1: %\documentclass[12pt,preprint]{aastex}
2:
3: %% This emulates the ApJ
4: \documentclass{emulateapj}
5:
6: \usepackage{lscape}
7: %\usepackage{apjfonts}
8: %\usepackage{deluxetable}
9: %\usepackage{emulateapj5}
10: %\usepackage{psfig}
11: %\def\ni{\noindent}
12: %\def\ang{${\rm \AA}$~}
13: %\def\Halpha{H$\alpha$~}
14: %\def\about{$\sim$}
15: \def\arcsec{$\,^{\prime\prime}$~}
16: \def\arcmin{$\,^\prime$~}
17: \def\deg{$^{\circ}$~}
18: \def\erg/cm2sec{ergs~cm$^{-2}$~s$^{-1}$}
19: \def\ergcm2{ergs~cm$^{-2}$}
20: \def\sqdeg{deg$^{-2}$~}
21: \def\mdot{$\dot{m}$~}
22: \def\X{$\times$}
23: \def\Fx{F$_x$~}
24: \def\Fv{F$_v$~}
25: \def\FxFv{{F$_x$/{F$_v$}}~}
26: \def\Lx{L$_x$~}
27:
28:
29: % abbreviations for journals for Science refs. follows here:
30: \def\apj{ ApJ}
31: \def\aap{ A\&A}
32: \def\mn{MNRAS}
33: \def\pasp{PASP}
34: \def\araa{ARAA}
35: \def\aj{AJ}
36: \def\apjs{ApJ Supp}
37:
38: \def\Chandra{${\it Chandra}$\ }
39: \def\HST{${\it HST}$\ }
40:
41:
42: \newcommand{\Msun}{\ifmmode {M_{\odot}}\else${M_{\odot}}$\fi}
43: \newcommand{\Lsun}{\ifmmode {L_{\odot}}\else${L_{\odot}}$\fi}
44: \newcommand{\Rsun}{\ifmmode {R_{\odot}}\else${R_{\odot}}$\fi}
45: \newcommand\HALP{{\rm H}\alpha}
46:
47: \newcommand\VI{V_{555}-I_{814}}
48: \newcommand\HR{{\rm H}\alpha-R_{675}}
49: \newcommand\jfh{$\,^{\rm h}$}
50: \newcommand\jfm{$\,^{\rm m}$}
51: \newcommand\jfs{$\,^{\rm s}$}
52:
53: \shorttitle{New Quiescent LMXBs in 47 Tuc}
54: \shortauthors{Heinke et al.}
55:
56:
57: %----------------------------------------------------------------------
58: \begin{document}
59: \title{Three Additional Quiescent Low-Mass X-ray Binary Candidates in 47 Tucanae}
60:
61: \author{C.~O.~Heinke\altaffilmark{1,2,3},
62: J.~E.~Grindlay\altaffilmark{1}, P.~D.~Edmonds\altaffilmark{1}}
63: \altaffiltext{1}{Harvard-Smithsonian Center for Astrophysics,
64: 60 Garden Street, Cambridge, MA 02138;\\
65: cheinke@cfa.harvard.edu, josh@cfa.harvard.edu, pedmonds@cfa.harvard.edu}
66:
67: \altaffiltext{2}{Northwestern University, Dept. of Physics \&
68: Astronomy, 2145 Sheridan Rd., Evanston, IL 60208}
69:
70: \altaffiltext{3}{Lindheimer Postdoctoral Fellow}
71:
72: %\slugcomment{To be submitted to ApJ...}
73:
74: \begin{abstract}
75: We identify through their X-ray spectra one certain (W37) and two probable
76: (W17 and X4) quiescent low-mass X-ray binaries (qLMXBs) containing
77: neutron stars in a long \Chandra X-ray exposure of the globular
78: cluster 47 Tucanae, in addition to the two previously known qLMXBs.
79: W37's spectrum is dominated by a blackbody-like component consistent
80: with radiation from the hydrogen atmosphere of a 10 km neutron star.
81: W37's lightcurve shows strong X-ray variability which we attribute to
82: variations in
83: its absorbing column depth, and eclipses with a probable 3.087 hour period. For
84: most of our exposures, W37's blackbody-like emission (assumed to be from the
85: neutron star surface) is almost completely obscured, yet some soft
86: X-rays (of uncertain origin) remain.
87: Two additional candidates, W17 and X4, present X-ray spectra dominated
88: by a harder
89: component, fit by a power-law of photon index $\sim$1.6-3.
90: An additional soft component is required for both W17 and X4, which
91: can be fit with a 10 km hydrogen-atmosphere neutron star model.
92: X4 shows significant variability, which
93: may arise from either its power-law or hydrogen-atmosphere
94: spectral component. Both W17 and X4
95: show rather low X-ray luminosities, $L_X$(0.5-10 keV)
96: $\sim5\times10^{31}$ ergs
97: s$^{-1}$. All three candidate qLMXBs would be
98: difficult to identify in other globular clusters, suggesting an
99: additional reservoir of fainter qLMXBs in globular clusters that may be of
100: similar numbers as the group of previously identified objects. The
101: number of millisecond pulsars inferred to exist in 47 Tuc is less
102: than 10 times larger than the number of qLMXBs in 47 Tuc, indicating
103: that for typical inferred lifetimes of 10 and 1 Gyr respectively,
104: their birthrates are comparable.
105: \end{abstract}
106:
107: \keywords{
108: accretion disks ---
109: binaries: close, eclipsing ---
110: binaries : X-rays ---
111: globular clusters: individual (NGC 104) ---
112: stars: neutron
113: }
114:
115: \maketitle
116: %----------------------------------------------------------------------
117:
118: \section{Introduction}\label{s:intro}
119:
120: Several soft X-ray transients, identified in outburst to be accreting
121: neutron star (NS)
122: systems, have been observed in quiescence \citep[see][]{Campana98a,
123: Rutledge02b}. These systems show
124: soft spectra, generally consisting of a thermal,
125: blackbody-like component, and in most cases a harder component extending to
126: higher energies, usually fit with a power-law of photon index 1-2.
127: The nature of their X-ray emission remains unsolved. \citet[see
128: also Campana et al. 1998]{Brown98}
129: advanced the idea that the soft thermal component seen in these
130: systems can be explained by the release, over long timescales, of heat
131: injected into the deep crust by pycnonuclear reactions driven during
132: accretion (the ``deep crustal heating'' model). This scenario
133: accurately predicts the quiescent luminosity of some qLMXBs, based on their
134: outburst history \citep{Rutledge00,Rutledge01b}, and
135: must operate at some level. (Recent observations of low quiescent
136: luminosities in some qLMXBs
137: \citep{Colpi01,Campana02,Yakovlev03,Wijnands04}
138: may require nonstandard cooling processes.)
139:
140: However, the deep crustal heating model cannot explain the hard power-law
141: component, which is attributed to continued accretion and/or a
142: shock from a pulsar wind \citep[see][]{Campana98a,Menou01}.
