \S
349: \ref{sec:optical}.}
350: \end{figure}
351:
352: \section{Search for the optical counterpart of M5B}
353: \label{sec:optical}
354:
355: We have used the astrometric information on M5B to search for an
356: optical counterpart in archival {\em HST} ACS/WFC data from programs GO
357: 10120 and GO 10615. The GO-10120 images were taken on 2004 August 1
358: through the F435W, F625W and F658N filters. Since the uncertainty in
359: the absolute astrometry of {\em HST} data is 1--2\arcsec, we first tie
360: the astrometry of the GO-10120 ACS images to the ICRS frame using
361: UCAC2 stars \cite[positional accuracy $\lesssim$0.070\arcsec~down to
362: the magnitude limit of the UCAC2 catalog,][]{zachea04}. Since UCAC2
363: standards in the small ACS field (3.4\arcmin$\times$3.4\arcmin) are
364: scarce, we use ground-based imaging of M5 to
365: derive secondary standards. We retrieved from the public archive of
366: the 2.5 m Isaac Newton Telecope (INT) a 30-s Sloan-r Wide
367: Field Camera image taken on 2004 June 8 and processed only the chip
368: that contained the core of M5 (field of view
369: $\sim$23\arcmin$\times$11\arcmin). Astrometric calibration of this
370: image was achieved using 308 UCAC2 stars with positions corrected for
371: proper motion to the epoch of the INT image. After fitting for shift,
372: rotation angle, scale factor and distortions, the final solution has
373: rms residuals of 0.050\arcsec~in right ascension and 0.047\arcsec~in
374: declination. We selected a set of 198 secondary standards from
375: unsaturated and relatively isolated stars in the INT image. These were
376: used in turn to compute an astrometric solution for the short (70s)
377: F435W distortion-corrected (using the {\em multidrizzle} software)
378: exposure. The resulting fit for shift, rotation angle and scale factor
379: has rms residuals of 0.017\arcsec~in right ascension and
380: 0.014\arcsec~in declination. We estimate the final 1$\sigma$ accuracy
381: of our ACS absolute astrometry as the quadratic sum of the errors in
382: the UCAC2 astrometry, the UCAC2--INT tie and the INT--{\em HST} tie,
383: i.e.~0.1\arcsec~(or 2 ACS pixels).
384:
385: The position of M5B at the epoch of the GO-10120 observations is shown
386: in Fig.~\ref{fig:optical_M5B}. No optical sources are detected within
387: the 2-$\sigma$ error circle of the radio position (this only includes
388: the uncertainty in the absolute astrometry as the uncertainties in the
389: radio position are negligible). The photometry shows the nearest
390: sources---indicated as ``1'' and ``2'' at distances 2.7 and 3.0\,$\sigma$,
391: respectively---to be
392: main-sequence (MS) stars located $\sim$1.3 and $\sim$6.1 mag (in F435W)
393: below the MS turnoff. Using the M5 turnoff magnitude from
394: Sandquist et al.~(1996) \nocite{sandea96} and assuming $B \approx
395: F435W$, the 14-Gyr isochrones from Bergbusch \& Vandenberg (1992)
396: \nocite{bergvand92} imply that ``1'' and ``2'' are $\sim$0.75 and
397: $\sim$0.4$M_{\sun}$ stars, respectively (for a distance modulus
398: $(m-M)=14.41$ and $E(B-V)=0.03$, see Sandquist et al 1996). To check
399: for fainter stars, we stacked F435W images taken with fraction-pixel offsets
400: from {\em HST} program GO 10615 (2006 Feb 15)
401: into a deep (8500 s) high-resolution (twice-oversampled)
402: masterframe. Psf photometry revealed no additional stars within 3$\sigma$
403: of M5B. A conservative upper limit for the detection limit is
404: m$_{F435W}\approx 26-26.5$ mag which corresponds to
405: $\sim$0.25--0.3$M_{\sun}$ for MS stars in M5 (although close to
406: star ``1'' the sensitivity is lower).
407:
408: In summary, we cannot exclude stars ``1'' or ``2'' as counterparts of
409: M5B. However, the astrometry suggests that it is more likely that the
410: true optical counterpart is fainter than our detection limit. That
411: implies that the companion is either a WD or a faint, low-mass MS star.
412:
413: \section{Periastron advance of M5B}
414: \label{sec:omega-dot}
415:
416: We have determined a highly significant estimate for the rate of
417: advance of periastron of M5B: $\dot{\omega}\,=\,0.0142(7)^\circ \rm
418: yr^{-1}$. The 1-$\sigma$ estimate provided directly by {\tt TEMPO} is
419: $\dot{\omega}\,=\,0.01422(43)^\circ \rm yr^{-1}$. To verify that these
420: values are realistic, we have kept $\dot{\omega}$ fixed, and fitted all
421: the remaining timing parameters, recording the resulting
422: $\chi^2$. Doing this for a range of values of $\dot{\omega}$, we
423: obtain the 1-$\sigma$ uncertainty as the half-width of the region
424: where $[\chi^2 (\dot{\omega}) - \chi^2 (\dot{\omega}_{\rm min}) < 1]$,
425: $\dot{\omega}_{\rm min}$ being the value that minimizes $\chi^2$
426: \cite{sna+02}. The result is $\dot{\omega}\,=\,0.001422(43)^\circ \rm
427: yr^{-1}$. Estimating all parameters using a Monte-Carlo Bootstrap
428: algorithm (\S \ref{sec:timing}) we obtain
429: $\dot{\omega}\,=\,0.01422(36)^\circ \rm yr^{-1}$. Furthermore, the
430: lack of significant higher derivatives of the spin frequency (see
431: Table~\ref{tab:parameters}) suggests that, with the present timing
432: precision, we are unable to detect any of this pulsar's timing noise
433: which can contaminate the physical interpretation of its timing
434: uncertainties. This is a necessary pre-condition for an accurate
435: estimation of parameter uncertainties. We therefore believe that the
436: $1 \sigma$ {\tt TEMPO} uncertainty estimates are essentially
437: accurate. We choose, however, to be more conservative by making our
438: 1-$\sigma$ uncertainties twice as large as the values suggested by the
439: Monte-Carlo method (see Table~\ref{tab:parameters}). This caution is
440: due to our use of two different datasets with a large time gap between
441: them.
442:
443: This time gap is common to many Arecibo timing data sets, such as that
444: of PSR~J0751+1807. Nice et al. (2005)\nocite{nss+05} claimed a mass of
445: $2.1\,M_{\sun}$ for that pulsar. This claim has recently been
446: retracted, the latest estimate for the pulsar mass now being
447: $1.26^{+0.14}_{-0.12} M_{\sun}$ (1$\sigma$, Nice 2007\footnote{See
448: also http://www.ns2007.org/talks/nice.pdf}).
449: The problem with the earlier estimate was not related to the gap as the
450: new estimate is based on essentially the same data. Its cause
451: was the use of a necessarily imperfect ephemeris when folding
452: the very earliest data. This resulted in orbital-dependent smearing of
453: the pulse profiles that led to an error in the
454: calculation of the orbital phase for the earliest data and an
455: over-estimate of the orbital period decay. Because those data were
456: folded online, this problem could only be solved by ignoring the
457: earliest TOAs. Our 430-MHz data contained the signals of more than one
458: pulsar, so we had to record them to tape. This allowed us to re-fold them
459: iteratively after the timing solution had been obtained (Anderson et
460: al. 1997). After updating the timing solution to 2008, we have also
461: re-folded all our L-band data; none of the pulse profiles used in this
462: work are therefore smeared due to imprecise folding.
463:
464: \subsection{Is $\dot{\omega}$ relativistic?}
465: \label{sec:relativistic}
466:
467: The possible contributions to $\dot{\omega}$ in a system containing a pulsar
468: and an extended star have been studied in detail by Lai, Bildsten \& Kaspi
469: (1995)\nocite{lbk95}. In their analysis of the binary pulsar
470: PSR~J0045$-$7319 they concluded that the only likely contribution to
471: $\dot{\omega}$ in such systems is from rotational deformation of the
472: companion. If we assume that PSR~J0045$-$7319 has a mass of
473: $1.4\,M_{\sun}$, the companion mass is $8.8\,M_{\sun}$ and its radius
474: $R_c \sim 6.4\,R_{\sun}$ \cite{bbs+95}; the orbital separation $a$ is
475: $\sim 126\,R_{\sun}$. Using these values they reached the conclusion
476: that the contribution to $\dot{\omega}$ from tidal deformations, which
477: are proportional to $(R_c / a)^3 \simeq 1.32 \times 10^{-4}$, is not
478: significant in that system. For M5B, again assuming a pulsar mass of
479: $1.4\,M_{\sun}$ with a MS companion of maximum mass $0.3\,M_{\sun}$
480: (see \S~\ref{sec:optical}) and a radius $R_c = 0.3\,r_{0.3}\,R_{\sun}$
481: (where $r_{0.3} \sim 1$), then the orbital inclination $i$ is $\sim
482: 24^{\circ}$, $a \sim 18\,R_{\sun}$ and $(R_c / a)^3 \simeq 4.5 \times
483: 10^{-6} r_{0.3}^{3}$. This is 30 times smaller than for
484: PSR~J0045$-$7319, becoming even less significant for smaller MS
485: companion masses and correspondingly higher inclinations.
486:
487: Using equations 68 and 79 of Wex (1998) we can estimate the
488: contribution to $\dot{\omega}$ due to rotational deformation:
489: \begin{equation}
490: \label{eq:rot}
491: \dot{\omega}_{\rm rot} = n \frac{k R_c^2 \hat{\Omega}^2}{a^2 (1 - e^2)^2}
492: \left( 1 - \frac{3}{2} \sin^2 \theta + \cot i \sin \theta \cos \theta
493: \cos \Phi_0 \right),
494: \end{equation}
495: where $n$ is the orbital angular frequency ($2 \pi / P_b = 1.06 \times
496: 10^{-5} \rm rad\,s^{-1}$), $k$ is the gyration radius (for a homogeneous
497: sphere this is 0.63, while for any centrally condensed objects
498: this will always be smaller: it is about 0.2 for a completely
499: convective star), $\hat{\Omega}$ is the rotation rate relative to break-up,
500: $\theta$ is the angle between the rotational and orbital angular momenta
501: (if the companion is non-degenerate, these tend to be aligned, so $\theta = 0$)
502: and $\Phi_0$ is the longitude of the ascending node in a reference frame
503: defined by the total angular momentum vector (see Fig.~9 of Wex 1998).
504:
505: For the situation discussed above (a 1.4-$M_{\sun}$ pulsar with the
506: largest possible MS companion, $R_c = 0.3 r_{0.3} R_{\sun}$), we have
507: $R_c / a = 16.55 \times 10^{-3} r_{0.3}$. We can also calculate the
508: break-up angular velocity:
509: $\Omega_{\rm max} = \sqrt{GM_c/R_c^3} = 1/ 0.3 \sqrt{GM_{\sun} /
510: (r_{0.3} R_{\sun})^3} = 2.09 \times 10^{-3} r_{0.3}^{(-3/2)}\,
511: \rm rad\,s^{-1}$. If the companion's rotation is tidally locked to
512: its orbit around the pulsar, then
513: $\hat{\Omega} = n / \Omega_{\rm max} = 5.06 \times 10^{-3} r_{0.3}^{(3/2)}$.
514: Therefore, $\dot{\omega}_{\rm rot} = 2.65 \times 10^{-15}
515: r_{0.3}^{5} {\rm rad\,s^{-1}} = 4.79 \times 10^{-6} (r_{0.3}^{5})^{\circ}
516: \rm yr^{-1}$ (or three times this if the companion were to be a
517: homogeneous sphere). This is $\sim 3 \times 10^{3}$ times smaller than
518: the value we dedetermined (\S \ref{sec:omega-dot}). If the companion
519: were to be significantly distended for its mass (i.e., if $r_{0.3} >
520: 1$), as is the case for the companion of PSR~J1740$-$5340
521: \cite{fpds01}, then $\dot{\omega}_{\rm rot}$ could be
522: significant. This scenario can be excluded, since no optical
523: counterpart is readily detectable within $\sim 2.5 \sigma$ of the pulsar
524: (see \S \ref{sec:optical}).
525:
526: If the companion were to be a WD, then it could be more massive than
527: $0.3\,M_{\sun}$ and still evade optical detection. Irrespective of its mass,
528: the contribution to $\dot{\omega}$ from the tidal deformation of a WD
529: is negligible, but that is not necessarily the case for the contribution
530: from rotational deformation. As an example, we re-calculate
531: $\dot{\omega}_{\rm rot}$ for a 0.3-$M_{\sun}$ WD. For WDs, we have $k = 0.45$
532: \cite{lp98}, more than twice as large as for fully convective
533: stars. For WDs $r_{0.3}$ is of the order of 0.1, i.e., the $(R_c/a)^2$
534: term in eq.~\ref{eq:rot} would be $\sim 10^{2}$ times smaller than
535: discussed for $r_{0.3} \sim 1$. However, a WD companion is not
536: likely to be tidally locked. If it were spinning fast, then
537: $\hat{\Omega} \sim 1$. While there is no special {\it a priori} reason
538: why this should be true, it is a possibility that cannot be excluded. This
539: would mean that $\hat{\Omega}^2$ could be $\sim 4 \times 10^{4}$ times
540: larger than discussed above, with $\dot{\omega}_{\rm rot}$ similar to the
541: observed $\dot{\omega}$. This is particularly so for the larger-sized WDs
542: (those with the lowest masses). Following Splaver et al. (2002), we
543: note first that if the companion is not tidally locked, the angular
544: momenta of the orbit and companion spin will probably not be aligned
545: ($\theta \neq 0$). In this case, the spin of the companion will induce
546: a precession of the orbital plane. This will cause a change in $i$,
547: affecting the projected semi-major axis of the orbit which will vary
548: with a rate $\dot{x}$. Rewriting equation 81 of Wex
549: (1998)\nocite{wex98}, we can relate $\dot{\omega}_{\rm{rot}}$ to $\dot{x}$:
550: \begin{equation}
551: \dot{\omega}_{\rm rot} =
552: \frac{\dot{x}}{x}\,
553: \left(
554: \tan i
555: \frac{1-\frac{3}{2}\sin^2\theta}{%
556: \sin\theta\cos\theta\sin\Phi_0}
557: + \cot \Phi_0
558: \right).
559: \end{equation}
560: This equation has the advantage that it does not depend on the mass
561: (or the nature) of the companion.
562: Thus, our observed 2-$\sigma$ upper limit of $|\dot{x}/x|< 9.6 \times
563: 10^{-14}\,\rm s^{-1}$ implies $|\dot{\omega}_{\rm{rot}}|\,<\,(1.7 \,
564: \times\,10^{-4})^\circ$\,yr$^{-1}$ times a geometric factor. In 80\% of
565: cases this geometric factor will be smaller than 10 and
566: the upper limit for $\dot{\omega}_{\rm{rot}}$ is similar to the present
567: measurement uncertainty for $\dot{\omega}$.
568:
569: To summarize, $\dot{\omega}_{\rm rot}$ can only be significant
570: if the companion is degenerate, rotating near breakup velocity, and
571: with its rotational angular momentum nearly aligned with the orbital
572: angular momentum, making $\dot{x}$ undetectable. Otherwise
573: $\dot{\omega}$ is relativistic.
574:
575: \begin{figure*}
576: \plotone{fig4.eps}
577: \caption{\label{fig:mass_mass} Constraints on the masses of M5B
578: and its companion. The hatched region is excluded by knowledge of
579: the mass function and by $\sin i \leq 1$. The diagonal dashed lines
580: correspond to a total system mass that causes a general-relativistic
581: $\dot{\omega}$ equal to, or within 1-$\sigma$ of, the measured value.
582: The six solid curves indicate constant inclinations. We also
583: display the probability density function for
584: the mass of the pulsar ({\em top}) and the mass of the companion
585: ({\em right}), and mark the respective medians with vertical
586: (horizontal) lines. For comparison, the gray bars indicate the range
587: of ``normal'' NS masses (see \S\ref{sec:masses}).}
588: \end{figure*}
589:
590: \subsection{Binary, pulsar and companion masses}
591: \label{sec:masses}
592:
593: When $\dot{\omega}$ is solely due to the effects of general
594: relativity, we can measure the total mass of a binary system:
595: \begin{equation}
596: \label{eq:total_mass}
597: {M} = \left(\frac{P_b}{2 \pi}\right)^{5/2}
598: \left[\frac{(1 - e^2)\,\dot{\omega}}{3}\right]^{3/2}
599: \left(\frac{1}{T_\sun}\right),
600: \label{eq:totmass}
601: \end{equation}
602: where $T_\sun \equiv G M_{\sun} / c^3\,=\,4.925490947 \mu$s. For M5B,
603: we obtain $M\,=\,(2.29\,\pm\,0.17)\,M_{\sun}$. For the nominal value
604: of $\dot{\omega}$ and a median $i$ of 60$^\circ$,
605: the mass of the companion is 0.173\,$M_{\sun}$ and the mass of the
606: pulsar is 2.11\,$M_{\sun}$. This is well above all NS masses that have
607: been precisely measured to date.
608:
609: We calculated a 2-D probability distribution function (pdf) for the
610: mass of the pulsar and the mass of the companion, assuming that the
611: pdf for $\dot{\omega}$ is a Gaussian with the half-width equal to
612: the 1-$\sigma$ uncertainty listed in Table~\ref{tab:parameters} and an
613: {\em a priori} constant probability for $\cos i$. The two-dimensional
614: pdf is then projected in both dimensions, resulting in 1-D pdfs for
615: the mass of the pulsar and the mass of the companion. These are
616: displayed graphically in Fig.~\ref{fig:mass_mass}. The pulsar
617: definitely has a mass smaller than 2.52\,$M_{\sun}$, and the
618: companion has a mass larger than 0.13\,$M_{\sun}$, the median and
619: 1-$\sigma$ limits for the pulsar and companion mass are
620: $2.08^{+0.18}_{-0.19} M_{\sun}$ and $0.172^{+0.107}_{-0.023} M_{\sun}$
621: respectively. There is a 99\%, 95\% and 90\% probability that the
622: pulsar is more massive than 1.38, 1.72 and 1.82\,$M_{\sun}$
623: respectively. There is a 0.77\% probability that $i$ is low
624: enough to make the NS mass fall within the range of the components of
625: double neutron star (DNS) systems: from 1.20~$M_{\sun}$ measured for
626: the companion of PSR~J1756$-$2251 \cite{fkl+05} to 1.44~$M_{\sun}$
627: measured for PSR~B1913+16 \cite{wt03}. For M5B, assuming its nominal
628: value of $\dot{\omega}$, these mass limits would imply that
629: $7.9^\circ\,<\,i\,<\,10.2^{\circ}$.
630:
631: \begin{deluxetable*}{ l c r c c c c c c c c}
632: \tabletypesize{\footnotesize}
633: \tablecolumns{10}
634: \tablewidth{0pc}
635: \tablecaption{Millisecond Pulsar Masses}
636: \tablehead{
637: \colhead{Name PSR} &
638: \colhead{GC} &
639: \colhead{$P$ (ms)} &
640: \colhead{$P_b$ (days)} &
641: \colhead{$e$} &
642: \colhead{$f/M_{\sun}$} &
643: \colhead{$M/M_{\sun}$\tablenotemark{a}} &
644: \colhead{$M_c/M_{\sun}$} &
645: \colhead{$M_p/M_{\sun}$} &
646: \colhead{Method\tablenotemark{b}} &
647: \colhead{Ref.\tablenotemark{c}}}
648: \startdata
649:
650: \multicolumn{11}{c}{Selected MSP Mass Measurements} \\
651:
652: \hline
653:
654: J0751+1807 & - & 3.47877 & 0.26314 & 0.00000 & 0.0009674 & - & - &
655: $1.26^{+14}_{-12}$ & $\dot{P_b}$, $s$ & 1 \\
656:
657: J1911$-$5958A & NGC~6752 & 3.26619 & 0.83711 & $<$0.00001 &
658: 0.002688 & 1.58$^{+0.16}_{-0.10}$ & 0.18(2) & $1.40^{+0.16}_{-0.10}$ & Opt. &
659: 2 \\
660:
661: J1909$-$3744 & - & 2.94711 & 1.53345 & 0.00000 & 0.003122 & $1.67^{+3}_{-2}$ & 0.2038(22) & $1.47^{+3}_{-2}$ & $r,s$ & 3 \\
662:
663: J0437$-$4715 & - & 5.75745 & 5.74105 & 0.00002 & 0.001243 & 2.01(20) &
664: 0.254(14) & 1.76(20) & $r,s$ & 4 \\
665:
666: J1903+0327 & - & 2.14991 & 95.1741 & 0.43668 & 0.139607 & 2.88(9) & 1.07(2) & 1.81(9) & $\dot{\omega}$, $s$ & 5 \\
667:
668: \hline
669:
670: \multicolumn{11}{c}{Binary systems with indeterminate orbital inclinations} \\
671:
672: \hline
673:
674: J0024$-$7204H & 47~Tucanae & 3.21034 & 2.35770 & 0.07056 & 0.001927 & 1.61(4) &
675: $> 0.164$ & $< 1.52$ & $\dot{\omega}$ & 6 \\
676:
677: J1824$-$2452C & M28 & 4.15828 & 8.07781 & 0.84704 & 0.006553 & 1.616(7) &
678: $> 0.260$ & $< 1.367$ & $\dot{\omega}$ & 7 \\
679:
680: \hline
681:
682: J1748$-$2446I & Terzan~5 & 9.57019 & 1.328 & 0.428 & 0.003658 & 2.17(2) &
683: $> 0.24$ & $< 1.96$ & $\dot{\omega}$ & 8 \\
684:
685: J1748$-$2446J\tablenotemark{d} & Terzan~5 & 80.3379 & 1.102 & 0.350 & 0.013066 & 2.20(4) &
686: $> 0.38$ & $< 1.96$ & $\dot{\omega}$ & 8 \\
687:
688: B1516+02B & M5 & 7.94694 & 6.85845 & 0.13784 & 0.000647 & 2.29(17) &
689: $> 0.13$ & $< 2.52$ & $\dot{\omega}$ & \S\ref{sec:masses} \\
690:
691: J0514$-$4002A\tablenotemark{e} & NGC~1851 & 4.99058 & 18.7852 &
692: 0.88798 & 0.145495 & 2.453(14) &
693: $> 0.96$ & $< 1.52$ & $\dot{\omega}$ & 9 \\
694:
695: J1748$-$2021B & NGC~6440 & 16.76013 & 20.5500 & 0.57016 & 0.000227 & 2.91(25) &
696: $> 0.11$ & $< 3.3$ & $\dot{\omega}$ & 10 \\
697:
698: \enddata
699: \tablenotetext{a}{Binary systems are sorted according to the total
700: estimated mass $M$.}
701: \tablenotetext{b}{Methods are: $\dot{P_b}$ - relativistic orbital
702: decay, $r,s$ - Shapiro delay ``shape'' and ``range'', ``Opt'' -
703: optically derived mass ratio, plus mass estimate based on spectrum
704: of companion, $\dot{\omega}$ - precession of periastron.}
705: \tablenotetext{c}{References are 1: \cite{nice07} (total and companion
706: masses not provided), 2: \cite{bkkv06},
707: 3: \cite{hbo06}, 4: \cite{vbs+08}, 5: \cite{crl+08}, 6:
708: \cite{fck+03}, 7: \cite{brf+08}, 8: \cite{rhs+05}, 9: (Freire, Ransom
709: \& Gupta 2007)\nocite{frg07}; 10: \cite{frb+08}}
710: \tablenotetext{d}{This pulsar is not technically a MSP, its spin
711: period is longer than those found in most DNS systems. However,
712: given the similarity of its orbital parameters to those of
713: Terzan~5~I, we assume that it had a similar formation history.}
714: \tablenotetext{e}{Because of its large companion mass and eccentricity,
715: this system is thought to have formed in an exchange interaction
716: \cite{fgri04}.}
717: \label{tab:msp_masses}
718: \end{deluxetable*}
719:
720: \section{Statistical evaluation of mass measurements}
721:
722: M5B has the second largest mass estimate among all known NSs
723: after PSR~J1748$-$2021B (NGC 6440B). Because of indeterminate
724: orbital inclinations, all mass estimates based solely on $\dot{\omega}$
725: are probabilistic statements: one more PK parameter is necessary
726: to have an unambiguous determination of $i$ and $M_p$. No such parameters
727: have yet been measured for M5B, NGC~6440B or any other eccentric MSP
728: binaries in GCs; this is in some cases due to their low timing
729: precision (like M5B, which is faint and has a broad pulse profile),
730: and in others to their small timing baselines (like
731: NGC~6440B). Nevertheless, unambiguous upper limits for the pulsar
732: masses and lower limits for
733: the companion masses can always be obtained from a measurement
734: of a relativistic $\dot{\omega}$ alone (see Table~\ref{tab:msp_masses}).
735: Furthermore, in systems where the mass function is very small and the
736: total binary mass is very large (as for M5B and NGC~6440B) there is a
737: much greater probability that most of the mass of the binary belongs to
738: the pulsar itself, as described in \S\ref{sec:masses}.
739:
740: \begin{figure}
741: \epsscale{1.15}
742: \plotone{fig5.eps}
743: \caption{\label{fig:binaries}
744: Cosine of the orbital inclination $i$ as a function of total binary
745: mass for the eccentric MSP binaries in GCs. For
746: each binary, the upper curve assumes a pulsar mass of $1.2\,M_{\sun}$
747: and the lower curve $1.44\,M_{\sun}$. These are different for each
748: binary because of their different mass functions. The vertical lines
749: indicate the total masses corresponding to $\dot{\omega}$ and its
750: $\pm\,1, 2$ and $3-\sigma$ uncertainties, based on the most recent
751: timing. If we assume randomly oriented orbits, then for any given total
752: mass value the vertical distance between the two curves gives us the
753: probability of the pulsar mass falling within the $1.2 - 1.44\,M_{\sun}$
754: range. With the exception of NGC~1851A, these probabilities are
755: significantly smaller for the more massive systems.}
756: \end{figure}
757:
758: \subsection{Evidence for high average neutron star masses}
759:
760: An interesting feature of the eccentric binary MSPs in GCs is
761: that as the binary mass increases, the mass function $f$ does not
762: increase (see Table~\ref{tab:msp_masses}). The exception is
763: NGC~1851A; a system thought to have resulted from an exchange
764: interaction \cite{fgri04}. If these binaries were to have
765: $M_p\,<\,1.44\,M_{\sun}$ then the increase in total mass would be due
766: to higher companion masses, resulting in a general trend to higher
767: mass functions. This is generally not the case.
768:
769: If we assume that all these GC MSPs have ``normal'' masses (between
770: 1.2 and $1.44\,M_{\sun}$), we can then calculate the orbital
771: inclinations of these binaries from their total masses and mass
772: functions. These are displayed graphically in
773: Fig.~\ref{fig:binaries}. Of the five massive GC systems, four seem to
774: have small orbital inclinations (i.e., with $\cos i\,>\,0.8$), the
775: exception being NGC~1851A. {\em A priori}, one would expect only one
776: out of five systems to have such a small $i$. We have used
777: a Kolmogorov-Smirnov test to compare the $\cos i$ values corresponding
778: to the nominal $\dot{\omega}$ values and $M_p = 1.44 M_{\sun}$ (only
779: possible to calculate for the five massive binaries) with a fake set of
780: 100 randomly oriented binary systems (i.e., with a uniform
781: distribution of $\cos i$). We obtain a 0.46\% probability that the
782: observed distribution is extracted from this set with random
783: orbital orientations. If we use instead $M_p = 1.2 M_{\sun}$, we can
784: calculate $\cos i$ for all the eccentric GC binaries. The probability
785: that the resulting distribution of $\cos i$ is selected from the set
786: with random orientations is $7.7 \times 10^{-5}$.
787:
788: Such low inclinations might be less unlikely were there a tendency
789: for pulsars to emit in a plane perpendicular to their orbit. During
790: accretion, orbital angular momentum is transferred to the NS,
791: making its rotation axis nearly perpendicular to its orbital
792: plane. If the angle between the magnetic and rotational axes of
793: pulsars ($\alpha$) is small, the magnetic axis will describe a narrow
794: cone nearly perpendicular to the orbital plane. However, no such
795: tendency for small $\alpha$ has been described in the literature.
796: If a pulsar has a small $\alpha$, its beam will probably illuminate
797: a smaller fraction of the sky, particularly if it is narrow. This
798: can only make the low-$\alpha$ objects {\em less} likely to be
799: detected\footnote{The pulse profiles also tend to be wider in these
800: nearly aligned rotators, further hindering their detection. We see
801: no correlation between the ``apparent'' inclinations in
802: Fig.~\ref{fig:binaries} and the pulse-widths, although the pulsar
803: profile of M5B is quite broad (Fig. \ref{fig:profiles}).}.
804: Furthermore, no such tendency towards low orbital inclinations is seen
805: among the two lighter binary MSPs in GCs, nor among the systems
806: with estimated orbital inclinations: NGC~6752A has $i > 70^\circ$
807: \cite{bkkv06}; PSR~J1909$-$3744 has $i = 86.6(1)^\circ$
808: \cite{jhb+05,hbo06}; PSR~J0437$-$4715 has $i \sim 43 ^\circ$
809: \cite{vbs+08} and the massive binary PSR~J1903+0327 has $i \sim 79
810: ^\circ$ \cite{crl+08}\footnote{In many cases, high orbital inclinations
811: enable the measurement of the mass of a MSP through Shapiro delay,
812: producing a selection of such inclinations among the systems with
813: measured mass. This should not be too strong in the
814: sample we selected, systems with the timing precision of
815: PSR~J1909$-$3744 and PSR~J0437$-$4715 can have their Shapiro delays
816: measured almost irrespective of their orbital inclination}.
817:
818: If the low mass functions of the massive binaries are not a result of
819: systematically low orbital inclinations, they can only be due to
820: systematically small companion masses. This implies that in the
821: majority of these systems the pulsar masses are significantly larger
822: than in the lighter binaries.
823:
824: \begin{figure}
825: \epsscale{1.18}
826: \plotone{fig6.eps}
827: \caption{\label{fig:psr_masses} Probability distribution functions
828: (pdfs) for the eccentric MSPs binaries in GCs (see also
829: Fig.~\ref{fig:binaries}). The mass pdfs of the MSPs in the less
830: massive binaries (those with $M < 2 M_{\sun}$) are represented by the
831: dashed curves. The distribution of masses is much broader than is
832: found for the components of DNS systems. Four out of a total
833: of seven systems seem to be significantly more massive than the most
834: massive NS in DNS systems, PSR~B1913+16.}
835: \end{figure}
836:
837: \subsection{Implications for the equation of state of dense matter}
838: \label{sec:massives}
839:
840: Because there is no physically plausible reason to assume that most
841: massive binaries have a small $i$ we now assume that the
842: probability density of $\cos i$ is constant. We use this to calculate the
843: mass pdfs from $\dot{\omega}$ as described in \S~\ref{sec:masses}
844: for all the pulsars in Fig.~\ref{fig:binaries}. The pdfs are displayed
845: graphically in Fig. \ref{fig:psr_masses}. If the probability that
846: $1.2\,M_{\sun}\,<\,M_p\,<\,1.44\,M_{\sun}$ is small
847: (as is the case for Terzan 5 I and J, M5B and NGC~6440B), it is a
848: direct indication that the required orbital inclination ranges are
849: very narrow (as shown in Fig.~\ref{fig:binaries}) and therefore
850: unlikely under the present assumption. For M5B, this probability is only
851: 0.77\%, while for NGC~6440B it is even smaller, only 0.10\%.
852: Multiplying such probabilities for all the massive MSPs in GCs, we
853: obtain a composite probability of $5.3 \times 10^{-9}$ that all
854: have masses between 1.2 and $1.44\,M_{\sun}$. It is therefore very
855: likely that some of these NSs are significantly more massive.
856:
857: As discussed above in \S \ref{sec:masses}, for M5B there is a 95\%
858: probability that the pulsar is more massive
859: than $1.72\,M_{\sun}$. This would exclude a third of the equations
860: of state considered in Lattimer \& Prakash (2007). However, in this
861: respect NGC~6440B should be far more constraining; there are 99 and 95\%
862: probabilities that the mass of that pulsar is $>$ 2.01 and
863: $2.36\,M_{\sun}$ respectively. We consider the M5B result to
864: be more secure: the non-detection of its companion (\S \ref{sec:optical})
865: almost guarantees that its $\dot{\omega}$ is purely relativistic (\S
866: \ref{sec:relativistic}).
867:
868: There are two MSPs listed in Table~\ref{tab:msp_masses} for which
869: large masses have already been determined, PSR~J0437$-$4715 and
870: PSR~J1903+0327. The latter in particular
871: has the potential for a precise, unambiguous measurement of a large
872: pulsar mass in the very near future. At the moment, it has not been
873: confirmed whether its $\dot{\omega}$ is relativistic or not, although
874: it is likely to be so. These results strengthen the case for the
875: existence of massive NSs.
876:
877: \section{Formation of massive neutron stars}
878:
879: The MSP mass estimates in Table~\ref{tab:msp_masses} and the
880: mass pdfs in Fig.~\ref{fig:psr_masses}, especially those of M5B and
881: NGC 6440B, suggest that the distribution of MSP masses could span a
882: factor of 2, a situation that is completely different to that found
883: for the components of DNSs. NGC~1851A and the MSPs in the ``light''
884: ($M\,<\,2\,M_{\sun}$) binaries have masses smaller than
885: $1.5\,M_{\sun}$, i.e., they are not significantly more massive than
886: mildly recycled NSs, despite having spin frequencies of
887: hundreds of Hz. In particular, the case of M28C demonstrated that if
888: all NSs start with $M_P > 1.2 M_{\sun}$, then some MSPs
889: can be recycled by accreting $<\,0.15\,M_{\sun}$ from their companions.
890: At the other end of the distribution, NGC~6440B could be twice as
891: massive.
892:
893: It could be that MSPs were born with this wide range of masses.
894: Hydrodynamical core collapse simulations \cite{tww96,bok+08} indicate
895: that stars below $\sim 18\,M_{\sun}$ form $\sim 1.20-1.35\,M_{\sun}$
896: NSs (such as 47~Tuc~H, M28C and NGC~1851A), while stars with
897: masses between $18 - 20\, M_{\sun}$ form $1.8\,M_{\sun}$ NSs (similar
898: to PSR~J0437$-$4715, PSR~J1903+0327, Terzan 5 I, J and M5B).
899: Above $20\,M_{\sun}$, stars experience partial fall-back of material
900: immediately after the supernova that can significantly increase the
901: mass of the stellar remnant, making it either a super-massive NS (like
902: NGC~6440B) or a black hole.
903:
904: This possibility raises the question of why such massive NSs, while
905: representing about half of the MSPs in Table~\ref{tab:msp_masses}, have
906: not been found among the 9 known DNS systems. Most of the secondary NSs
907: in DNS systems have masses between 1.2 and $1.3\,M_{\sun}$, and recently
908: van den Heuvel (2007)\nocite{heu07} suggested that these were formed by
909: electron capture (EC) supernovae. The nine primary NSs
910: in DNS systems were still likely formed in normal (iron core
911: collapse) supernovae. The predicted percentage of massive NSs is quite
912: small, and with only 9 known DNS we are unlikely to see any massive NSs
913: as the primary \cite{bok+08}, a situation that is very different from
914: what is derived for the MSPs.
915:
916: If the extra mass of some MSPs were instead acquired during the long
917: accretion episodes that recycled them, we can explain naturally why
918: we only see massive NSs as MSPs (not just in
919: GCs, but also in the Galaxy, e.g. Verbiest et al. 2008, Champion
920: et al. 2008) but not in DNS systems.
921:
922: \section{Conclusion}
923:
924: We have measured the positions and proper motions of M5A and B.
925: This has allowed a detailed search for the companion of M5B. However,
926: no object was detected within $2.5 \sigma$ of the position of M5B
927: to a magnitude limit of 26-26.5, indicating that its companion is
928: either a low-mass MS star or a WD. We have measured the rate of
929: advance of periastron for this binary system, concluding that it is
930: very likely due solely to the effects of general relativity. In this
931: case, the total mass of the binary is $2.29 \pm 0.17 M_{\sun}$,
932: similar to the total masses of Terzan~5~I and J. Like those pulsars
933: and NGC~6440B, the relatively low mass function for M5B indicates that
934: most of the system mass is likely to be in the pulsar, which we
935: estimate to be $2.08 \pm 0.19 \,M_{\sun}$
936: (1 $\sigma$). There is a 95\% probability that the mass of this pulsar
937: is above $1.72\,M_{\sun}$. If confirmed, this would exclude
938: about a third of the equations of state that are now accepted as possible
939: descriptions of the bulk properties of super-dense matter.
940:
941: Together with other recent results, the large mass derived for M5B
942: suggests that MSPs have a very broad mass distribution; half
943: of these objects seem to be significantly more massive than
944: $1.44\,M_{\sun}$. It is likely that all NSs began with the a narrow
945: mass range like that found in DNS. They then accretted different
946: amounts of matter (in some cases as much as their starting mass)
947: during their evolution to the MSP phase.
948:
949: \acknowledgements
950:
951: We thank S. M. Ransom and I. H. Stairs for many of the L-band
952: observations made since 2001, Patrick Lazarus and Melissa Ilardo for
953: help with data reduction, Chris Salter, Marten van Kerkwijk and the
954: referee, Matthew Bailes, for their many constructive suggestions.
955: J.W.T.H. thanks NSERC and the Canadian Space Agency for a postdoctoral
956: fellowship and supplement respectively. The Arecibo Observatory, a
957: facility of the National Astronomy and Ionosphere Center, is operated
958: by Cornell University under a cooperative agreement with the National
959: Science Foundation. The Borg was funded by a New Opportunities Grant
960: from the Canada Foundation for Innovation. This paper makes use of
961: data obtained from the Isaac Newton Group Archive which is maintained
962: as part of the CASU Astronomical Data Centre at the Institute of
963: Astronomy, Cambridge.
964:
965: \begin{thebibliography}{}
966:
967: \bibitem[{Anderson} {et al.} 1997]{awkp97}
968: {Anderson}, S.~B., {Wolszczan}, A., {Kulkarni}, S.~R., \& {Prince}, T.~A. 1997,
969: \apj, 482, 870
970:
971: \bibitem[{Bassa} {et al.} 2006]{bkkv06}
972: {Bassa}, C.~G., {van Kerkwijk}, M.~H., {Koester}, D., \& {Verbunt}, F. 2006,
973: \astap, 456, 295
974:
975: \bibitem[B\'egin {et al.} 2008]{brf+08}
976: B\'egin, S., Ransom, S.~M., Freire, P.~C.~C., Stairs, I.~H., Hessels, J.~W.~T.,
977: Katz, J., Kaspi, V., \& Camilo, F. 2008, \apj.
978: \newblock in preparation
979:
980: \bibitem[Bell {et al.} 1995]{bbs+95}
981: Bell, J.~F., Bessell, M.~S., Stappers, B.~W., Bailes, M., \& Kaspi, V.~M. 1995,
982: \apjlett, 447, L117
983:
984: \bibitem[Belczynski et al. 2008]{bok+08}
985: Belczynski, K, O'Shaughnessy, R., Kalogera, V., Rasio, F., Taam, R. \&
986: Bulik, T. 2008, arXiv:0712.1036v1
987:
988: \bibitem[{Bergbusch} \& {VandenBerg} 1992]{bergvand92} {Bergbusch},
989: P.~A. \& {Vandenberg}, D.~A. 1992, \apjs, 81, 163
990:
991: \bibitem[{Champion} {et al.} 2008]{crl+08}
992: {Champion}, D. {et al.} 2008,
993: \newblock private communication
994:
995: \bibitem[Damour \& Deruelle 1985]{dd85}
996: Damour, T. \& Deruelle, N. 1985, Ann. Inst. H. Poincar\'e (Physique
997: Th\'eorique), 43, 107
998:
999: \bibitem[Damour \& Deruelle 1986]{dd86}
1000: Damour, T. \& Deruelle, N. 1986, Ann. Inst. H. Poincar\'e (Physique
1001: Th\'eorique), 44, 263
1002:
1003: \bibitem[{Dinescu}, {Girard}, \& {van Altena} 1999]{dga99}
1004: {Dinescu}, D., {Girard}, T., \& {van Altena}, W. 1999, \apj, 117, 1792
1005:
1006: \bibitem[Dowd, Sisk, \& Hagen 2000]{dsh00}
1007: Dowd, A., Sisk, W., \& Hagen, J. 2000, in { Pulsar Astronomy - 2000 and Beyond,
1008: {IAU} Colloquium 177}, ed.\ M. Kramer, N. Wex, \& R. Wielebinski, (San
1009: Francisco: Astronomical Society of the Pacific), 275
1010:
1011: \bibitem[{Efron} \& {Tibshirani} 1993]{et93}
1012: {Efron}, B. \& {Tibshirani}, R.~J. 1993, { An Introduction to the Bootstrap},
1013: (New York: Chapman \& Hall)
1014:
1015: \bibitem[{Faulkner} {et al.} 2005]{fkl+05}
1016: {Faulkner}, A.~J. {et al.} 2005, \apjlett, 618, L119
1017:
1018: \bibitem[Ferraro et al. 2001]{fpds01} Ferraro, F.~R.,
1019: Possenti, A., D'Amico, N., \& Sabbi, E.\ 2001, \apjl, 561, L93
1020:
1021: \bibitem[{Freire} {et al.} 2003]{fck+03}
1022: {Freire}, P.~C., {Camilo}, F., {Kramer}, M., {Lorimer}, D.~R., {Lyne}, A.~G.,
1023: {Manchester}, R.~N., \& {D'Amico}, N. 2003, \mnras, 340, 1359
1024:
1025: \bibitem[{Freire} {et al.} 2004]{fgri04}
1026: {Freire}, P.~C., {Gupta}, Y., {Ransom}, S.~M., \& {Ishwara-Chandra}, C.~H.
1027: 2004, \apjlett, 606, L53
1028:
1029: \bibitem[{Freire}, {Ransom}, \& {Gupta} 2007]{frg07}
1030: {Freire}, P.~C.~C., {Ransom}, S.~M., \& {Gupta}, Y. 2007, \apj, 662, 1177
1031:
1032: \bibitem[{Freire} {et al.} 2008]{frb+08}
1033: {Freire}, P. C.~C., {Ransom}, S.~M., {B\'egin}, S., {Stairs}, I.~H., {Hessels},
1034: J. W.~T., {Frey}, L.~H., \& {Camilo}, F. 2008, \apj, 675, 670
1035:
1036: \bibitem[Harris 1996]{har96} Harris, W.~E. 1996, \aj, 112, 1487.
1037: \newblock See
1038: http://www.physics.mcmaster.ca/resources/globular.html for updated
1039: version of table.
1040:
1041: \bibitem[{Hessels} {et al.} 2007]{hrs+07}
1042: {Hessels}, J.~W.~T., {Ransom}, S.~M., {Stairs}, I.~H., {Kaspi}, V.~M., \&
1043: {Freire}, P.~C.~C. 2007, \apj, 670, 363
1044:
1045: \bibitem[van den Heuvel 2007]{heu07} van den Heuvel,
1046: E.~P.~J.\ 2007, American Institute of Physics Conference Series, 924, 598
1047:
1048: \bibitem[{Hotan} {et al.} 2006]{hbo06}
1049: {Hotan}, A.~W., {Bailes}, M. \& {Ord}, S.~M. 2006, \mnras, 369, 1502
1050:
1051: \bibitem[Jacoby {et al.} 2005]{jhb+05}
1052: Jacoby, B.~A., Hotan, A., Bailes, M., Ord, S., \& Kuklarni, S.~R. 2005,
1053: \apjlett, 629, L113
1054:
1055: \bibitem[Lai, Bildsten, \& Kaspi 1995]{lbk95}
1056: Lai, D., Bildsten, L., \& Kaspi, V.~M. 1995, \apj, 452, 819
1057:
1058: \bibitem[{Lattimer} \& {Prakash} 2007]{lp07}
1059: {Lattimer}, J.~M. \& {Prakash}, M. 2007, \physrep, 442, 109
1060:
1061: \bibitem[Livio \& Pringle 1998]{lp98} Livio, M., \&
1062: Pringle, J.~E.\ 1998, \apj, 505, 339
1063:
1064: \bibitem[Nice et al. 2005]{nss+05} Nice, D.~J., Splaver,
1065: E.~M., Stairs, I.~H., L{\"o}hmer, O., Jessner, A., Kramer, M., \& Cordes,
1066: J.~M.\ 2005, \apj, 634, 1242
1067:
1068: \bibitem[Nice 2007]{nice07} Nice, D.~J., 2007, to appear in the proceedings
1069: of "40 Years of Pulsars: Millisecond Pulsars, Magnetars, and More",
1070: August 12-17, 2007, McGill University, Montreal, Canada, C. Bassa, Z. Wang,
1071: A. Cumming and V. Kaspi Eds.
1072:
1073: \bibitem[Odenkirchen et al. 1997]{obgt97} Odenkirchen, M.,
1074: Brosche, P., Geffert, M., \& Tucholke, H.-J.\ 1997, New Astronomy, 2, 477
1075:
1076: \bibitem[Press {et al.} 1992]{ptvf92}
1077: Press, W.~H., Teukolsky, S.~A., Vetterling, W.~T., \& Flannery, B.~P. 1992, {
1078: Numerical Recipes: {T}he Art of Scientific Computing, 2$^{nd}$ edition},
1079: (Cambridge: Cambridge University Press)
1080:
1081: \bibitem[{Ransom} {et al.} 2005]{rhs+05}
1082: {Ransom}, S.~M., {Hessels}, J.~W.~T., {Stairs}, I.~H., {Freire}, P.~C.~C.,
1083: {Camilo}, F., {Kaspi}, V.~M., \& {Kaplan}, D.~L. 2005, Science, 307, 892
1084:
1085: \bibitem[Rasio \& Heggie 1995]{rh95}
1086: Rasio, F.~R. \& Heggie, D.~C. 1995, \apj, 445, L133
1087:
1088: \bibitem[{Sandquist} {et al.} 1996]{sandea96} {Sandquist}, E.~L.,
1089: {Bolte}, M., {Stetson}, P.~B. \& {Hesser}, J.~E. 1996, \apj, 470, 910
1090:
1091: \bibitem[Scholz et al. 1996]{soh+96} Scholz, R.-D.,
1092: Odenkirchen, M., Hirte, S., Irwin, M.~J., Borngen, F.,
1093: \& Ziener, R.\ 1996, \mnras, 278, 251
1094:
1095: \bibitem[Splaver {et al.} 2002]{sna+02}
1096: Splaver, E.~M., Nice, D.~J., Arzoumanian, Z., Camilo, F., Lyne, A.~G., \&
1097: Stairs, I.~H. 2002, \apj, 581, 509
1098:
1099: \bibitem[Standish 1998]{sta98b}
1100: Standish, E.~M. 1998, { JPL Planetary and Lunar Ephemerides, DE405/LE405, Memo
1101: IOM 312.F-98-048}, (Pasadena: JPL).
1102: \newblock http://ssd.jpl.nasa.gov/iau-comm4/de405iom/de405iom.pdf
1103:
1104: \bibitem[Taylor 1992]{tay92}
1105: Taylor, J.~H. 1992, Philosophical Transactions of the Royal Society of London,
1106: A, 341, 117
1107:
1108: \bibitem[Timmes et al. 1996]{tww96} Timmes, F.~X., Woosley,
1109: S.~E., \& Weaver, T.~A.\ 1996, \apj, 457, 834
1110:
1111: \bibitem[Verbiest et al. 2008]{vbs+08} Verbiest, J. P. W., Bailes, M.,
1112: van Straten, W. Hobbs, G. B., Edwards, R. T., Manchester, R. N.
1113: Bhat, N. D. R., Sarkissian, J. M., Jacoby, B. A. and Kulkarni, S. R. 2008,
1114: Astrophysical Journal, in press, arXiv:0801.2589
1115:
1116: \bibitem[Webbink 1985]{web85}
1117: Webbink, R.~F. 1985, in { Dynamics of Star Clusters, {IAU} {S}ymposium {N}o.
1118: 113}, ed.\ J. Goodman \& P. Hut, (Dordrecht: Reidel), 541
1119:
1120: \bibitem[{Weisberg} \& {Taylor} 2003]{wt03}
1121: {Weisberg}, J.~M. \& {Taylor}, J.~H. 2003, in { Radio Pulsars}, ed.\ M. Bailes,
1122: D.~J. Nice, \& S.E. Thorsett, (San Francisco: Astronomical Society of the
1123: Pacific), 93
1124:
1125: \bibitem[Wex 1998]{wex98}
1126: Wex, N. 1998, \mnras, 298, 67
1127:
1128: \bibitem[Wolszczan {et al.} 1989]{wakp89}
1129: Wolszczan, A., Anderson, S., Kulkarni, S., \& Prince, T. 1989.
1130: \newblock IAU circular 4880
1131:
1132: \bibitem[{Zacharias} {et al.} 2004]{zachea04} {Zacharias}, N.,
1133: {Urban}, S.~E., {Zacharias}, M.~I., {Wycoff}, G.~L., {Hall}, D.~M.,
1134: {Monet}, D.~G. \& {Rafferty}, T.~J. 2004, \aj, 127, 3043
1135:
1136: \end{thebibliography}
1137:
1138: \typeout{get arXiv to do 4 passes: Label(s) may have changed. Rerun}
1139:
1140: \end{document}
1141: