0810.0007/ms.tex
1: \documentclass[preprint]{aastex}
2: %\usepackage{graphicx}
3: %\renewcommand{\textfraction}{0.0}
4: %\renewcommand{\topfraction}{1.0}
5: %\renewcommand{\bottomfraction}{0.0}
6: 
7: \begin{document}
8: \newcommand{\srca}{GX~17$+$2}
9: \newcommand{\srcb}{4U~1705$-$44}
10: \newcommand{\srcc}{4U~1728$-$34}
11: 
12: \title{Column Densities Towards Three Bursting Low-Mass X-ray Binaries from 
13: High Resolution X-ray Spectroscopy}
14: 
15: \author{Patricia Wroblewski, Tolga G\"uver, and Feryal \"Ozel}
16: 
17: \affil{University of Arizona, Departments of Astronomy and Physics, 
18: 933 N. Cherry Ave., Tucson, AZ 85721}
19: 
20: \begin{abstract}
21: 
22: We measured the galactic hydrogen column densities to the neutron-star
23: binaries GX~17$+$2, 4U~1705$-$44, and 4U~1728$-$34 by modeling the Mg
24: and Si absorption edges found in high-resolution X-ray spectra
25: obtained by the Chandra X-ray Observatory. We found for GX~17$+$2,
26: $N_{\rm H} = (2.38 \pm 0.12) \times 10^{22}$~cm$^{-2}$, for
27: 4U~1705$-$44, $N_{\rm H} = (2.44 \pm 0.09) \times 10^{22}$~cm$^{-2}$,
28: and for 4U~1728$-$34, $N_{\rm H} = (2.49 \pm 0.14) \times
29: 10^{22}$~cm$^{-2}$. These values are in reasonable agreement with the
30: hydrogen column densities inferred earlier from modeling of the
31: continuum spectra of the sources. Our results can be used to constrain
32: the uncertainties of model parameters of the X-ray spectra of these
33: sources that are correlated to the uncertainties of the hydrogen
34: column density. In the case of continuum spectra obtained during
35: thermonuclear X-ray bursts, they will significantly reduce the
36: uncertainties in the spectroscopically measured masses and radii of
37: the neutron stars.
38: 
39: \end{abstract}
40: 
41: \keywords{X-rays: ISM, X-rays: binaries, stars: neutron, 
42: stars: individual (GX~17$+$2, 4U~1705$-$44, 4U~1728$-$34)}
43: 
44: \section{Introduction}
45: 
46: Low-mass X-ray binaries (LMXBs) that show thermonuclear
47: bursts are ideal targets for measuring the masses and radii of neutron
48: stars. During these short-lived flashes, the flux from the stellar
49: surface dominates the persistent luminosity by up to two orders
50: of magnitude and can often reach the local Eddington limit. Owing to
51: their low magnetic fields, inhomogeneities in the surface emission are
52: neither observed (e.g., Galloway et al.\ 2003), nor theoretically
53: expected, during the peak of the bursts.  Modeling time resolved X-ray
54: spectra during these bursts can, thus, lead to an accurate measurement
55: of either the gravitational redshift, or, if the distance to the
56: source is also known, to a measurement of the mass and the radius of
57: neutron star (van Paradijs 1978; Damen et al.\ 1990; Lewin, van
58: Paradijs, \& Taam 1993; \"Ozel 2006; \"Ozel, G\"uver, \& Psaltis
59: 2008).
60: 
61: Accurate measurements of column densities towards the bursting LMXBs
62: are important for two reasons. First, X-ray spectra are attenuated by
63: the interstellar medium (ISM) due to photoelectric absorption and
64: scattering by dust grains. When X-ray spectra, particularly in the
65: soft X-ray band (0.1-10~keV), are used to measure the effective
66: temperature of the neutron star, $T_{\rm eff}$, and its radius R, this
67: extinction affects the modeling of continuum spectra, the inferred
68: temperature, and, thus, the measured stellar radius. Because of the $R
69: \sim T_{\rm eff}^{-2}$ scaling of the measured radius, uncertainties in the
70: column density can give rise to large errors in radius. The spectral
71: parameters are indeed very sensitive to the assumed or allowed values
72: of the hydrogen column density. We show in Figure~1 an example of the
73: correlated uncertainties between the temperature and the column
74: density for the X-ray spectrum of 4U~1728$-$34 taken with the Rossi
75: X-ray Timing Explorer Proportional Counter Array during a
76: thermonuclear burst. Such correlations alone can lead to $\gtrsim
77: 15\%$ uncertainties in stellar radii.
78: 
79: Second, measuring the column of interstellar extinction towards an
80: X-ray source facilitates the use of a novel distance determination
81: technique based on red clump stars along the line of sight to the
82: source. The red clump stars are excellent infrared standard candles
83: (Paczynski \& Stanek 1998; Lopez-Corredoira et al.\ 2002) that can be
84: used to measure the run of reddening with distance. This method
85: establishes a distance ladder in the galaxy, which can then be
86: compared to the extinction of X-ray sources that is determined by
87: their high resolution spectra to determine their distances.
88: 
89: An independent measurement of the X-ray column density through a
90: grating observation, where the photoelectric absorption edges of each
91: element within the observational window can be determined, is valuable
92: for correctly computing the spectral parameters of these binary
93: systems. This is a method that has been previously applied to LMXBs by
94: Juett, Schulz, \& Chakrabarty (2004) and Juett et al. (2006), who
95: measured the O, Ne, and Fe edges in their spectra obtained with the
96: Chandra High Energy Transmission Grating (HETG) and XMM-Newton. It has
97: also been employed in the case of Anomalous X-ray pulsars by Durant \&
98: van Kerkwijk (2006).
99: 
100: In this paper, we analyze Chandra HETG spectra taken during numerous
101: observations of GX~17$+$2, 4U~1705$-$44, and 4U~1728$-$34.  These
102: binaries do not have previously measured hydrogen column densities
103: from absorption edges in high-resolution X-ray grating spectra. We
104: report the equivalent column densities for each of the measurable
105: edges, investigate any possible systematic uncertainties, and compare
106: our findings with column densities obtained previously from continuum
107: spectra for these sources.
108: 
109: All three sources have shown frequent thermonuclear (Type-I) X-ray
110: bursts, with a total number as high as 106 in the case of 4U~1728$-$34
111: (Galloway et al.\ 2008). A large number of these bursts (e.g., 69 for
112: the last source) also showed characteristic photospheric radius
113: expansion properties, indicating that the local Eddington limit has
114: been reached. Thus, they are ultimately good candidates for the
115: measurement of the neutron star mass and radius. Given their locations
116: in the galactic plane, extinction towards these sources can also be
117: used to study the properties of the ISM in the Galaxy.
118: 
119: \section{Observations and Data Reduction}
120: 
121: The low-mass X-ray binaries GX~17$+$2, 4U~1705$-$44, and 4U~1728$-$34
122: were observed multiple times with the High Energy Transmission Grating
123: Spectrometer (HETG) onboard the Chandra X-ray Observatory. Our study
124: includes a total of nine archival observations of these three sources.
125: The HETG consists of the Medium Energy Gratings (MEG), with a
126: $2.5-31~{\rm \AA}$ wavelength range and a resolution of $\delta
127: \lambda = 0.023~{\rm \AA}$ (for the first-order spectra), and the High
128: Energy Gratings (HEG),with a $1.2-15~{\rm \AA}$ range and a resolution
129: of $\delta \lambda = 0.012~{\rm \AA}$.
130: 
131: The three observations of GX~17$+$2, with 30.18~ks, 23.68~ks, and
132: 24.08~ks durations, were all taken in the continuous clocking (CC)
133: mode between 2004 and 2006. Both observations of 4U~1705$-$44 were
134: taken in the timed exposure (TE) mode, for a total exposure of
135: 25.13~ks in 2001 and 27.25~ks in 2005. 4U~1728$-$34 has been observed
136: four times, with exposure times of 10~ks (in 2002), 151.84~ks (in
137: 2006), 49.49~ks (in 2006), and 39.71 (in 2006). The first observation
138: in 2002 was in TE mode, while all the subsequent observations have
139: been performed in the CC mode. We list in Table~1 the observations
140: included in our study, together with the average count rate in each
141: observation.
142: 
143: The spectral data were reduced following standard CIAO
144: threads\footnotemark[1] using the CIAO version 4.0.1 and CALDB
145: 3.4.3. Different threads were used for continuous clock and timed
146: event modes due to the variations in the calibration. For observations
147: performed in a timed exposure mode, the high fluxes of the sources can
148: cause a pile-up in the zeroth order (undispersed) images. This can
149: give rise to problems in determining the exact zero-order position of
150: the source, which can ultimately shift the wavelength scale of the
151: grating spectrum. For this reason, we have made use of the
152: findzo.sl\footnotemark[2] files, which use the MEG arms and the
153: frame-shift streak to find the best fit zero-order
154: position. Additionally, for both types of observations, we excluded
155: any observed Type-I bursts, which have different continuum spectra and
156: may also have local heavy-element features.
157: 
158: Detector response files (ARFs) were created for the $+1$ and $-1$ MEG
159: and HEG spectra for both the TE mode and the CC mode observations.
160: Note that the ARFs include the correction for any time-dependent
161: change due to a contaminant on ACIS.  For the TE mode observations,
162: the $+1$ and $-1$ orders were combined together for both the MEG and
163: the HEG spectra. However, two concerns rendered the MEG $+1$ order of
164: the CC mode observations unusable. First, there is an instrumental
165: feature in the O edge on the MEG $+1$ side (Juett, A., Private
166: Communication). Second, all the observations we studied were carried
167: out with either a $0.33^\prime$ or $0.1667^\prime$ offset, which
168: caused the wavelength of the Mg edge to fall in or near chip gaps. In
169: the CC mode, this resulted in a complete reduction of the effective
170: area at that wavelength, while in the TE mode, there were sufficient
171: counts that were included in the analysis. Figure~\ref{effarea} shows
172: a sample observation from 4U~1728$-$34, where the effective areas of
173: each grating and order can be seen. Finally, all MEG data were binned
174: by a factor of 4 and the HEG data were binned by a factor of 8 to
175: increase the statistics while achieving the same energy resolution on
176: both gratings.
177: 
178: \section{Spectral Analysis}
179: 
180: We originally sought to measure the edges of neutral Mg, Si, Ne, O,
181: and Fe in the HETG spectra of GX~17$+$2, 4U~1705$-$44, and
182: 4U~1728$-$34. However, the source count rates were typically very low
183: at long wavelengths, leading to negligible flux around the O and Fe
184: edges for all the sources. Near the Ne edge, at a wavelength of
185: $14.31~{\rm \AA}$ (Juett et al.\ 2006), the continuum was detectable
186: but there were not enough counts for any of the sources to yield
187: statistically significant results. We thus considered only the Mg and
188: Si edges in our analyses.
189: 
190: We focused on small wavelength regions around the Mg and Si edges and
191: assumed that the continuum spectrum of each source could be fit with a
192: power-law in each of these regions, $F^{\rm c}_\lambda \propto
193: (\lambda/\lambda_{\rm edge})^\alpha$, with a break at the absorption
194: edge of interest. We used fixed edge wavelengths: $\lambda_{\rm Mg}=
195: 9.5~{\rm \AA}$ for the location of the Mg edge and $\lambda_{\rm
196: Si}=6.72~{\rm \AA}$ (Ueda et al. 2005) for the Si edge. We modeled
197: each edge by fitting a function of the form
198: \begin{eqnarray}
199: F_{\lambda} = \left \{
200: \begin{array}{lr}
201: F^{\rm c}_{\lambda} & {\rm for} \lambda > \lambda_{\rm edge}  \\
202: F^{\rm c}_{\lambda} \exp\left[-A \left( \frac{\lambda}{\lambda_{\rm edge}} \right)^3 \right]   
203: & {\rm for} \lambda \leq \lambda_{\rm edge} 
204: \end{array} \right.
205: \end{eqnarray}
206: 
207: We adjusted the specific wavelength range that we used in the fits for
208: each source because of small variations in the intrinsic continuum
209: spectra of the sources. However, between the multiple observations of
210: a given a source, we froze the range that we analyzed.  Specifically,
211: for GX~17$+$2, we used $5.8-7.2~{\rm \AA}$ for Si and $8.6-10.5~{\rm
212: \AA}$ for Mg. For 4U~1705$-$44, we chose a range of $5.7-7.5~{\rm
213: \AA}$ for Si and 8.5$-$10.3~${\rm \AA}$ for Mg. Finally, for
214: 4U~1728$-$34, we used $5.9-7.3~{\rm \AA}$ for Si and $8.5-10.3~{\rm
215: \AA}$ for Mg.  Sample spectra for all three sources in these
216: wavelength ranges are shown in Figures 3 and 4.
217: 
218: We fit the MEG and HEG data both separately and simultaneously to
219: detect any possible systematic differences. For the data taken in the
220: TE mode, the slopes obtained by fitting power-law functions to the
221: continuum of both gratings were in good agreement with each other.
222: Note that the pile-up fraction in one of the observations of
223: 4U~1705$-$44 (Obsid: 1923) was nearly 15\% around the Si edge in the
224: MEG spectrum and this result was excluded from the final analysis. In
225: the CC mode observations, however, there were more than 3$\sigma$
226: differences in the power-law indices obtained for the HEG and MEG
227: data. This is thought to be due to calibration uncertainties in this
228: mode. Because of this, only MEG data were used for all of the fits (of
229: the CC mode data) due to the larger photon collecting area of this
230: grating compared to the HEG in the wavelength region of interest
231: ($5.5-15.5~{\rm \AA}$).
232: 
233: In addition to a power-law continuum and the function modeling the
234: edges, we allowed for a number of emission and absorption line
235: features that appear in each spectral region and whose positions are
236: fixed. For consistency, we included the same number of absorption and
237: emission lines in the spectral fits of each observation of each
238: source, regardless of the apparent presence or absence of a feature in
239: that spectrum, as was done by Juett et al.\ (2006). When the features
240: were not detectable or were not well-constrained in a given
241: observation, we report a zero flux for the lines. A negative amplitude
242: corresponds to an absorption feature.
243: 
244: Around the Si absorption edge, we fixed the wavelengths of the five
245: most prominent features in the wavelength range for each source using
246: Chandra's atomic lines database ATOMDB\footnotemark[3]. The most
247: prominent features in this region are the two Si XIV and three Si XIII
248: atomic lines. In Figure~3, we show three sample spectra from each
249: source around the Si edge, along with the fitted power-law continuum,
250: the edge, and the line features. Table~2 shows the calculated fluxes
251: of the emission lines for each observation.
252: 
253: Similarly, in the wavelength range around the Mg edge, we included
254: Mg~XI, Fe~XXI, and Ne~X emission features, which are the
255: strongest. The Mg~XI features at 9.16~{\rm\AA}, 9.22~{\rm \AA}, and
256: 9.31~{\rm \AA} are part of a He-like triplet formed by the resonance,
257: intercombination, and forbidden transitions (Porquet et al.\ 2001).
258: We show in Figure~4 the results of our fits to the grating spectra
259: obtained from one of the observations of each source. We report the
260: corresponding line fluxes in Table~3 and report our findings for the
261: Si and Mg edges in the next section.
262: 
263: \footnotetext[1]{http://asc.harvard.edu/ciao/threads/}
264: \footnotetext[2]{http://space.mit.edu/cxc/analysis/findzo/Data/findzo.sl}
265: \footnotetext[3]{http://cxc.harvard.edu/atomdb/WebGUIDE/}
266: 
267: \section{Results and Discussion}
268: 
269: We used the Chandra grating observations of GX~17$+$2, 4U~1705$-$44,
270: and 4U~1728$-$34 to measure absorption coefficients for the Mg and Si
271: edges. In Tables 2 and 3, we also report the fluxes of spectral lines
272: that were measured in each observation around the Si and Mg edges,
273: respectively. We then used the absorption coefficients along with the
274: cross-sections from Gould \& Jung (1991) to find the column densities
275: for each element. We converted the Mg and Si column densities to
276: hydrogen column densities, $N_{\rm H}$, using the ISM abundances given
277: by Wilms, Allen, \& McCray (2000). We report all of the column
278: densities in Table~4. We then averaged the results obtained with each
279: absorption edge over the observations, which makes it possible to
280: discriminate any possible systematic differences between the column
281: densities obtained from the different edges.
282: 
283: For GX~17$+$2, the $N_{\rm H}$ values derived from the Mg edge ranged
284: from $(2.01-2.77) \times 10^{22}~{\rm cm}^{-2}$, while the values
285: derived from the Si edge ranged from $(1.03-2.69) \times 10^{22}~{\rm
286: cm}^{-2}$ for the three observations. Averaging over the column
287: densities derived from the Mg edge yields $(2.38 \pm 0.12) \times
288: 10^{22}~{\rm cm}^{-2}$, while over those derived from the Si edge results
289: in $(1.78 \pm 0.09) \times 10^{22}~{\rm cm}^{-2}$. We note that the
290: lower $N_{\rm H}$ value derived from the Si edge can be attributed to
291: the small edge measured in one single observation and that the
292: remaining five measurements are consistent with each other to within
293: 2$\sigma$. 
294: 
295: 4U~1705$-$44 showed a somewhat smaller scatter with equivalent $N_{\rm
296: H}$ ranging from $(2.08-2.59) \times 10^{22} {\rm cm}^{-2}$ from the
297: Mg edge measurements and $(2.95 - 4.12) \times 10^{22} {\rm cm}^{-2}$
298: from the Si edge measurements. Averaging over the column densities
299: derived from the Mg edge yields $(2.44 \pm 0.09) \times 10^{22}~{\rm
300: cm}^{-2}$, and $(3.64 \pm 0.12) \times 10^{22}~{\rm cm}^{-2}$ from the
301: Si edge. Because the difference between these two measurements are
302: larger than that expected from statistical uncertainties, we treat it
303: as a systematic difference, potentially caused by calibration errors
304: or by anomalous abundances.
305: 
306: Finally, 4U~1728$-$34 had equivalent hydrogen column densities in the
307: range $(2.39-8.23) \times 10^{22}~{\rm cm}^{-2}$ in individual edge
308: measurements. However, the highest of these values, corresponding to
309: the Silicon edge in observation 2748 (N[Si]$_{\rm H} = (8.23 \pm 0.29)
310: \times 10^{22}~{\rm cm}^{-2}$), arises from a previously reported Si
311: overabundance that is thought to be caused from calibration
312: problems in that specific observation (D'A{\'{\i}} et al.\ 2006). As a
313: result, we excluded this measurement when calculating the average
314: hydrogen column density for this source. We again calculated separate
315: average values for each edge: we obtained $N_{\rm H} = (2.49 \pm 0.14)
316: \times 10^{22}~{\rm cm}^{-2}$ from the Mg edge and $(4.27 \pm 0.14)
317: \times 10^{22}~{\rm cm}^{-2}$ from the Si edge.
318: 
319: We detect a systematic difference between the $N_{\rm H}$ values
320: obtained from Mg and Si edges in all of the observations. These could
321: naturally be attributed to non-standard ISM abundances or to anomalous
322: intrinsic abundances of the sources. However, we also observed that
323: the scatter between the absorption coefficients at the Si edge is also
324: much larger between the different observations of each source. Noting
325: that the response of the detector suffers from a very large
326: discontinuity at the Si edge but is smooth across the Mg edge, we
327: investigated whether these differences could be due to small
328: inaccuracies in the detector response across the Si edge.  In
329: particular, we explored whether there is any dependence of this
330: systematic variation on the source count rate during the observations.
331: We found that when the count rates of the sources are on average
332: higher ($\sim 30$~cts~s$^{-1}$), the N$_{\rm H}$ values found using
333: the two edges become consistent with each other to within $2\sigma$,
334: and that the discrepancy between the two values increases as the
335: source count rate decreases. This is especially evident in the two
336: spectra of 4U~1705$-$44, where one of the two spectra of this source
337: was obtained when the source count rate was 38.6~cts~s$^{-1}$ while
338: the second observation was performed during a smaller count rate epoch
339: (12.6~cts~s$^{-1}$). We note that between these observations the
340: N$_{\rm H}$ value found from the Mg edge does not change
341: significantly, while the value found from the Si edge increases and
342: the difference between the two measurement becomes as much as
343: 10$\sigma$.
344: 
345: Given that the intrinsic metal abundance of the X-ray binary or the
346: metal abundances in the ISM is unlikely to change significantly for
347: the same source within 4 years, we conclude that the difference might
348: arise because of the presence of the significant intrumental Si edge,
349: where small uncalibrated residuals of this structure affect our edge
350: measurements. Because of this, we will use $N_{\rm H}$ measurements
351: from the Mg edges as the preferred values.
352: 
353: We have also compared our measurements of hydrogen column densities
354: towards these three sources determined from the overall suppression of
355: low-resolution continuum spectra.
356: 
357: Previous estimates of the column density towards GX~17$+$2 were made
358: from modeling the continuum spectra obtained with ROSAT observations.
359: Predehl \& Schmitt (1995) found $N_{\rm H}$ values to the source in
360: the range $(1.79-2.34) \times 10^{22}$~cm$^{-2}$, assuming a blackbody,
361: thermal bremsstrahlung or a power-law model for the continuum.  These
362: values span the range obtained from the Mg and Si edge measurements
363: and are thus in agreement with our results.
364: 
365: Two estimates of the column density towards 4U~1705$-$44 exist based
366: on observations with different satellites and using different
367: continuum models. Predehl \& Schmitt (1995) reported a range of
368: hydrogen column densities towards towards the source using ROSAT
369: observations. Continuum models included a blackbody, thermal
370: bremsstrahlung or a power-law, as in the case of GX~17$+$2, and
371: resulted in a range $N_{\rm H} = (1.23 - 1.63) \times 10^{22}~{\rm
372: cm}^{-2}$. Barret \& Olive (2002), on the other hand, reported a
373: column density $N_{\rm H} = 2.4 \times 10^{22}~{\rm cm}^{-2}$ by
374: fitting a Comptonized blackbody model to the RXTE Proportional Counter
375: Array data. The former values are lower than the model-independent
376: column densities we find here, while the latter value is consistent
377: within $1\sigma$ with our results when only the Mg column is taken
378: into account.
379: 
380: Observations of 4U~1728$-$34 by ROSAT, BeppoSax, and Chandra
381: observatories have also yielded a range of estimates for the hydrogen
382: column density towards this source when the data were fit with
383: different continuum models. Schulz (1999) analyzed the ROSAT data
384: using power-law and blackbody continuum models, which yielded hydrogen
385: column densities $N_{\rm H} = (3.44 \pm 0.3) \times 10^{22}~{\rm
386: cm}^{-2}$ and $N_{\rm H} = (2.89 \pm 0.22) \times 10^{22}~{\rm
387: cm}^{-2}$, respectively. BeppoSAX data were fit with a Comptonized
388: blackbody model to give $N_{\rm H} = (2.73 \pm 0.05) \times
389: 10^{22}~{\rm cm}^{-2}$ (Piraino, Santangelo, \& Kaaret 2000).
390: Finally, the spectrum obtained with Chandra HETG was modeled with a
391: Comptonized blackbody to give 2.61$^{+.06}_{-.07} \times
392: 10^{22}~\rm{cm}^{-2}$ (D'A{\'{\i}} et al.\ 2006). Note that while
393: 4U~1728$-$34 has also been observed with the RXTE Proportional Counter
394: Array and Integral, the hydrogen column density was not well
395: constrained from these observations. All but the ROSAT measurements
396: are consistent with the column density of $N_{\rm H}=(2.49 \pm 0.14)
397: \times 10^{22}~{\rm cm}^{-2}$ measured from the Mg edge.
398: 
399: \acknowledgments
400: We thank Adrienne Juett for numerous helpful discussions and
401: suggestions on the analysis of the grating spectra as well as her
402: comments on the manuscript. We also thank Michael Nowak for his help
403: with Chandra HETG. This work was supported by NSF grant AST~07-08640.
404: 
405: \begin{thebibliography}
406: 
407: \bibitem[Barret \& Olive(2002)]{2002ApJ...576..391B} Barret, D., \& Olive, 
408: J.-F.\ 2002, \apj, 576, 391 
409: 
410: \bibitem[D'A{\'{\i}} et al.(2006)]{2006A&A...448..817D} D'A{\'{\i}}, A., 
411: et al.\ 2006, \aap, 448, 817 
412: 
413: \bibitem[Damen et al.(1990)]{1990A&A...237..103D} Damen, E., Magnier,
414: E., Lewin, W.~H.~G., Tan, J., Penninx, W., \& van Paradijs, J.\ 1990,
415: \aap, 237, 103
416: 
417: \bibitem[Durant \& van Kerkwijk(2006)]{2006ApJ...650.1082D} Durant, M., 
418: \& van Kerkwijk, M.~H.\ 2006, \apj, 650, 1082 
419: 
420: \bibitem[Galloway et al.(2003)]{2003ApJ...590..999G} Galloway, D.~K., 
421: Psaltis, D., Chakrabarty, D., \& Muno, M.~P.\ 2003, \apj, 590, 999
422: 
423: \bibitem[Galloway et al.(2008)]{2006astro.ph..8259G} Galloway, D.~K.,
424: Muno, M.~P., Hartman, J.~M., Psaltis, D., \& Chakrabarty, D.\ 2008,
425: ApJS, in press (arXiv:astro-ph/0608259)
426: 
427: \bibitem[Gould \& Jung(1991)]{1991ApJ...373..271G} Gould, R.~J., \& Jung, 
428: Y.-D.\ 1991, \apj, 373, 271 
429: 
430: \bibitem[Juett et al.(2004)]{2004ApJ...612..308J} Juett, A.~M., Schulz, 
431: N.~S., \& Chakrabarty, D.\ 2004, \apj, 612, 308 
432: 
433: \bibitem[Juett et al.(2006)]{2006ApJ...648.1066J} Juett, A.~M., Schulz, 
434: N.~S., Chakrabarty, D., \& Gorczyca, T.~W.\ 2006, \apj, 648, 1066 
435: 
436: \bibitem[Lattimer \& Prakash(2001)]{2001ApJ...550..426L} Lattimer,
437: J.~M., \& Prakash, M.\ 2001, \apj, 550, 426
438: 
439: \bibitem[Lattimer \& Prakash(2007)]{2007PhR...442..109L} Lattimer,
440: J.~M., \& Prakash, M.\ 2007, \physrep, 442, 109
441: 
442: \bibitem[Lee et al.(2001)]{2001AJ....122.3136L} Lee, J.-W., Carney, B.~W., 
443: Fullton, L.~K., \& Stetson, P.~B.\ 2001, \aj, 122, 3136 
444: 
445: \bibitem[Lewin et al.(1993)]{1993SSRv...62..223L} Lewin, W.~H.~G., van 
446: Paradijs, J., \& Taam, R.~E.\ 1993, Space Science Reviews, 62, 223 
447: 
448: \bibitem[L{\'o}pez-Corredoira et al.(2002)]{2002A&A...394..883L} L{\'o}pez-Corredoira, 
449: M., Cabrera-Lavers, A., Garz{\'o}n, F., \& Hammersley, P.~L.\ 2002,
450: \aap, 394, 883
451: 
452: \bibitem[{\"O}zel(2006)]{2006Natur.441.1115O} {\"O}zel, F.\ 2006, \nat, 
453: 441, 1115
454: 
455: \bibitem[Ozel et al 2008]{} \"Ozel, F., G\"uver, T., \& Psaltis, D., ApJ, 
456: submitted
457: 
458: \bibitem[Paczynski \& Stanek(1998)]{1998ApJ...494L.219P} Paczynski, B., \& 
459: Stanek, K.~Z.\ 1998, \apjl, 494, L219 
460: 
461: \bibitem[Piraino et 
462: al.(2000)]{2000A&A...360L..35P} Piraino, S., Santangelo, A., \& Kaaret, P.\ 2000, \aap, 360, L35 
463: 
464: \bibitem[Predehl \& Schmitt(1995)]{1995A&A...293..889P} Predehl, P., \& Schmitt, 
465: J.~H.~M.~M.\ 1995, \aap, 293, 889 
466: 
467: \bibitem[Porquet et al.(2001)]{2001A&A...376.1113P} Porquet, D., Mewe, R., 
468: Dubau, J., Raassen, A.~J.~J., \& Kaastra, J.~S.\ 2001, \aap, 376, 1113 
469: 
470: \bibitem[Schulz(1999)]{1999ApJ...511..304S} Schulz, N.~S.\ 1999, \apj, 511, 
471: 304 
472: 
473: \bibitem[Ueda et al.(2005)]{2005ApJ...620..274U} Ueda, Y., Mitsuda, K., 
474: Murakami, H., \& Matsushita, K.\ 2005, \apj, 620, 274
475: 
476: \bibitem[van Paradijs(1978)]{1978Natur.274..650V} van Paradijs, J.\ 1978, 
477: \nat, 274, 650 
478: 
479: \bibitem[Wilms et al.(2000)]{2000ApJ...542..914W} Wilms, J., Allen, A., 
480: \& McCray, R.\ 2000, \apj, 542, 914 
481: 
482: 
483: \end{thebibliography}
484: 
485: 
486: \clearpage
487: 
488: \begin{figure*} \centering
489: \includegraphics[angle=0, scale=0.70]{f1.ps}
490: \caption{ The correlation between column density and temperature is
491: shown for 4U~1728$-$34 as found for one of the photospheric radius
492: expansion bursts from an RXTE observation. The three contours show the
493: 1$\sigma$, 2$\sigma$, and 3$\sigma$ confidence levels. }
494: \label{correlation} 
495: \end{figure*}
496: 
497: 
498: \begin{figure*}
499: \centering
500:    \includegraphics[scale=0.70]{f2.ps} \caption{A sample spectrum of
501:    4U~1728$-$34 that shows the $+1$ and $-1$ orders of both MEG and
502:    HEG gratings. In addition to intrinsic differences between the
503:    effective areas of the HEG and the MEG gratings, there is also a
504:    chip gap at the MEG $+1$ order aroung 9.5~$\AA$ which causes a
505:    significant reduction in the counts in this region.}
506: \label{effarea}
507: \end{figure*}
508: 
509: \begin{figure*}
510: \centering
511:    \includegraphics[scale=0.70]{f3.ps}
512:    \caption{The fitted spectra for the Silicon edge. For
513:      4U~1705$-$44, GX~17$+$2, and 4U~1728$-$34, obsids 5500, 4564, and
514:      6568, are shown, respectively. Details of these fits can be seen in 
515:      Table~2. }
516: \label{Siedge}
517: \end{figure*}
518: 
519: \begin{figure*}
520: \centering
521: \includegraphics[scale=0.70 ]{f4.ps}
522: \caption{The fitted spectra for the Magnesium
523: edge. For 4U~1705$-$44, GX~17$+$2, and 4U~1728$-$34, obsids 5500, 4564, 
524: and 6568, are shown, respectively. Details of these fits can be seen in 
525: Table~3. }
526: \label{Mgedge}
527: \end{figure*}
528: 
529: \begin{deluxetable}{lccccc}
530: \tablecolumns{4}
531: \tablewidth{460.0pt}
532: \tablecaption{List of Chandra Observations by Source Used in this Study}
533: \tablehead{Source & Obsid & Mode & Date & Exp. Time (ks) & Average Count Rate (ct~s$^{-1}$)}
534: \startdata
535: GX~17$+$2  & 4564  & CC &   2004 July 11 & 30.18 & 30.43   \\
536:            & 6629  & CC &   2006 May 10  & 23.68 & 31.52  \\
537:            & 6630  & CC &   2006 Aug 19  & 24.08 & 29.82  \\
538: 4U~1705$-$44 & 5500  & TE & 2001 July 01 & 25.13 & 12.62 \\
539:              & 1923  & TE & 2005 Oct 26  & 27.25 & 38.60 \\
540: 4U~1728$-$34 & 2748  & TE & 2002 Mar 04  & 10.00 & 7.29  \\
541:              & 6568  & CC & 2006 July 17 & 49.49 & 6.53  \\
542:              & 6567  & CC & 2006 July 18 & 151.84 & 7.84 \\
543:              & 7371  & CC & 2006 July 22 & 39.71 & 4.53 \\
544: \enddata
545: \end{deluxetable}
546: 
547: \begin{deluxetable}{lccccccc}
548: \tablecolumns{8}
549: \tablewidth{460.0pt}
550: \tablecaption{Fluxes of Spectral Lines around the Si Edge$^1$} 
551: \tablehead{
552:  &      & \multicolumn{5}{c}{Flux ($10^{-5}$ photons $\rm cm^{-1}$ $\rm sec^{-1}$ $\rm\AA^{-1}$)} \\
553:  &      & Si XIV$_a^{2}$          & Si XIV$_b$    & Si XIII$_c$   &  Si XIII$_d$  & Si XIII$_e$ \\
554:  Source   & Obsid & 6.1804~$\rm\AA$ & 6.1858~$\rm\AA$ & 6.6479~$\rm\AA$ & 6.6882~$\rm\AA$ & 6.7403~$\rm\AA$}
555: \startdata
556: GX~17$+$2 & 4564 &-16.0$\pm$5.99$^{1}$& $--$           & -17.3$\pm$4.8   & -10.1$\pm$4.8 & 15.8$\pm$4.6 \\
557:           &6629 & $--$                & $--$           & $--$            & -13.1$\pm$9.0 & 26.1$\pm$8.5\\
558:           &6630 & -11.7$\pm$6.85      & $--$           & -11.7$\pm$5.8   & $--$           &  9.64$\pm$5.38\\
559: 4U~1705$-$44 &1923 & $--$ & $--$ & $--$ & -11.03$\pm$5.46 & $--$     \\
560:           &5500 &  -3.43$\pm$1.76     & 3.93$\pm$1.69  &$--$             & -5.59$\pm$1.46 & 5.42$\pm$1.44 \\
561: 4U~1728$-$34    &2748 & $--$          & 1.66$\pm$1.43  & 2.07$\pm$1.20   & 3.34$\pm$1.16  & $--$ \\
562:           &6567 & 2.06$\pm$1.99       & 2.80$\pm$1.89  &$--$             & -2.13$\pm$1.66 & 7.36$\pm$1.60\\
563:           &6568 &  $--$               &$--$            & $--$            & $--$           & $--$ \\
564:           &7371 &  -5.07$\pm$2.12     & 4.69$\pm$2.04  & -4.81$\pm$1.8   & $--$           & 3.87$\pm$1.76\\
565: \enddata
566: \tablenotetext{1}{Errors correspond to 1$\sigma$ statistical uncertainties.}
567: \tablenotetext{2}{Transitions from $^{a} 4\rightarrow1$ $^{b} 3\rightarrow1$ $^{c} 7\rightarrow1$ $^{d} 5\rightarrow1$ $^{e} 2\rightarrow1$}
568: \end{deluxetable}
569: 
570: \begin{deluxetable}{lccccccc}
571: \tablecolumns{8}
572: \tablewidth{500.0pt}
573: \tablecaption{Fluxes of Spectral Lines around the Mg Edge$^1$} 
574: \tablehead{
575:          & & \multicolumn{4}{c}{Flux ($10^{-5}$ photons $\rm cm^{-1}$
576:          $\rm sec^{-1}$ $\rm\AA^{-1}$)} \\ & & Mg XI$_a^{2}$ & Mg
577:          XI$_b$ & Mg XI$_c$ & Fe XXI & Ne X \\ Source & Obsid & 9.1687~$\rm\AA$ 
578:          & 9.2267~$\rm\AA$  & 9.3143~$\rm\AA$  & 9.4797~$\rm\AA$  & 10.2385~$\rm\AA$  \\}
579: \startdata
580: GX~17$+$2 &4564 &$--$          & $--$          &3.95$\pm$2.17    & 2.95$\pm$2.17     & -4.37$\pm$1.58\\
581:           &6629 &$--$          & $--$          &$--$             & $--$              & $--$ \\
582:           &6630 &$--$          & $--$          &  $--$           & $--$              & -4.98$\pm$1.91\\
583: 4U~1705$-$44&1923 &$--$        & $--$          &  $--$           & $--$             &$--$ \\
584: 	  &5500 &$--$          & 2.72$\pm$1.11& 1.52$\pm$1.22  & $--$              & $--$ \\
585: 4U~1728$-$34&2748 &$--$        & $--$        & $--$            & $--$              & $--$ \\
586:           &6567 &1.51$\pm$0.71 & 1.69$\pm$0.71  & $--$            & -0.93$\pm$0.57      &  $--$\\
587:           &6568 &$--$          & $--$          &$--$            & $--$              & $--$  \\
588:           &7371 &2.90$\pm$0.94 &1.54$\pm$0.84 & $--$             & $--$           & -0.61$\pm$0.56  \\
589: \enddata
590: \tablenotetext{1}{Errors correspond to 1$\sigma$ statistical uncertainties.}
591: \tablenotetext{2}{Transitions from $^{a} 7\rightarrow1$ $^{b} 6\rightarrow1$ $^{c} 2\rightarrow1$}
592: 
593: \end{deluxetable}
594: \begin{deluxetable}{lcccccc}
595: \tablecolumns{7}
596: \tablewidth{460.0pt}
597: \tablecaption{Column Densities from Mg and Si Edges$^1$}
598: \tablehead{
599: Source     & Obsid & N$_{\rm Mg}$& N[Mg]$_{\rm H}$ & N$_{\rm Si}$ & N[Si]$_{\rm H}$ & $\langle N_{\rm H} \rangle$ \\
600:            &       & $10^{17}~{\rm cm}^{-2}$ & $10^{22}~{\rm cm}^{-2}$ & $10^{17}~{\rm cm}^{-2}$ 
601:            & $10^{22}~{\rm cm}^{-2}$ & $10^{22}~{\rm cm}^{-2}$ }
602: \startdata
603: GX~17$+$2    & 4564  & 5.06$\pm$0.41  & 2.01$\pm$0.16 & 1.92$\pm$0.26 & 1.03$\pm$0.14  &  \\
604:              & 6629  & 6.95$\pm$1.08  & 2.77$\pm$0.43 & 5.01$\pm$0.44 & 2.69$\pm$0.23  &  \\
605:              & 6630  & 6.96$\pm$0.44  & 2.77$\pm$0.18 & 4.33$\pm$0.29 & 2.32$\pm$0.16  &  2.38$\pm$ 0.12$^{2}$\\
606: 4U~1705$-$44 & 1923  & 6.50$\pm$0.26  & 2.59$\pm$0.10 & 5.49$\pm$0.36 & 2.95$\pm$0.19  &  \\
607:              & 5500  & 5.23$\pm$0.41  & 2.08$\pm$0.16 & 7.66$\pm$0.30 & 4.12$\pm$0.16  &  2.44$\pm$ 0.09 \\
608: 4U~1728$-$34 & 2748  & 6.38$\pm$0.81  & 2.54$\pm$0.32 & 15.3$\pm$0.55 & 8.23$\pm$0.29  &  \\
609:              & 6567  & 6.08$\pm$0.51  & 2.42$\pm$0.20 & 8.36$\pm$0.33 & 4.49$\pm$0.18  &  \\
610:              & 6568  & 6.80$\pm$0.85  & 2.71$\pm$0.34 & 7.32$\pm$0.47 & 3.93$\pm$0.25  &  \\
611:              & 7371  & 6.01$\pm$0.99  & 2.39$\pm$0.39 & 7.22$\pm$0.98 & 3.88$\pm$0.53  &  2.49$\pm$0.14\\
612: 
613: \enddata
614: \tablenotetext{1}{Errors correspond to 1-$\sigma$ statistical uncertainties.}
615: \tablenotetext{2}{The average value for all observations per source for the Mg edge.} 
616: 
617: \end{deluxetable}
618: 
619: \end{document}