143: Continued accretion has also been suggested as an explanation for the
144: thermal component, as the radiation spectrum from matter accreting radially
145: onto a neutron star should be similar to that from the ionized
146: hydrogen atmosphere of a hot non-accreting neutron star
147: \citep{Zampieri95}.
148: Models without any continuing accretion have difficulty explaining the
149: short-timescale
150: ($\sim10^4$ s) variability observed from Aquila X-1 and Cen X-4
151: \citep{Rutledge02b, Campana03}, and the days-timescale variability
152: observed from Cen X-4
153: \citep{Campana97} and other qLMXBs. However, many qLMXBs in globular
154: clusters show neither significant variability nor an additional
155: power-law spectral component \citep[hereafter HGL03]{Heinke03a}.
156: Understanding the
157: emission process will be critical for modeling the observed spectrum
158: and deriving fundamental physical parameters, such as mass, radius,
159: and magnetic field \citep{Lattimer04,Brown98}.
160:
161: Globular clusters are
162: overabundant in accreting NS systems compared to the field, with
163: thirty-eight probable
164: qLMXBs and active LMXBs known so far \citep{Heinke03d}. The
165: well-studied globular cluster 47
166: Tucanae (NGC 104; hereafter 47 Tuc) is ideal for X-ray studies of its
167: binary populations \citep{Grindlay01a} due to its
168: close distance \citep[4.85$\pm0.18$ kpc, ][]{Gratton03} and low
169: reddening \citep[$E(B-V)=0.024\pm0.004$, ][]{Gratton03}. Two qLMXBs
170: in 47 Tuc,
171: X5 and X7 \citep[originally detected by ROSAT,][]{Verbunt98},
172: have been spectrally identified by HGL03. HGL03 constrained the range
173: of mass and radius space for X7
174: using several different possible assumptions about its spectrum, and
175: showed that X5 is an eclipsing system with an 8.67 hour period and
176: strong dipping activity.
177:
178: We have shown \citep[HGL03,][]{Heinke03d} that the identified qLMXBs in 47
179: Tuc, and in globular clusters generally, have little to no hard
180: power-law component in contrast to the best-studied field systems,
181: Cen X-4 and Aql X-1 \citep{Rutledge01b, Rutledge02b}.
182: The exceptions thus far have been the transient LMXB in NGC 6440, which
183: entered outburst 13 months after the \Chandra\ observation
184: \citep{intZand01}, and the transient LMXB in Terzan 5
185: \citep{Wijnands03b}, both relatively bright qLMXBs.
186: If the strength of the power-law component (and short-term variability)
187: measures continuing low-level
188: accretion \citep{Campana98a,Rutledge02a}, then the absence of these
189: indicators in globular cluster systems might be taken to indicate
190: extremely low levels of accretion
191: activity. Recently, \citet{Jonker04} noted that the relative
192: strength of the spectral component fit with power-laws in quiescent NS
193: transients seems to increase for X-ray luminosities significantly
194: larger or smaller than $L_X\sim10^{33}$ ergs s$^{-1}$, where its
195: relative strength is often at a minimum \citep[but cf.][]{Wijnands03b}.
196: They suggested that this spectral component arises from
197: different processes (accretion vs.~pulsar wind shock?) in systems with
198: X-ray luminosities above or below $10^{33}$ ergs s$^{-1}$.
199:
200: In late 2002, we obtained deeper \Chandra observations of 47 Tuc
201: \citep{Heinke04a}, allowing detailed spectral modeling of X-ray sources
202: as faint as $L_X\sim10^{31}$ ergs s$^{-1}$. Most sources were found
203: to be consistent with powerlaw or thermal plasma models, but some
204: sources required more complicated models. The highly variable X-ray
205: source W37 (=CXOGLB J002404.9-720451) showed a steep spectrum well-described
206: by a blackbody, and the X-ray sources W17 (=CXOGLB J002408.3-720431) and
207: X4 (=W125, or CXOGLB J002353.9-720350) showed two-component spectra in which
208: the soft component could be well-represented by a model of a hydrogen
209: atmosphere appropriate for a neutron star. X4 was first identified in
210: ROSAT HRI data \citep{Verbunt98}, while W17 and W37 are located in
211: areas too crowded to
212: have been identified in lower-resolution ROSAT data at their current fluxes.
213: None of these three
214: have been positively identified with optical counterparts of any
215: category \citep{Edmonds03a, Edmonds03b}. In this paper, we study
216: these three strong qLMXB candidates in detail.
217: We describe the observations and our reduction in \S~\ref{s:obs},
218: the timing analysis in \S~\ref{s:timing},
219: and the spectral analyses in \S~\ref{s:spec}. Discussion and conclusions are
220: in \S~\ref{s:disc}.
221:
222: \section{Observations and Analysis}\label{s:obs}
223:
224: The data used in this paper are from the 2000 and 2002
225: \Chandra
226: observations of the globular cluster 47 Tuc. Both sets of
227: observations and their initial reduction are described in
228: detail in \citet{Heinke04a}; prior analyses of the 2000 dataset
229: are described in \citet{Grindlay01a,Grindlay02} and HGL03. Both
230: observations interleaved short exposures
231: using subarrays (chosen to reduce pileup, which is not a significant
232: issue for the three sources in this paper) with longer exposures. The
233: observations are summarized in Table~\ref{tab:obs}; we note that
234: OBS\_ID 3385
235: suffered strongly increased background, so we do not use that short
236: observation. We reprocessed (using CIAO 3.0) both the 2000
237: and 2002 observations to remove the 0\farcs25 pixel
238: randomization added in standard processing, implement the CTI
239: correction algorithm on the 2000 data, and use updated (Feb. 2004,
240: CALDB v. 2.26)
241: calibration files. We did not remove events
242: flagged as cosmic-ray afterglows by the standard processing.
243: 290 sources
244: were detected using the WAVDETECT routine in the 0.3-6 and 0.3-2 keV
245: energy bands, and ten additional sources (missed by WAVDETECT due to
246: crowding) were added to the source list by hand.
247:
248: We used the ACIS\_EXTRACT routine \citep{Broos02} from the Tools for
249: X-ray Analysis
250: (TARA\footnote{http://www.astro.psu.edu/xray/docs/TARA/}) to construct
251: polygonal extraction regions generally chosen to match the 90\%
252: encircled energy (at 1.5 keV) from the \Chandra point-spread
253: function (PSF).
254: We extracted source spectra and background spectra from nearby
255: source-free regions, and corrected the ancillary response function
256: (ARF) for the energy-varying fraction of the PSF enclosed by the
257: extraction region. We used the Jan.~2004 release of the ACIS
258: contamination model to correct for the hydrocarbon buildup on the
259: detectors \citep{Marshall04}. We adjusted all event times to the
260: solar system barycenter using
261: satellite orbit files provided by the \Chandra X-ray Center.
262:
263: \subsection{Timing Analysis}\label{s:timing}
264:
265: Kolmogorov-Smirnov (KS) tests on the 2002 event arrival times from W37
266: and X4
267: confirm that both are clearly variable on short
268: timescales, with KS probabilities of constancy reaching $10^{-47}$ and
269: $3\times10^{-4}$ for the first 2002 observations of W37 and X4
270: respectively. W17 shows only marginal evidence for variability, with
271: KS probabilities reaching only 0.04. X4 falls on a chip gap in the
272: 2000 data, and W37 is much fainter in the 2000 data. W37 receives a
273: photon flux over five
274: times higher in the first 2002 observation than in any other
275: observation, and 75\% of the 2002 photons from W37 arrive in the first
276: quarter of the data. X4 appears to have dimmed slightly between the first
277: and fourth long 2002 observations (at 99\% conf.).
278:
279: We show the
280: first (and most variable) sections of the 2002
281: lightcurves for W37 and X4
282: in Figures~\ref{fig:w37lc} and \ref{fig:x4lc}. Generating a power
283: spectrum for W37 (using XRONOS,
284: \footnote{http://heasarc.gsfc.nasa.gov/docs/xanadu/xronos/xronos.html})
285: gives a possible peak around 11000 seconds along with strong red
286: noise. An epoch folding search using all the 2002 observations finds
287: a periodicity at 11112.5 ($\pm15$)
288: seconds. Careful inspection of W37's lightcurves reveals
289: likely eclipses, indicated in Figures~\ref{fig:w37lc} and
290: \ref{fig:w37fold} (left; a folded lightcurve), consistent with the
291: 11112.5 second
292: period. The two clearest eclipses are separated by 6.2 hours, so we
293: also include a lightcurve folded on twice our preferred period
294: (Figure~\ref{fig:w37fold}, right), showing two apparent eclipses. This
295: indicates that our preferred period is probably correct, but the low
296: count rate of the data may allow for alternative solutions.
297: The folded lightcurve suggests eclipses $550\pm200$ seconds in length.
298:
299: We note
300: that W37's lightcurve is similar to that of X5,
301: presented in HGL03 and Heinke et al.~(2005, in prep), suggesting that
302: the strong variability seen in W37 is also probably due to a varying column
303: of cold gas associated with an accretion disk viewed nearly
304: edge-on. However, W37 may show residual emission during eclipses,
305: which is not seen for X5 (HGL03). The period indicates a companion
306: mass of order 0.34 \Msun, if the companion has a density appropriate
307: for lower main sequence stars \citep[][eq. 4.9]{Frank92}. Since
308: these companions are sometimes bloated
309: \citep{King99,Podsiadlowski02,Orosz03,Kaluzny03b}, the true mass
310: may be smaller. Assuming a 1.4 \Msun\ neutron star, the orbital separation
311: would be $\sim9\times10^{10}$ cm.
312:
313:
314: We calculate the hardness ratio for W37 as the ratio of the counts
315: detected in the 1-6 keV and 0.3-1 keV bands (this choice divides the
316: detected counts roughly evenly), and plot this ratio in the lower
317: panel of Figure~\ref{fig:w37lc}. Comparison with the upper panel shows that
318: decreases in the observed flux correlate well with
319: increasing hardness. We confirm this with a Spearman rank-order
320: correlation test \citep{Press92} comparing the binned (1200 second)
321: lightcurve with the similarly binned hardness ratio. This gives
322: $r_s$=-0.446, a negative correlation, with chance probability
323: $5.6\times10^{-4}$.
324: This is a strong sign that the variation is due to
325: changing photoelectric absorption. We divide W37's long 2002
326: observations into high and low flux states (e.g., the high flux state
327: portions indicated in Figure~\ref{fig:w37lc}), which we use to extract
328: spectra.
329:
330: X4's total and hardness ratio lightcurve for OBS\_IDs 2735 and 3384
331: are shown in Figure~\ref{fig:x4lc}. No
332: periodicities are apparent in X4's lightcurve or power spectra.
333: X4's spectrum appears to harden as it brightens, in contrast to
334: W37, which indicates the variability is not due to increasing
335: photoelectric absorption. We confirm this with a Spearman rank-order
336: correlation test on OBS\_ID 2735, finding $r_s$=0.444 and chance probability
337: $6.0\times10^{-4}$ for the binned (1200 second) lightcurve and hardness
338: ratio. We extract high
339: and low portions of X4's lightcurve from the first observation for
340: spectral fitting, which are marked in Figure~\ref{fig:x4lc}.
341:
342:
343: \subsection{Spectral Analysis}\label{s:spec}
344:
345: In \citet{Heinke04a}, we identify W37, X4, and W17 as
346: possible qLMXBs by their agreement with models consisting of a
347: hydrogen-atmosphere model and a power-law, along with their failure to fit
348: single-component models typical of thermal plasma (such as seen in
349: cataclysmic variables and X-ray active binaries). Here, we analyze
350: time-resolved
351: spectra for W37 and X4, and use the merged spectra (2000 and
352: 2002) for W17 (since it shows no variability). We use
353: the XSPEC {\it phabs} model to describe neutral absorbing gas. We use
354: the \citet{Lloyd03} model to describe thermal blackbody-like emission
355: from the hydrogen atmosphere of a neutron star, fixing the redshift
356: to 0.306 (appropriate for a 1.4 \Msun, 10 km neutron star) and take
357: the distance to be 4.85 kpc \citep{Gratton03}. The total number of
358: counts in the 2002 spectra of W37, W17 and X4 were 1277, 909, and
359: 1426, while from the 2000 data 46, 114, and 62 counts were available.
360: We binned spectra with more than 1000 counts with 40 counts/bin,
361: spectra from the 2000 observations or with fewer than 50 counts at 10
362: counts/bin, and other spectra at 20 counts/bin.
363:
364: \subsubsection{W17}\label{s:w17}
365:
366: W17 shows no variability on short or long timescales, so we
367: fit the combined 2002 and combined 2000 datasets simultaneously. We
368: try a variety of absorbed single-component models including blackbody,
369: bremsstrahlung, MEKAL, power-law, and hydrogen atmosphere neutron star
370: models, but do not find good fits with any. The best fit of these, a
371: power-law model, gives $\chi^2_{\nu}=1.36$, for a 2\% null hypothesis
372: probability (nhp). Models consisting of two mekal components also do not
373: produce good fits. Adding a soft component, such as the hydrogen-atmosphere
374: neutron star model of \citet{Lloyd03}, to the power-law model improves
375: the statistics tremendously ($\chi^2_{\nu}=1.05$, nhp=37\%; F-test
376: gives probability of $9\times10^{-5}$ that the extra component is not
377: needed). The inferred radius of the neutron star model,
378: $R=11.8^{+9.7}_{-3.9}$ km, is perfectly consistent with the canonical
379: 1.4 \Msun, 10 km neutron star predictions, although it cannot usefully
380: constrain the neutron star structure. (The inferred radius for a
381: blackbody would be only $1.2^{+0.1}_{-0.4}$ km.) A qLMXB model
382: is thus an excellent explanation for this spectrum, shown in
383: Figure~\ref{fig:w17spec}. The parameters for a
384: fit with a Lloyd H-atmosphere model and power-law are listed in
385: Table~\ref{tab:spec}. The power-law component provides a majority
386: ($65^{+28}_{-22}$\%) of the 0.5-10 keV flux, and the thermal component is
387: quite weak, with unabsorbed X-ray luminosity
388: $L_X(0.5-2.5)=1.7\times10^{31}$ ergs s$^{-1}$ and inferred bolometric
389: luminosity $L_{bol}=6.7\times10^{31}$ ergs s$^{-1}$.
390:
391: \subsubsection{W37}\label{s:w37}
392:
393: We identify pieces of W37's lightcurves at high and
394: low flux levels in the four long observations, with only a low flux
395: level in the second long observation; the (barycentered) times are
396: listed in Table~\ref{tab:add}. We
397: extracted spectra from these periods, as well as from the summed 2000
398: observations, and background spectra from
399: surrounding source-free
400: regions, producing eight spectra of varied quality.
401:
402: The hardening
403: of the spectrum with decreasing flux suggests that variation of the
404: column density is responsible for the spectral variations, and so we
405: allow the gas column to vary between spectral segments. However, no
406: absorbed single-component model was able to describe the eight
407: spectra simultaneously by only varying the absorption and one
408: other model parameter. (Allowing all the parameters to vary freely allows
409: some reasonable fits, but loses physical meaning. For instance, a
410: blackbody with strongly varying normalization, temperature and
411: absorbing column, while a reasonable fit, is difficult to interpret.)
412: No single- or double-temperature thermal plasma model
413: (represented by VMEKAL models in XSPEC, set to cluster abundances)
414: provides a good fit to the brightest portions of the data. The brightest
415: portions of the data are best fit by a blackbody or hydrogen
416: atmosphere neutron star model with varying absorption. If described by
417: a blackbody, the inferred radius is only
418: $1.4\pm0.2$ km in size. This component is well-described by a
419: hydrogen-atmosphere neutron star model with a constant
420: physical radius of order 10 km
421: ($12.3^{+5.8}_{-3.5}$ km) and constant temperature ($82^{+10}_{-9}$ eV).
422: Therefore we infer that W37 is a qLMXB containing a neutron star.
423:
424: The fainter
425: portions show additional soft flux beyond what can be accounted for by
426: an absorbed H-atmosphere (or blackbody) component. This suggests a
427: second component that does not suffer the same absorption as the
428: primary spectral component. We model this component with a separately
429: absorbed power-law, although a variety of models can account for this
430: low-count component equally well (including a 0.6 keV bremsstrahlung
431: spectrum, a
432: 0.1 keV blackbody with inferred 1 km radius, or a 0.13 keV thermal
433: plasma model), and we do not know its physical origin.
434: We hold the faint
435: component, its absorption, and the H-atmosphere model parameters fixed
436: between data portions, and
437: vary only the absorption to the H-atmosphere model. This is the
438: XSPEC model PHABS ( POW + PHABS * HATM ), varying only the second PHABS
439: component between observations. We find a good
440: fit ($\chi^2_{\nu}=1.24$ for 57 degrees of freedom, or dof, and an
441: nhp of 10.6\%). We are unable to impose
442: serious constraints upon this model if additional parameters are
443: allowed to vary, since the faint portions of the data have very few
444: counts. Therefore we cannot usefully constrain any temporal or
445: spectral variability in the faint component.
446: We show part of this fit (only 4
447: of 8 spectra to reduce confusion) in Figure~\ref{fig:w37spec}, and
448: give details of the fit in
449: Tables~\ref{tab:spec} and \ref{tab:add}. The fitted absorption column
450: to the probable neutron star varies by a factor $>100$, decreasing to
451: a value consistent with the cluster column during parts of the
452: first 2002 observation. The faint component
453: is responsible for a very small portion ($2^{+17}_{-1}$\%) of the
454: unabsorbed flux, and produces a photon index of $3^{+5}_{-1}$.
455: The absorption column to the faint
456: component is poorly constrained ($N_H=4.5^{+84}_{-4.5}\times10^{20}$)
457: but consistent with the column to the cluster. The
458: lowest-flux spectrum (from OBS\_ID 2736), dominated by the faint
459: component, is one of the spectra shown in Figure~\ref{fig:w37spec}.
460:
461: \subsubsection{X4}\label{s:x4}
462:
463: X4 varies during the first 2002 observation, but is not clearly
464: variable within other observations. We separately fit the high-flux
465: and low-flux parts of the first 2002 observation (OBS\_ID 2735), and
466: data from each of the other three long 2002 observations, plus one
467: spectrum from the combined 2000
468: observations. We found that the remaining 2002 observations were
469: spectrally indistinguishable, so we combined all 2002 data except
470: OBS\_ID 2735 into one spectrum, and fit the high and low-flux parts of
471: 2735, the rest of the 2002 data, and the combined 2000 data
472: simultaneously. Fits with any single absorbed component, with all
473: parameters allowed to vary, failed to give reasonable fits to the
474: spectra. A power-law fit gave the lowest $\chi^2_{\nu}=1.45$,
475: nhp=3.9\%, if $N_H$ is allowed to vary substantially along with the
476: power-law parameters. However, if $N_H$ is not allowed to vary during
477: the 2002 data (as indicated in \S~\ref{s:timing} above), then
478: $\chi^2_{\nu}$=1.52, nhp=1.9\%. Adding a H-atmosphere model
479: improves the fit substantially (F-statistic=4.38, prob. 1.9\% extra
480: component not required), and results in an inferred radius consistent
481: with the 10 km canonical NS radius. We conclude that X4 is a good
482: candidate for a qLMXB. Alternate models involving two VMEKAL
483: components with all temperatures and normalizations free do not
484: produce good fits.
485:
486: The variability during OBS\_ID 2735 demands that at least one
487: component of the spectrum varies. Varying $N_H$ alone does not produce
488: a good fit ($\chi^2_{\nu}=2.41$, nhp=$1\times10^{-4}$\%). Varying the
489: power-law component does produce a
490: good fit ($\chi^2_{\nu}=1.30$, nhp=11\%; see Table~\ref{tab:spec}).
491: We show this fit for the two components of OBS\_ID 2735
492: (Fig.~\ref{fig:x4spec1}) and for the remainder of the 2002
493: observations, indicating the contributions of the two components
494: (Fig.~\ref{fig:x4spec2}). Varying only the
495: H-atmosphere component's temperature and radius also produces a good fit
496: ($\chi^2_{\nu}=1.35$, nhp=7.3\%; see Table~\ref{tab:spec}).
497: Varying both the
498: power-law and H-atmosphere parameters ($\chi^2_{\nu}=1.21$, nhp=20\%)
499: is only a marginal improvement over varying a single component, and
500: requires correlated swings in numerous parameters (which we assume is
501: less likely than one or two varying parameters). Varying
502: only the power-law normalization also produces a good fit
503: ($\chi^2_{\nu}=1.33$, nhp=7.5\%), with a power-law photon index of
504: 2.2$^{+0.3}_{-0.3}$. Varying the H-atmosphere
505: temperature alone does not produce a good fit ($\chi^2_{\nu}=1.62$,
506: nhp=0.7\%), nor does varying the normalization (and inferred radius)
507: alone ($\chi^2_{\nu}=1.85$, nhp=0.07\%).
508: If the thermal component is due to continued accretion, and the
509: variation is due to changes in the accretion rate onto the
510: neutron star, we would expect only the temperature to vary, not the
511: inferred radius. It is possible that the emitting area varies if
512: accretion is ongoing, but variation in the emitting area has not been
513: clearly seen in field qLMXBs with higher statistics
514: \citep{Rutledge02b,Campana03}.
515: Therefore we conclude that the power-law probably varies, and that we
516: do not have strong evidence for variability of the thermal component in this
517: source.
518:
519: We can also compare
520: X4's flux in our \Chandra observations to its flux in ROSAT HRI
521: observations from 1992-1996 (which contain very little spectral
522: information). \citet{Verbunt98} report 49$\pm9$ counts
523: from X4 (which we confidently identify with our source) in 58820 seconds of
524: exposure. Using a power-law spectral fit to only the 0.5-2.5 keV \Chandra
525: data (with $N_H$ fixed to the cluster value, $1.3\times10^{20}$
526: cm$^{-2}$), we find a power-law index of 2.9$\pm0.1$, with
527: (absorbed) 0.5-2.5 keV X-ray fluxes
528: ranging between $2.7\times10^{-14}$ to $1.0\times10^{-14}$ ergs
529: s$^{-1}$ cm$^{-2}$ (for different parts of the \Chandra data, as
530: above). Using
531: this spectrum, the PIMMS tool\footnote{Available at
532: http://asc.harvard.edu/toolkit/pimms.jsp} gives an absorbed 0.5-2.5
533: keV X-ray flux of $2.8\times10^{-14}$ ergs s$^{-1}$ cm$^{-2}$ for the
534: ROSAT HRI data. This flux is consistent with the upper end of the range
535: of fluxes observed with \Chandra.
536:
537: \section{Discussion}\label{s:disc}
538:
539: The 47 Tuc X-ray sources W37, W17 and X4 are very probably qLMXBs. No
540: other identified source types
541: observed in globular clusters show spectral components consistent with
542: the emission from a 10 km radius hydrogen atmosphere, as observed in
543: these objects. Millisecond pulsars show thermal emission from a much
544: smaller area, and at lower luminosities
545: \citep{Zavlin02,Grindlay02,Bogdanov04}. Magnetic
546: CVs show soft, blackbody-like components plus harder emission, but
547: with emission radii that
548: are much larger (generally hundreds of km) than the implied radii
549: (0.5-2 km) from blackbody fits to these sources. We also
550: note that these three objects are the three X-ray brightest objects
551: (excepting X7, another known qLMXB) not yet
552: positively identified in our \HST optical identification program
553: \citep{Edmonds03a, Edmonds03b}. If W17 were a cataclysmic variable
554: (or, even less likely, an X-ray active binary), its faint upper
555: limit of $U\gtrsim24$ would produce an unusual X-ray to optical flux
556: ratio, which led \citet{Edmonds03a} to
557: suggest this object as a possible qLMXB. X4 fell outside the \HST
558: fields of view analyzed by \citet{Edmonds03a}, but its X-ray spectral
559: similarities to W17 are convincing. W37's error circle showed no
560: blue or variable objects, but two bright main-sequence stars fall
561: within a 2$\sigma$ error circle (not unusual considering its projected
562: location at the center of the cluster), likely obscuring the true
563: (fainter) companion \citep[as for X7,][]{Edmonds02a}. The far-UV imaging
564: observations of
565: \citet{Knigge02} include W37's position, and may be able to provide stronger
566: constraints on a possible UV counterpart. We conclude that W37's varying
567: $N_H$ and
568: eclipses indicate that it is a transiently accreting qLMXB, with
569: marked similarities to X5 (HGL03). We also note that W37 is the
570: second qLMXB without a recorded outburst to have an identified period,
571: after X5, and that its short period implies a very dim low-mass companion.
572:
573: We can not rule out possibilities other than qLMXBs for W17 and X4.
574: It is possible that
575: W17 and X4 are neutron stars without Roche-lobe filling low-mass
576: companions, accreting from the intracluster medium \citep{Pfahl01} or
577: from the normal stellar winds of low-mass stellar companions in close
578: orbits, in so called pre-LMXB systems
579: \citep{Willems03}. Either of these is possible, but we judge a
580: stellar wind model to be more likely, because neutron stars that
581: accrete from the intracluster medium must also be able to accrete the
582: higher-density wind from a low-mass star, and because all the (slowly
583: spinning) neutron
584: stars in the cluster should display signals of accretion from the
585: intracluster medium if any do. The low bolometric luminosities from
586: the neutron star surfaces of W17 and X4 indicate a very low level of
587: time-averaged mass transfer in the \citet{Brown98} model, or enhanced
588: neutrino cooling. Mass transfer from a pre-LMXB by a stellar wind thus
589: seems a reasonable possibility.
590: However, similarly low quiescent luminosities and power-law dominated
591: spectra have recently been found for several transient LMXBs:
592: $L_X$(0.5-10 keV)=$5\times10^{31}$ ergs s$^{-1}$ for SAX
593: J1808.4-3658 in quiescence \citep{Campana02}, $L_X$(0.5-10
594: keV)=$9\times10^{31}$ ergs s$^{-1}$ ($d$/8.5 kpc)$^2$ for XTE
595: J2123-058 \citep{Tomsick04}, and $L_X$(0.5-10 keV)=$1\times10^{32}$
596: ergs s$^{-1}$($d$/4.9 kpc)$^2$ for SAX J1810.8-2609
597: \citep{Jonker03}.
598: This suggests that W17 and X4 may be normal qLMXBs.
599: They seem to be generally
600: consistent with the anticorrelation between X-ray luminosity, and the
601: fraction of that luminosity seen in the power-law component, suggested
602: by \citet{Jonker04}. On the other hand, W37 \citep[like the qLMXB in
603: NGC 6397,][]{Grindlay01b}, appears to show a smaller power-law
604: component than expected from the anticorrelation of \citet{Jonker04}.
605:
606:
607: The faint soft component in W37's spectrum must be
608: associated with W37,
609: since the position of the source does not change between the brightest
610: and faintest intervals (as would be expected if another X-ray source
611: produced this component). The X-rays might
612: be produced by the rapidly rotating companion star's corona, as
613: suggested \citet{Bildsten00} and disproved \citep{Lasota00,Kong02} for
614: quiescent black hole systems. We can
615: estimate the coronal saturation luminosity of a
616: companion as ${\rm log}
617: L_{cor}=-2.9+2 {\rm log} R_d$, where $L_{cor}$ and $R_d$ are
618: measured in units of the solar luminosity and radius
619: respectively \citep{Fleming89}. Assuming a companion mass of 0.3
620: \Msun\ (appropriate for a
621: lower main-sequence star in a 3-hour orbit; see \S~\ref{s:timing}), we
622: compute a radius
623: $2.5\times10^{10}$ cm, and thus a coronal saturation luminosity
624: ${\rm log} L_{cor} < 29.8$ ergs s$^{-1}$, which is significantly less
625: than the $5\times10^{30}$ ergs s$^{-1}$ required for the additional
626: component. Scattering of W37's X-rays in an
627: accretion disk wind might produce a similar spectrum, as observed in
628: the eclipsing CVs OY Car, DQ Her and UX UMa
629: \citep{Naylor88,Mukai03,Pratt04}. We are not able to make predictions
630: about the efficiency of scattering in such a wind, but we do not see
631: any outstanding objections to this scenario.
632:
633: Alternatively, this faint less-absorbed component might be identified
634: with the power-law component observed
635: in many field qLMXBs \citep[e.g., ][]{Campana98a, Rutledge01a,
636: Rutledge01b}, and in W17 and X4. If this identification is correct,
637: then this
638: component must be generated in an extended environment around the
639: neutron star, and not near the surface. A rough estimate of its size
640: can be estimated by assuming that the place of its generation is large
641: compared to the size of variations in the accretion disk. (This assumes
642: that the variations in $N_H$ observed for the neutron star component are
643: due to material in the accretion disk.) The
644: accretion disk is probably of order $4\times10^{10}$ cm in radius, for
645: typical masses of the components (1.4 and 0.1 \Msun) and an accretion
646: disk size one-half the orbital separation. Assuming
647: that variations in the accretion disk height may be up to one-tenth the
648: radius of the disk, this suggests that the environment that generates the
649: faint soft component should be larger than $4\times10^{9}$ cm. This
650: is significantly larger than the light-cylinder radius for neutron
651: stars spinning at millisecond periods. If the X-rays are generated
652: within or at the edge of the magnetosphere, the size of the emitting
653: region implies that a pulsar
654: mechanism is operating, and that the observed X-rays may be a shock
655: from the pulsar wind as suggested by \citet{Campana98a}.
656: The physical nature of the additional component to
657: W37's spectrum cannot be conclusively determined from this small
658: amount of data, but the possibility that we are seeing a pulsar wind
659: in a system with an accretion disk is intriguing.
660:
661: W17 and X4 indicate that a class of faint qLMXBs dominated by harder
662: nonthermal emission may exist in similar numbers as the class of
663: qLMXBs dominated by thermal emission from the neutron star surface
664: \citep{Heinke03d}. Additional qLMXBs
665: with quiescent properties similar to SAX J1808.4-3658 hiding among the
666: low-luminosity sources in 47 Tuc cannot be ruled out, since the
667: neutron star atmosphere is not always clearly detectable
668: \citep{Campana02}.
669: These particular identifications would be difficult with
670: shorter exposures, as have been performed for other globular
671: clusters, due to
672: the burying of their telltale thermal components under a stronger
673: nonthermal component and/or a blanket of absorbing gas, and
674: their significantly lower fluxes. However, it will be worthwhile to
675: look for similar objects in the nearest, best-studied clusters.
676: The numbers of faint qLMXBs with $L_X<10^{32}$ ergs s$^{-1}$ may be
677: similar to those of the identified qLMXBs. We note that there are few
678: hard qLMXBs among sources with
679: $L_X=10^{32-33}$ ergs s$^{-1}$, using the results of detailed X-ray
680: and optical studies of 47 Tuc, NGC 6397, and NGC 6752
681: \citep{Edmonds03a,Edmonds03b,Grindlay01b,Pooley02b}, but a significant
682: population below $L_X<10^{32}$
683: ergs s$^{-1}$ cannot be ruled out. If we extrapolate from these two
684: faint, nonthermally-dominated qLMXBs to the globular cluster system
685: \citep[using their close encounter frequencies, as in][and Poisson
686: statistics]{Heinke03d},
687: an additional 56$^{+61}_{-36}$ such
688: systems are suggested, added to the $\sim$100 qLMXBs already
689: anticipated \citep{Pooley03,Heinke03d}.
690: Study of faint qLMXB systems may shed light
691: on the interior physics of neutron stars, the physics of accretion at
692: low mass transfer rates, and possibly the transition as LMXBs evolve to
693: become millisecond pulsars \citep{Burderi02, Burderi03}.
694:
695: Early estimates of the numbers of millisecond
696: pulsars ($\sim$10000) and their suggested progenitors, LMXBs (12), in
697: globular clusters indicated (for LMXB lifetimes 1/10 those of
698: millisecond pulsars, $\sim$1 and 10 Gyr) a large discrepancy in
699: their respective birthrates \citep{Kulkarni90,Hut91}, thus implying
700: that other formation processes may have created some globular cluster
701: pulsars \citep{Bailyn90a}.
702: Radio timing surveys of 47 Tuc have identified 22 millisecond pulsars
703: \citep{Camilo00, Freire01a}, and X-ray, optical and radio
704: imaging surveys indicate that the total number of millisecond pulsars
705: in 47 Tuc is of order 30 \citep{Heinke04a,Edmonds03b,McConnell04,Grindlay02}.
706: When these constraints are compared with the current number of
707: likely quiescent LMXBs in 47 Tuc (5), the birthrate discrepancy between
708: millisecond pulsars and LMXBs disappears, indicating that LMXBs are
709: probably sufficient to produce the observed millisecond pulsars.
710:
711:
712: \acknowledgements
713:
714: We thank the Penn State team, particularly P. Broos, for the
715: development and support of the ACIS\_EXTRACT software.
716: We thank D. Lloyd for the use of his neutron
717: star atmosphere models.
718: We also thank P. Jonker and R. Wijnands for useful discussions and
719: insightful comments, and the anonymous referee for a useful report.
720: C.~O.~H. acknowledges the support of \Chandra grant GO2-3059A, and of the
721: Lindheimer fund at Northwestern University.
722: %Facilities:\facility{CXO(ACIS)}.
723:
724: \bibliography{src_ref_list}
725: \bibliographystyle{apj}
726:
727:
728: %figures here....
729:
730: \clearpage
731:
732: %\vspace*{0.3cm}
733:
734: \begin{figure}
735: \figurenum{1}
736: \epsscale{0.8}
737: \plotone{f1.eps}
738: \caption{Top: Lightcurve of W37 (0.3-8 keV), in 300 second
739: bins, from first two 2002 observations (OBS\_IDs 2735 and 3384,
740: 65 and 5 ksec, respectively). Time is
741: labeled in units of the best-fit period, 11112.5
742: seconds, with integer phases occurring at the times of mid-eclipse.
743: Horizontal bars indicate portions of the data taken for a high-state
744: spectrum (2735,H); the remainder of OBS\_ID 2735 is taken as a
745: low-state spectrum (2735,L).
746: Bottom: Hardness ratio (1-6 keV/0.3-1 keV) lightcurve for
747: W37, corresponding to the same timespan as above, but with 1200 second
748: bins. An anticorrelation can be seen between hardness and count
749: rate.}\label{fig:w37lc}
750: \end{figure}
751:
752:
753: \clearpage
754:
755: \begin{figure}
756: \figurenum{2}
757: \epsscale{0.8}
758: \plotone{f2.eps}
759: \caption{Top: 0.3-8 keV lightcurve of X4 (W125), in
760: 1200 second bins, from first two 2002 observations. Time is labeled
761: in hours. A horizontal bar marks the portion of the data taken for
762: the high-state spectrum. Bottom: Hardness ratio (1-6 keV/0.3-1 keV)
763: lightcurve for X4, with 5000 second bins. A correlation can be seen
764: between hardness and count rate. } \label{fig:x4lc}
765: \end{figure}
766:
767: %\clearpage
768:
769: \begin{figure}
770: \figurenum{3}
771: \epsscale{0.7}
772: %\plottwo{f3a.eps}{f3b.eps}
773: \includegraphics[angle=270,scale=.36]{f3a.eps}
774: \includegraphics[angle=270,scale=.36]{f3b.eps}
775: \caption{Left: Folded lightcurve of all W37 2002
776: data on 11112.5 second period, with the data repeated over two phases
777: for clarity. A clear eclipse can be seen at phase 0. Each bin is
778: 185.2 seconds long. Right: Folded lightcurve of all W37 2002 data on
779: 22225 second period, showing eclipses at phases 0 and 0.5.
780: }\label{fig:w37fold}
781: \end{figure}
782:
783: \clearpage
784:
785: \begin{figure}
786: \figurenum{4}
787: \epsscale{0.6}
788: \includegraphics[angle=270,scale=.60]{f4.eps}
789: \caption{X-ray spectrum of W17 from the 2002 observations, fit with an
790: absorbed hydrogen atmosphere
791: model and power-law as in Table~\ref{tab:spec}. The contributions from
792: the two components are shown, with the power-law dominating at
793: higher energies.}\label{fig:w17spec}
794: \end{figure}
795:
796:
797: \begin{figure}
798: \figurenum{5}
799: %%\epsscale{0.6}
800: \includegraphics[angle=270,scale=.60]{f5_color.eps}
801: \caption{X-ray spectra of W37, taken from four different
802: portions of the data with different fluxes. The spectra were fit with
803: an absorbed hydrogen atmosphere model plus a separately absorbed
804: power-law. Only the absorption on the hydrogen-atmosphere model was
805: allowed to change between data segments. Eight portions were fit
806: to produce the results in Tables~\ref{tab:spec} and ~\ref{tab:add},
807: but only four are shown here to
808: reduce confusion. From highest to lowest flux, these are [2735,H],
809: dashed line; [2735,L], squares and dotted (red) line; [2737,L],
810: triangles and solid (blue) line;
811: [2736], stars and dash-dotted (green) line. (See the electronic edition of the
812: Journal for a color version of this figure.)}\label{fig:w37spec}
813: \end{figure}
814:
815:
816: \begin{figure}
817: \figurenum{6}
818: \epsscale{0.6}
819: \includegraphics[angle=270,scale=.60]{f6_color.eps}
820: \caption{X-ray spectra of X4 from two parts of the
821: first 2002 observation, fit with an absorbed
822: hydrogen atmosphere model and a power-law (with the power-law
823: component allowed to vary between observations) as in
824: Table~\ref{tab:spec}. The lower-flux spectrum (red) shows the
825: greatest differences at higher energies, indicating the difference is
826: not due to photoelectric absorption. (See the electronic edition of the
827: Journal for a color version of this figure.)}\label{fig:x4spec1}
828: \end{figure}
829:
830:
831: \begin{figure}
832: \figurenum{7}
833: \epsscale{0.6}
834: \includegraphics[angle=270,scale=.60]{f7.eps}
835: \caption{X-ray spectra of X4 from the rest of the 2002
836: observations (except those in Fig.~\ref{fig:x4spec1}), fit with an absorbed
837: hydrogen atmosphere model and a power-law (with the power-law
838: component allowed to vary between observations) as in
839: Table~\ref{tab:spec}. The contributions from the two components are
840: shown, with the power-law dominating at higher energies.}\label{fig:x4spec2}
841: \end{figure}
842:
843: %tables
844:
845: \clearpage
846:
847: \begin{deluxetable}{lccccr}
848: \tabletypesize{\footnotesize}
849: \tablewidth{4.8truein}
850: \tablecaption{\textbf{Summary of \Chandra Observations}}
851: \tablehead{
852: \colhead{Seq,OBS\_ID} & \colhead{Start Time} & \colhead{Exposure} &
853: \colhead{Aimpoint} & \colhead{Frametime} & \colhead{CCDs}
854: }
855: \startdata
856: 300003,078 & 2000 Mar 16 07:18:30 & 3875 & ACIS-I & 0.94 & 1/4 \\
857: 300028,953 & 2000 Mar 16 08:39:44 & 31421 & ACIS-I & 3.24 & 6 \\
858: 300029,954 & 2000 Mar 16 18:03:03 & 845 & ACIS-I & 0.54 & 1/8 \\
859: 300030,955 & 2000 Mar 16 18:33:03 & 31354 & ACIS-I & 3.24 & 6 \\
860: 300031,956 & 2000 Mar 17 03:56:23 & 4656 & ACIS-I & 0.94 & 1/4 \\
861: 400215,2735 & 2002 Sep 29 16:59:00 & 65237 & ACIS-S & 3.14 & 5 \\
862: 400215,3384 & 2002 Sep 30 11:38:22 & 5307 & ACIS-S & 0.84 & 1/4 \\
863: 400216,2736 & 2002 Sep 30 13:25:32 & 65243 & ACIS-S & 3.14 & 5 \\
864: 400216,3385 & 2002 Oct 01 08:13:32 & 5307 & ACIS-S & 0.84 & 1/4 \\
865: 400217,2737 & 2002 Oct 02 18:51:10 & 65243 & ACIS-S & 3.14 & 5 \\
866: 400217,3386 & 2002 Oct 03 13:38:21 & 5545 & ACIS-S & 0.84 & 1/4 \\
867: 400218,2738 & 2002 Oct 11 01:42:59 & 68771 & ACIS-S & 3.14 & 5 \\
868: 400218,3387 & 2002 Oct 11 21:23:12 & 5735 & ACIS-S & 0.84 & 1/4 \\
869: \enddata
870: \tablecomments{Times in seconds. Subarrays are indicated by
871: fractional numbers of CCDs.
872: } \label{tab:obs}
873: \end{deluxetable}
874:
875:
876:
877: \clearpage
878: \thispagestyle{empty}
879: \begin{landscape}
880: \begin{deluxetable}{lccccccccr}
881: \tabletypesize{\scriptsize}
882: %\rotate
883: \tablewidth{8.4truein}
884: \tablecaption{\textbf{X-ray Spectral Model Parameters}}
885: \tablehead{
886: \colhead{\textbf{Obs}} & \colhead{$F_{\rm X,abs}$(0.5-10 keV)}
887: & \colhead{$N_H$} & \colhead{kT} & \colhead{R} &
888: \colhead{$L_{X,NS}$(0.5-10 keV)} & \colhead{$L_{bol,NS}$} &
889: \colhead{$\Gamma$} &
890: \colhead{$L_{X,PL}$(0.5-10 keV)} & \colhead{$\chi^2_{\nu}$/DoF(nhp)} \\
891: & ($10^{-14}$ ergs s$^{-1}$) & ($10^{20}$ cm$^{-2}$) & (eV) & (km) & ($10^{31}$ ergs s$^{-1}$) &
892: ($10^{31}$ ergs s$^{-1}$) & & ($10^{31}$ ergs s$^{-1}$) & \\
893: }
894: \startdata
895: \tableline
896: \multicolumn{10}{c}{\bf W17, 2000 and 2002 data} \\
897: \tableline
898: - & $1.5^{+0.3}_{-0.4}$ & $4.1^{+1.7}_{-1.3}$ & $46^{+11}_{-12}$ &
899: $15.0^{+15}_{-5.3}$ & $1.5$ & $7.3$ & $1.9^{+0.4}_{-0.3}$ &
900: $3.3$ & 1.28/43(10\%) \\
901: \tableline
902: \multicolumn{10}{c}{\bf W37, 2000 and 2002 data} \\
903: \tableline
904: - & -$^a$ & -$^a$ & $82^{+10}_{-9}$ & $12.3^{+5.8}_{-3.5}$ & 26 & 53 & 3.0$^{+4.8}_{-0.9}$ & 0.56 & 1.24/57(11\%) \\
905: \tableline
906: \multicolumn{10}{c}{\bf X4, $N_H$ and H-atmosphere fixed, power-law varies} \\
907: \tableline
908: 2735 high & 3.7$^{+1.3}_{-1.3}$ & $5.0^{+0.7}_{-1.4}$ & $53^{+13}_{-8}$
909: & $10.8^{+7.5}_{-4.7}$ & 2.0 & 7.3 & $2.4^{+0.3}_{-0.3}$ & 9.8 & 1.30/38(11\%) \\
910: 2735 low & 1.3$^{+0.5}_{-0.7}$ & - & - & - & - & - &
911: $3.1^{+0.5}_{-0.7}$ & 2.5 & - \\
912: Other 2002 & 2.0$^{+0.4}_{-0.9}$ & - & - & - & - & - &
913: $2.3^{+0.4}_{-0.3}$ & 4.6 & - \\
914: 2000 & 1.3$^{+0.7}_{-0.8}$ & - & - & - & - & - &
915: $2.4^{+1.0}_{-1.0}$ & 2.4 & - \\
916: \tableline
917: \multicolumn{10}{c}{\bf X4, $N_H$ and power-law fixed, H-atmosphere varies} \\
918: \tableline
919: 2735 high & $3.2^{+0.3}_{-2.0}$ & $3.9^{+0.6}_{-1.2}$ & $97^{+24}_{-19}$ & 4.1+2.9-1.7 & 6.6 & 12 & 2.1$^{+0.2}_{-0.4}$ & 3.8 & 1.35/38(7\%) \\
920: 2735 low & $1.7^{+0.3}_{-1.1}$ & - & $44^{+12}_{-11}$ & $19^{+22}_{-9}$ & 2.0 & 10 & - & - & - \\
921: Other 2002 & $1.9^{+0.3}_{-0.7}$ & - & $66^{+12}_{-11}$ & $7.0^{+2.0}_{-2.6}$ & 2.7 & 7.1 & - & - & - \\
922: 2000 & $1.6^{+1.9}_{-1.1}$ & - & $28^{+32}_{-1,b}$ & $70^{+26}_{-62}$ & 1.5 & 24 & - & - & - \\
923: \enddata
924: \tablecomments{ Spectral fits to W37, W17 and separate parts of X4 data,
925: using \citet{Lloyd03} hydrogen-atmosphere neutron star model plus a
926: power-law model. For X4, some parameters are held fixed between
927: observations, while others are allowed to vary. All errors are 90\%
928: confidence limits. Distance of
929: 4.85 kpc is assumed. Flux measurements are absorbed (no correction
930: for $N_H$), while $L_X$ and $L_{bol}$ are unabsorbed.
931: Neutron star H-atmosphere radius and temperature for assumed
932: grav. redshift of 0.306, implying 10 km, 1.4 \Msun\ NS; this
933: tests for consistency with the standard model. \\
934: $^a$ These parameters vary among W37's observations, and are presented in
935: Table~\ref{tab:add}. \\
936: $^b$ This parameter reached the hard limit of the model.
937: }\label{tab:spec}
938: \end{deluxetable}
939: \clearpage
940: \end{landscape}
941:
942: \begin{deluxetable}{lcccr}
943: \tabletypesize{\footnotesize}
944: \tablewidth{5.0truein}
945: \tablecaption{\textbf{Additional Parameters for W37}}
946: \tablehead{
947: \colhead{\textbf{Obs.}} & \colhead{Cts} & \colhead{$F_X$, (0.5-2.5 keV)} &
948: \colhead{$N_H$} & \colhead{Intervals of spectral extractions} \\
949: & & ($10^{-14}$ ergs s$^{-1}$) & ($10^{20}$ cm$^{-2}$) & ($10^3$ s)
950: }
951: \startdata
952: 2735,H & 618 & $7.4^{+0.8}_{-0.5}$ & $5.9^{+3.2}_{-1.9}$ & 29.006:43.006, 53.006:66.102 \\
953: 2735,L & 338 & $3.2^{+0.9}_{-0.3}$ & $31^{+5}_{-5}$ & 0:29.006, 43.006:53.006 \\
954: 2736 & 40 & $0.2^{+0.8}_{-0.1}$ & $724^{+1e6,a}_{-371}$ & 74.104:140.209 \\
955: 2737,H & 94 & $2.7^{+0.8}_{-0.4}$ & $38^{+9}_{-7}$ & 266.385:278.006, 287.506:289.506 \\
956: 2737,L & 87 & $0.7^{+0.7}_{-0.2}$ & $146^{+29}_{-20}$ & 278.006:287.506, 289.506:332.490 \\
957: 2738,H & 29 & $1.5^{+1.2}_{-0.3}$ & $67^{+25}_{-14}$ & 1038.010:1045.010 \\
958: 2738,L & 71 & $0.5^{+0.7}_{-0.2}$ & $201^{+48}_{-32}$ & 980.879:1038.010, 1045.010:1051.550 \\
959: 2000 & 46 & $0.4^{+0.7}_{-0.1}$ & $248^{+76}_{-43}$ & - \\
960: \enddata
961: \tablecomments{ Spectral fits to various parts of W37 data,
962: using \citet{Lloyd03} hydrogen-atmosphere neutron star model and
963: power-law model (as in Table~\ref{tab:spec}), with $N_H$ the only variable
964: parameter. Times are barycentered
965: and should be added to reference time 149706994.3 (seconds after MJD 50814.0). All errors
966: are 90\% confidence limits. Distance of
967: 4.85 kpc is assumed. \\
968: $^a$ This parameter reached the hard limit of the model.
969: }\label{tab:add}
970: \end{deluxetable}
971:
972:
973: \end{document}
974: