1: %\documentclass[12pt,preprint]{aastex}
2: \documentclass{emulateapj}
3:
4: \usepackage{xspace}
5: \usepackage{graphicx, floatflt}
6: %\usepackage{multicol}
7:
8: \newcommand{\NH}{\mbox{${\rm N}_{\rm H}$}} % Defines NH
9: \newcommand{\etal}{{\it et al.}\xspace}
10: \newcommand{\chandra}{{\sl Chandra}\xspace}
11:
12: \begin{document}
13:
14: \title{X-ray Dust Scattering at Small Angles: The Complete Halo around
15: GX13+1}
16:
17: \author{Randall K. Smith} \affil{NASA Goddard Space Flight
18: Center, Greenbelt, MD 20771} \affil{Department of Physics and
19: Astronomy, The Johns Hopkins University, 3701 San Martin Drive,
20: Baltimore, MD 21218}
21:
22: \begin{abstract}
23:
24: The exquisite angular resolution available with \chandra should allow
25: precision measurements of faint diffuse emission surrounding bright
26: sources, such as the X-ray scattering halos created by interstellar
27: dust. However, the ACIS CCDs suffer from pileup when observing bright
28: sources, and this creates difficulties when trying to extract the
29: scattered halo near the source. The initial study of the X-ray halo
30: around GX13+1 using only the ACIS-I detector done by \citet{SES02}
31: suffered from a lack of sensitivity within $50''$\ of the source,
32: limiting what conclusions could be drawn.
33:
34: To address this problem, observations of GX13+1 were obtained with the
35: \chandra HRC-I and simultaneously with the RXTE PCA. Combined with
36: the existing ACIS-I data, this allowed measurements of the X-ray halo
37: between 2-1000''. After considering a range of dust models, each
38: assumed to be smoothly distributed with or without a dense cloud along
39: the line of sight, the results show that there is no evidence in this
40: data for a dense cloud near the source, as suggested by
41: \citet{Xiang05}. Finally, although no model leads to formally
42: acceptable results, the \citet{WD01} and nearly all of the composite
43: grain models from \citet{ZDA04} give poor fits.
44: \end{abstract}
45:
46: \keywords{dust, extinction --- scattering --- X-rays: ISM}
47: \section{Introduction}
48:
49: Practically every band of the electromagnetic spectrum affects or is
50: affected by interstellar (IS) dust grains. In the IR, PAHs emit lines
51: and small grains emit continuum radiation; in the UV/optical, small
52: grains both extinct and scatter light. In X-rays, large dust grains
53: ($>0.1 \mu$m) scatter X-rays, creating halos around point sources.
54: The classic paper by \citet[][MRN77]{MRN} used the observed optical
55: extinction to determine the size distribution of dust grains between
56: 0.005-0.25$\mu$m. Newer models, such as \citet[][WD01]{WD01}, have
57: extended the modeling to include polycyclic aromatic hydrocarbons
58: (PAHs) to match the observed IR emission as well as other constraints
59: on grain abundances. Recently, \citet[][ZDA04]{ZDA04} found that a
60: wide range of dust compositions and size distributions could fit the
61: existing data, and suggested that new observational constraints from
62: X-ray halos are needed to select amongst these models.
63:
64: X-ray dust scattering halos are created by the small-angle scattering
65: of X-rays as they pass through dust grains. When an incoming X-ray
66: interacts with the electrons in a grain large compared to the X-ray
67: wavelength, the resulting Rayleigh scattering adds coherently in the
68: forward direction leading to small-angle scattering; see \citet{vdH57}
69: and \citet{ML91} for details, and \citet{Draine03} for a comprehensive
70: review. More generally, the scattering problem can be posed as that
71: of a wave interacting with a sphere, in which case the Mie solution
72: applies \citep[{\protect{\it e.g.}}][]{SD98}. In either approach, the
73: scattering depends largely on the grain size distribution, with lesser
74: dependencies on the grain composition and position along the line of
75: sight.
76:
77: Observations of X-ray scattered halos have just begun to significantly
78: impact dust models. \citet[][(SES02)]{SES02} described \chandra
79: observations of GX13+1 with the ACIS-I detector and showed that dust
80: grains do not have large ($\gtrsim 0.8$) vacuum fractions considered
81: by \citet{MW89}. SES02 also found that the extremely large grains
82: found by {\sl Ulysses}\ in the solar neighborhood \citep{Landgraf00,
83: WSD01} do not seem to be common throughout the Galaxy. Despite these
84: successes, SES02 could not distinguish between the MRN77 and WD01
85: models. This was in part due to calibration uncertainties as well as
86: inherent limitations of the data. Despite \chandra's excellent
87: angular resolution, ACIS-I observations of GX13+1 could not measure
88: the halo within $50''$, due to massive pileup in the ACIS-I detectors.
89: \citet{Draine03} and \citet{Xiang05} have both noted that this result
90: is therefore insensitive to dust near the source, as scattering from
91: dust within the last 25\% of the distance would lead to features
92: primarily within the excluded $50''$. To address this shortcoming, I
93: obtained a short \chandra HRC-I observation of GX13+1. The
94: multichannel plate design of the HRC-I is far less sensitive to large
95: count rates, which allows GX13+1's radial profile to be measured to
96: within 2'' of the source, far closer than previously possible.
97:
98:
99: %*********************************************************************
100:
101: \section{Observations}
102:
103: GX13+1 was observed simultaneously with the \chandra HRC-I and RXTE
104: Proportional Counter Array (PCA) on February 8, 2005 for 9.1 ksec
105: (ObsID 6093) and 6.2 ksec (P90173), respectively. CIAO v3.3 software
106: was used to process the \chandra data, which showed significant
107: background flares in addition to the flux from the bright source.
108: Standard processing was used for the RXTE data.
109:
110: \subsection{Selecting Good Events}
111:
112: The full-field lightcurve included significant periods when the count
113: rate approached the 184 cts s$^{-1}$\ telemetry limit. The expected
114: HRC-I background rate for the full field is $\sim 50$\,cts s$^{-1}$
115: \citep{POG06}. Despite the brightness of the source, the telemetry
116: saturation was in fact primarily due to the particle background.
117: After excluding a 2' radius circle around GX13+1, the average count
118: rate was $<50$\,cts s$^{-1}$, but with excursions above 100 cts
119: s$^{-1}$\ where telemetry saturation would affect the data. To
120: eliminate this problem only time periods where the total counts in the
121: field ({\it i.e}, $> 2'$\ from GX13+1) were $<45$\,cts s$^{-1}$\ were
122: included. Although this reduced the total good time to 3.58 ksec,
123: $\sim 200,000$\ counts were detected within $2'$ of GX13+1 for a
124: source count rate of 54.8 cts s$^{-1}$. All of these effects can be
125: seen in Figure~\ref{fig:ltcrv}, which shows the total HRC-I count
126: rate, the HRC-I count rate within $2'$\ of GX13+1, and the RXTE PCA
127: lightcurve (between 2-9 keV) during the observation. Within $2''$\ of
128: GX13+1 the count rate during these 3.58 ksec was 42.7 cts s$^{-1}$.
129: According to \S4.2.3.1 of the \chandra Proposer's Observatory Guide,
130: the encircled energy within $2''$\ is $\approx 90$\%, rising to
131: $\approx 95\%$\ within $10''$. Based only on these values, it appears
132: that $\sim 78$\% of the total counts are ``on-axis'' while 22\% are
133: scattered by a combination of interstellar dust and the \chandra
134: mirrors.
135: \begin{figure}
136: \includegraphics[totalheight=2.3in]{f1.eps}
137: \caption{GX13+1 lightcurve from HRC-I (red) and RXTE (in blue; divided
138: by 4 for clarity). The full HRC-I field is shown in black, and the
139: selected ``good time intervals'' are shown in green. When the full
140: HRC-I spikes much above 100 cts/s, the telemetry limit affects the source
141: count rate.\label{fig:ltcrv}}
142: \end{figure}
143:
144: The extremely high flux from the source combined with the desire to
145: get the highest possible spatial resolution required an unusual
146: instrument configuration. In collaboration with the CXC Operations
147: team, the HRC-I detector was positioned so that the source would
148: appear in one corner of the HRC-I, while still being on-axis to the
149: HRMA. This offset retained \chandra's spatial resolution but ensured
150: the source was far away from the normal aimpoint.
151: Figure~\ref{fig:image}[Left] shows the full field of the HRC-I, with
152: GX13+1 at one corner.
153:
154: \begin{figure*}
155: \includegraphics[totalheight=2.8in]{f2a.eps}
156: \hfill
157: \includegraphics[totalheight=2.8in]{f2b.eps}
158: \caption{[Left] The full HRC-I observation of GX13+1, with the
159: excluded rectangular region marked. [Right] Expanded image of GX13+1
160: on the HRC-I, showing detector ``jet'' and excluded region. The 2''
161: radius circle shows the near-source region excluded due to likely
162: detector non-linearity.\label{fig:image}}
163: \end{figure*}
164:
165: In Figure~\ref{fig:image}[Right], a ``jet''extending to the NE and
166: containing $\sim 1000$\,counts can be seen. This jet is a well-known
167: detector artifact \citep{Murray00} which is normally removed by the
168: standard processing to a level of $<0.1\%$\ of the total source flux
169: \citep{MurrayGhost}. In the case of GX13+1, this jet is $\sim 0.5\%$\
170: of the apparent source count rate. The most likely cause is the high
171: source count rate interfering with the the on-board electronic event
172: processing (Dr. Michael Juda, private communication). Although the
173: jet could be eliminated with aggressive filtering, this would also
174: invalidate the standard calibration. It was therefore decided to
175: simply ignore all events from the ``jet''-side of the source, as shown
176: by the box region in Figure~\ref{fig:image}[Left].
177:
178: \subsection{Extracting the Spectrum and Flux}
179:
180: The surface brightness of the X-ray halo must be normalized by the
181: source flux to make absolute measurements of the dust column density.
182: GX13+1 is observed almost constantly since it is a RXTE All-Sky
183: Monitor (ASM) source with a average rate of 20-30 cts/s. However, as
184: the RXTE ASM has little spectral sensitivity and the HRC-I has no
185: effective energy resolution, simultaneous RXTE PCA observations of
186: GX13+1 were taken to obtain a useful spectrum of the source. Although
187: the RXTE PCA itself has only moderate resolution and little
188: sensitivity below 2 keV, GX13+1's spectrum is dominated by 2-4 keV
189: photons (SES02). The PCA spectrum is shown in
190: Figure~\ref{fig:xtespec}[Left] as fit with a simple model consisting
191: of an absorbed multi-color disk model plus a blackbody, following
192: \citet{Ueda04}. The column density was fixed at the value found by
193: \citet{Ueda04} from the Chandra HETG, $_{\rm H} =
194: 3.2\times10^{22}$\,cm$^{-2}$, since this result is far more accurate
195: than one obtained from the PCA. The best-fit inner temperature of the
196: multicolor disk and the blackbody were 1.73 keV and 3.52 keV, with
197: absorbed 1-10 keV fluxes of $7.43\times10^{-9}$\,erg
198: cm$^{-2}$s$^{-1}$\ and $2.30\times10^{-10}$\,erg cm$^{-2}$s$^{-1}$,
199: respectively. Despite the large reduced $\chi^2_{\nu}$ ($>50$, driven
200: by systematic errors), this model is an adequate fit for this work
201: since only the total flux and the approximate spectral shape are
202: needed to calculate the predicted response of the \chandra HRC-I.
203: Nonetheless, it is important to note that fluxes measured with the
204: RXTE PCA are systematically high by 10-15\% in the 2-10 keV range,
205: compared to other X-ray observatories \citep{Jahoda05}.
206: %The measured flux from the HRC-I was therefore reduced by 15\% (from
207: %the RXTE value) to bring the count rates predicted by PIMMS in line
208: %with the observed values and to account for the discrepancy in the
209: %RXTE PCA calibration.
210:
211:
212: \begin{figure*}
213: \includegraphics[totalheight=2.2in]{f3a.eps}
214: \includegraphics[totalheight=2.2in]{f3b.eps}
215: \caption{[Left] The RXTE PCA spectrum of GX13+1, fit with an absorbed
216: multicolor disk plus blackbody model. [Right] The best-fit model
217: folded through the HRC-I response. The integrated count rate
218: predicted is 64.6 cts/s. \label{fig:xtespec}}
219: \end{figure*}
220:
221: To check the expected count rate, we folded this spectrum through the
222: HRC effective area file for on-axis Cycle 7 data ({\tt
223: hrciD2005-11-30pimmsN0008.fits}), as shown in
224: Figure~\ref{fig:xtespec}[Right]. The total predicted source count
225: rate in the HRC-I is 64.6 cts s$^{-1}$, $\sim 18$\% larger than the
226: observed HRC-I count rate of 54.8 cts s$^{-1}$\ within 2' of GX13+1.
227: % see makeplot_ueda.sl for details
228: The discrepancy is primarily due to the overestimation from the RXTE
229: PCA calibration, with an additional complication due to spatial
230: variation in the response of the HRC-I that reduces the effective area
231: of the detector corners relative to the center \citep{Donnelly03}
232:
233: The \chandra PSF, measured as a ratio of the surface brightness to the
234: source flux, is the background for this observation. The RXTE PCA, a
235: non-imaging detector, includes both the direct source flux and the
236: scattered halo photons, which must be removed to avoid
237: double-counting. However, as the goal is to measure the scattered halo
238: fraction itself, this problem is recursive. I addressed this by
239: assuming a column density of $3.2\times10^{22}$\,cm$^{-2}$\ and
240: calculating the total scattered fraction for the MRN77, WD01, and
241: ZDA04 BARE-GR-B models, weighted by the HRC-I response. The resulting
242: halo fraction ranged from 13-26\%. This predicted halo strength is
243: consistent with the result that 22\% of the total source counts are
244: between $2''-120''$. Therefore, for purposes of calculating the
245: background PSF, the RXTE PCA flux was reduced by 20\% to exclude the
246: halo contribution, with a 7\% systematic error. This reduction is in
247: addition to the 15\% reduction described above. The 7\% error is
248: likely not the dominant term in the systematic error, however. The
249: observation of a very bright source in one corner of the HRC-I
250: detector is at the extreme edge of the available calibration, and so
251: careful consideration of all uncertainties will be required.
252:
253: Another concern regarding this observation was that a significant
254: short-term change in the source flux, on the order of 12-24 hours,
255: would also affect the halo in a time-delayed manner \citep[\protect{\it
256: e.g.}][]{Vaughan04}. At smaller angles the delay could be even longer.
257: The RXTE ASM data was checked for a 10 day period before the
258: observation, but no strong or significant variation was seen.
259: Although Type I X-ray bursts have been seen from GX13+1 which show
260: $3-4\times$\ the normal flux, they only last $\sim 15$\,seconds
261: \citep{Matsuba95}. In this case, no bursts were seen in either the
262: HRC-I or PCA lightcurves, and indeed the halo observation would not be
263: sensitive to such a small variation.
264:
265: \subsection{Point-spread function\label{ss:psf}}
266:
267: An accurate measurement of the Chandra HRC-I point-spread-function
268: (PSF) between 2-100'' from the source is crucial to this observation.
269: An accurate raytrace model (ChaRT
270: \footnote{http://asc.harvard.edu/soft/ChaRT/cgi-bin/www-saosac.cgi}),
271: of the Chandra HRMA has been calibrated for near-source ($< 2''$)
272: photons
273: %\footnote{http://asc.harvard.edu/cal/docs/cal_present_status.html\#psf}
274: , SES02 showed that at large scattering angles this model
275: significantly underestimates the PSF, leading to substantial problems
276: in the analysis. Therefore, SES02 relied upon an ACIS-I observation
277: of Her X-1 as a PSF calibrator, but this source is affected by pileup
278: within $\sim 10$'' and therefore cannot be used in the 2-10'' range.
279:
280: As shown in Figure~\ref{fig:xtespec}[Right], the spectrum of GX13+1
281: peaks at $\sim 2$\,keV. The HRC-I's lack of spectral response means
282: that spectral differences between any calibration source and GX13+1
283: will lead to additional complications. The best possible calibration
284: source would be a bright, hard, and lightly-absorbed X-ray source
285: observed on-axis with the HRC-I. The X-ray binary LMC X-1 matches
286: these requirements reasonably well, and two \chandra observations
287: (ObsID 1200, 1201) of the source have been done. However, they were
288: both done early in the mission (August 1999) before the HRMA final
289: focus was set and are thus unsuitable. Since then, the brightest hard
290: X-ray source with little absorption and a known (albeit variable) flux
291: to be observed with the HRC-I is 3C273 (ObsID 461 on Jan 22, 2000).
292: Figure~\ref{fig:PSF_3C273} shows 3C273's surface brightness, divided
293: by its source flux, as observed with the \chandra HRC-I (excluding the
294: well-known jet region). The spectrum was taken from a \chandra HETG
295: observation done twelve days earlier (ObsID 459) which is well-fit by
296: an absorbed power-law with $\Gamma = 1.67\pm0.01$\ and $F_X(2-10{\rm
297: keV}) = (1.08\pm0.03)\times10^{-10}$\,ergs cm$^{-2}$s$^{-1}$. The
298: absorption column was fixed at the Galactic value, N$_{\rm H} =
299: 1.8\times10^{20}$\,cm$^{-2}$. The predicted HRC-I count rate (based
300: on the CXC PIMMS tool) for this spectrum is 8.7 cts/s, while the
301: actual source count rate was 26\% higher at 11 cts/s. As the source
302: is variable, this was taken as showing little change and the flux was
303: simply assumed to have increased by 26\% during the HRC-I observation.
304: A significant but undetected change in the spectral shape could also
305: cause this change in observed flux and might affect our results.
306: However, BeppoSAX observations of 3C273 over a period of 10 days
307: showed only small changes in the 2-10 keV spectral shape, with
308: $\Gamma$\ values ranging from $1.56\pm0.02$ to $1.64\pm0.02$\ while
309: the flux varied by 15\% \cite{Haardt98}. The potential systematic
310: error caused by uncertainty in the true spectrum of 3C273 during its
311: observation with the HRC-I is thus smaller than the error due to
312: unavoidable differences between the spectra of 3C273 and GX13+1.
313:
314: The core of the PSF of 3C273 was fit with a Gaussian term centered at
315: 0 with FWHM of $1.007_{-0.005}^{+0.004}$''\ and amplitude
316: $1676\pm42$\,arcmin$^{-2}$. In addition, the best-fit model included
317: two power-law terms with $\Gamma_1 = 4.06\pm0.05, \Gamma_2 =
318: 2.40_{-0.02}^{+0.01}$\ and amplitudes $A_1 =
319: (3.4_{-0.9}^{+1.3})\times10^{-7}, A_2 =
320: (2.12_{-0.09}^{+0.13})\times10^{-4}$\ arcmin$^{-2}$\ at 100''. The
321: particle and sky background was fit with a constant,
322: $(2.60\pm0.02)\times10^{-3}$\,arcmin$^{-2}$. As
323: Figure~\ref{fig:PSF_3C273} shows the fit is quite good over a large
324: range of surface brightnesses, with the somewhat large reduced
325: $\chi^2_{\nu} = 2.6$\ likely due to the extreme precision of the
326: measurement compared to the relatively simple model.
327: \begin{figure}
328: \includegraphics[totalheight=2.3in]{f4.eps}
329: \caption{[Top] The radial profile of 3C273's surface brightness,
330: divided by the source flux, fit with the sum of a Gaussian plus two
331: power laws and a constant. [Bottom] The ratio of the data and model,
332: showing small excursions at $1.5''$\ and $10''$, but generally good
333: agreement. \label{fig:PSF_3C273}}
334: \end{figure}
335:
336: \section{Results}
337:
338: The HRC-I observations are most useful between 2-100'', since beyond
339: that radius the ACIS-I data can measure the energy-resolved X-ray
340: halo. Therefore, the ACIS-I data were reprocessed (with CIAO 3.3) and
341: reanalyzed following the approach described in SES02 except as noted
342: below. Both the HRC-I and ACIS-I results were used in the final
343: analysis. We note that in reprocessing the ACIS data, the source flux
344: measurement, done via the CCD transfer ``streak'', was redone with a
345: better calibration and improved handling of the background subtraction
346: which resulted in an overall $\sim 15$\% decrease in the measured
347: source flux. The calibration changes include a spatially-varying
348: modification of order $\pm5\%$\ in the quantum efficiency uniformity
349: in CALDB 2.28, and an energy-dependent increase of up to 16\% in the
350: overall effective area which was added in CALDB v3.2.1.
351:
352: The data were fit using the CIAO fitting engine {\sl Sherpa}\ using
353: scattering models based on the exact Rayleigh-Gans (RG) approximation
354: \citep{SD98}. This model assumes the grains are spherical but uses
355: the energy-dependent optical constants rather than the Drude
356: approximation when calculating the scattering efficiency.
357: \citet{SD98} noted that the full Mie treatment is necessary for X-rays
358: $< 2$\,keV, since the RG overestimates the total scattering at low
359: energies. The HRC-I is sensitive to X-rays between 0.08-10 keV, with
360: peak efficiency between 0.7-2.0 keV. In all cases the halo model for
361: the HRC-I was calculated using an average value weighted by the
362: spectrum of GX13+1 and efficiency of the HRC-I. However, since
363: GX13+1's spectrum as observed by the HRC-I (see
364: Figure~\ref{fig:xtespec}[Right]) is dominated by photons with $E >
365: 2$\,keV, the use of the simpler RG treatment is justified.
366:
367: The initial analysis assumed the dust was smoothly distributed along
368: the line of sight. Unlike SES02, where the predicted PSF was
369: subtracted from the data, here the PSF was incorporated into the
370: fitting directly to allow for an explicit inclusion of uncertainty in
371: the PSF. Fits to the ACIS-I and HRC-I data included a constant factor
372: that allowed for calibration uncertainty in the overall PSF, caused
373: primarily by systematic errors in the source flux. For both the ACIS-I
374: and HRC-I data this multiplier was allowed to vary by up to 10\%. In
375: many cases the fit pushed the multiplier to an extremum of the range,
376: showing that systematic uncertainties remain in the data, although it
377: is not clear what component dominates them. In SES02, systematic
378: errors in the PSF manifested as energy-dependent column density fits,
379: since to first order an error in the PSF could be adjusted by changing
380: the overall halo scattering. As the total halo intensity is inversely
381: proportional to energy, this effect is often a linear dependence of
382: the best-fit N$_{\rm H}$\ on energy. To check for this, I allowed the
383: value of N$_{\rm H}$\ to vary independently in the HRC-I and each energy
384: band of the ACIS-I data. I used the F-test to determine that in only
385: one case (ZDA04 BARE-GR-B) were the best-fit ACIS-I N$_{\rm H}$\ values
386: better described by a linear energy-dependent model than by a constant
387: value. Even in this case, the F-test significance was only 3.5\%, a
388: negligible value given the number of different models tried. It seems
389: unlikely, therefore, that there is a significant error in the relative
390: power in the dust-scattered (halo) and mirror-scattered (PSF) photons.
391:
392: Table~\ref{tab:smofits}\ shows the best-fit N$_{\rm H}$\ results for
393: the HRC-I and the ``average'' ACIS-I value fit and the total
394: $\chi^2_{\nu}$\ assuming smoothly-distributed dust along the line of
395: sight for the MRN77, WD01, and the 15 ZDA04 models. These can be
396: compared to the value of $3.2\times10^{22}$\,cm$^{-2}$\ found by
397: \citet{Ueda04}. The ACIS-I column densities are $\sim 20\%$\ larger
398: than the HRC-I values, although the models with the lowest
399: $\chi^2_{\nu}$\ values tend to have the best agreement. The most
400: likely cause of this discrepancy is cumulative errors in the source
401: flux measurements combined with calibration differences between the
402: ACIS-I and HRC-I detectors.
403: % need to rerun smooth zda7
404:
405: \begin{table}[t]
406: \caption{Smooth Dust Model Parameters\label{tab:smofits}}
407: \begin{tabular}{llll}
408: \hline \hline
409: Model &N$_{\rm H}$(HRC)& N$_{\rm H}$(ACIS) & $\chi^2_{\nu}$ \\
410: &$10^{22}$\,cm$^{-2}$&$10^{22}$\,cm$^{-2}$& \\ \hline
411: MRN77 &$2.4\pm0.2$ &$2.85\pm0.05$& 2.0 \\
412: WD01 &$1.51\pm0.02$&$2.0\pm0.2 $ & 3.6 \\
413: BARE-GR-S &$2.7\pm0.2$ &$2.9\pm0.1 $ & 1.9 \\
414: BARE-GR-FG&$2.6\pm0.2$ &$3.00\pm0.04$& 1.9 \\
415: BARE-GR-B &$3.6\pm0.3$ &$3.4\pm0.4$ & 2.8 \\
416: BARE-AC-S &$2.5\pm0.2$ &$3.0\pm0.1$ & 2.1 \\
417: BARE-AC-FG&$2.5\pm0.1$ &$3.0\pm0.2$ & 2.2 \\
418: BARE-AC-B &$3.3\pm0.3$ &$3.60\pm0.05$& 2.2 \\
419: COMP-GR-S &$2.02\pm0.02$&$2.9\pm0.4$ & 4.9 \\
420: COMP-GR-FG&$2.23\pm0.04$&$3.0\pm0.3$ & 3.5 \\
421: COMP-GR-B &$3.0\pm0.2$ &$3.71\pm0.09$& 1.9 \\
422: COMP-AC-S &$2.41\pm0.02$&$3.7\pm0.6 $ & 7.0 \\
423: COMP-AC-FG&$2.67\pm0.02$&$3.8\pm0.5 $ & 5.0 \\
424: COMP-AC-B &$4.24\pm0.04$&$6.7\pm0.9$ & 5.7 \\
425: COMP-NC-S &$11.1\pm0.6$ &$13.5\pm0.9$ & 2.4 \\
426: COMP-NC-FG&$2.81\pm0.02$&$4.6\pm0.8$ & 8.6 \\
427: COMP-NC-B &$3.39\pm0.03$&$5.9\pm1.1$ & 11.7 \\ \hline
428: \end{tabular}
429: \end{table}
430:
431: Figure~\ref{fig:Smooth} shows the profile of the HRC-I data along with
432: the ACIS-I data at $2.5\pm0.1$\,keV, near the median energy of the
433: spectrum as observed by the HRC-I. This figure shows the level of
434: agreement between the HRC-I and ACIS-I data agree with each other in
435: the overlap region ($50-100''$), as well as the large ($> 20\times$)
436: difference in the HRC-I and ACIS-I backgrounds in the $500-1000''$\
437: region. The radial profile shown in Figure~\ref{fig:Smooth}\ is fit
438: assuming the line of sight (LOS) dust is ``smoothly-distributed'' and
439: has a composition and size distribution described by WD01 [Left], and
440: the ZDA04 BARE-GR-S [Right] models. Although both models agree with
441: the overall shape of the radial profile, the ratio plots show that the
442: WD01 model underestimates the ACIS-I data in the $300-500''$\ range,
443: while the BARE-GR-S model underestimates the HRC-I data in the
444: $10-50''$\ range.
445:
446: \begin{figure*}
447: \includegraphics[totalheight=2.3in]{f5a.eps}
448: \includegraphics[totalheight=2.3in]{f5b.eps}
449: \caption{[Left] The X-ray halo from smoothly-distributed WD01-type
450: grains fit to the radial profile of GX13+1's surface brightness
451: divided by the source flux. The HRC-I observations are fit with a
452: red line; the ACIS-I data (at 2.5 keV) with a blue line. The ratio
453: of the HRC-I data to the model is shown below with crosses and the
454: ACIS-I data with circles. [Right] Same, for the ZDA04 BARE-GR-S
455: model.
456: \label{fig:Smooth}}
457: \end{figure*}
458:
459: The smoothly distributed dust fits show small discrepancies that might
460: be indications of dusty molecular clouds along the line of sight.
461: These would appear as ``bumps'' in the profile whose position and
462: strength depends upon the relative distance to the cloud and its
463: column density. I therefore refit the HRC-I and ACIS-I data using a
464: two-component model that included a smoothly-distributed component
465: plus a thin cloud with variable position and column density; the cloud
466: is treated as a sheet with negligible thickness. Models using only a
467: single thin cloud with no smooth component were also considered; these
468: gave generally poor fits independent of the dust model used, and so
469: were abandoned. This is not unexpected since (a) SES02 was unable to
470: fit a single cloud using only the ACIS-I data and (b) GX13+1 is
471: reasonably near the Galactic center ($(l,b) = (13.5^{\circ},
472: 0.1^{\circ})$, $D = 7\pm1$\,kpc; \citet{Bandyopadhyay99}) where a
473: sightline dominated by a single cloud would be unusual.
474:
475: To reduce fit time, the column density of both the smooth component
476: and a cloud was fixed to be the same for all datasets, as was the
477: position of the cloud along the line of sight. While more realistic
478: than allowing the cloud column density to vary as a function of X-ray
479: energy, this has the effect of magnifying residual systematic errors.
480: As noted previously, the halo strength diminishes with energy while the
481: relative PSF strength increases which can create a trend in the
482: best-fit column density as a function of energy. However, since in
483: only one case out of fourteen was such a trend seen previously, it
484: seems unlikely that the systematic errors are driving the resulting
485: best-fit parameters in the smooth plus cloud model.
486:
487: The best-fit parameters for each model are shown in
488: Table~\ref{tab:clfits}. None of the fits are formally acceptable
489: ($\chi^2_{\nu}$\ ranges from 2.1 to 9.9), although some are clearly
490: better than others. In cases where the best-fit position is 0, the
491: 1$\sigma$\ upper limit is shown.
492:
493: \begin{table}[t]
494: \caption{Smooth Plus Cloud Model Parameters\label{tab:clfits}}
495: \begin{tabular}{lllll}
496: \hline \hline
497: Model &N$_{\rm H}$(smooth) &N$_{\rm H}$(cloud) & Relative & $\chi^2_{\nu}$ \\
498: &$10^{22}$\,cm$^{-2}$&$10^{22}$\,cm$^{-2}$& Position & \\ \hline
499: MRN77 &$2.60_{-0.05}^{+0.08}$&$0.69_{-0.03}^{+0.06}$&$<0.003$&2.4 \\
500: WD01 &$1.38\pm0.02$&$0.46\pm0.01$&$<0.001$ &3.5 \\
501: BARE-GR-S &$3.02\pm0.01$&$0.15\pm0.01$&$0.89\pm0.01$&2.1 \\ %4
502: BARE-GR-FG&$2.95\pm0.01$&$0.11\pm0.01$&$0.91\pm0.01$&2.2 \\ %5
503: BARE-GR-B &$3.51\pm0.01$&$0.57\pm0.01$&$0.82\pm0.01$& 2.4\\ %6
504: BARE-AC-S &$2.6\pm0.1$ &$0.21\pm0.06$&$<0.001$ &2.4 \\ %7
505: BARE-AC-FG&$2.6\pm0.1$ &$0.23\pm0.06$&$<0.001$ & 2.4\\ %8
506: BARE-AC-B &$3.4\pm0.1$ &$0.35\pm0.14$ &$0.85\pm0.01$& 2.1 \\ %9
507: COMP-GR-S &$1.6\pm0.1$ &$0.82_{-0.09}^{+0.07}$&$<0.0003$& 2.9\\ %10
508: COMP-GR-FG&$2.0\pm0.1$ &$0.62_{-0.08}^{+0.06}$&$<0.004$&2.6 \\ %11
509: COMP-GR-B &$3.2\pm0.1$ &$0.23\pm0.05$&$<0.01$&2.2 \\ %12
510: COMP-AC-S &$1.6_{-0.1}^{+0.2}$&$1.3\pm0.1$&$<0.0002$ &3.7 \\%13
511: COMP-AC-FG&$2.2_{-0.1}^{+0.2}$&$1.0\pm0.1$&$<0.0003$&3.3 \\ %14
512: COMP-AC-B &$2.9\pm0.2$ &$2.3\pm0.1$&$<0.0004$ &3.6 \\ %15
513: COMP-NC-S &$11.4_{-0.3}^{+0.6}$&$1.3\pm0.3$&$<0.0008$& 2.5\\%16
514: COMP-NC-FG&$1.6_{-0.2}^{+0.3}$&$1.9\pm0.2$&$<0.0001$ &4.3 \\%17
515: COMP-NC-B &$1.6_{-0.5}^{+0.9}$&$2.5_{-0.5}^{+0.4}$&$<0.00001$ &6.5 \\ \hline
516: \end{tabular}
517: \end{table}
518:
519: \section{Discussion}
520:
521: \citet{SES02} analyzed the ACIS-I observations of GX13+1's dust
522: scattered halo and found that the dust size distribution does not
523: extend to very large ($> 1\mu$m) grains and that grains do not have a
524: large vacuum fraction. However, the ACIS-I data could not distinguish
525: between the MRN77 and WD01 models, as the two distributions lead to
526: similar scattering profiles at large angles. Similarly, the data left
527: open the possibility that there might be a substantial population of
528: grains near the source \citep{Draine03,Xiang05}. These would create a
529: near-source scattered halo that was obscured by pileup in the ACIS-I
530: detector. The primary goal of the HRC-I observation was to remove
531: these uncertainties by measuring the halo near the source. This would
532: determine which dust model best fit the data, as well as detecting (or
533: put limits upon) variation in the dust distribution along the line of
534: sight.
535:
536: The fit results shown in Tables~\ref{tab:smofits} and \ref{tab:clfits}
537: contain a few surprises. Just as in \citet{SES02}, the WD01 model had
538: the smallest column density of any of the models when fit with either
539: smoothly distributed dust or after adding a cloud. However, the
540: overall result was a significantly worse fit than found with either
541: the MRN77 or many of the ZDA04 models. As Figure~\ref{fig:Smooth}
542: shows, the smooth WD01 model fits the HRC data well, but
543: underestimates the halo measured by ACIS between $150''-400''$, while
544: the ZDA04 BARE-GR-S model underestimates the halo measured by the HRC
545: between $10-50''$. Examining the other models show that these two
546: cases are representative. Adding a single dust cloud to the model
547: results in a solution with a cloud 70-90\% of the distance to the
548: source if the pure smooth model underestimates the halo around $30''$.
549: Conversely, adding a cloud component to smooth dust models that
550: underestimate the halo around $300''$\ tend to put the cloud near the
551: Sun.
552:
553: \begin{figure*}
554: \includegraphics[totalheight=2.2in]{f6.eps}
555: \caption{[Top] Radial profile of GX13+1 from the HRC-I (upper points) with
556: PSF calibration data from 3C273 (lower points). A close
557: approximation to the \citet{Xiang05} model for GX13+1 using the WD01
558: model is shown in blue, with the PSF fit to the 3C273 data is shown
559: in red. [Bottom] Ratio of GX13+1 data to \citet{Xiang05} model.
560: \label{fig:xiangwd01}}
561: \end{figure*}
562:
563: Although uncertainties remain due to calibration issues, the overall
564: quality of the fits shown in Table~\ref{tab:smofits} and
565: Figure~\ref{fig:Smooth} do not support the proposition that a
566: significant cloud of dust is present near GX13+1. In particular,
567: although it is not identical to the \citet{Xiang05} model,
568: Figure~\ref{fig:xiangwd01} shows the result from a model similar to
569: their fit to the zero-order HETG for GX13+1 using WD01-type dust.
570: This approximation to their model puts a cloud with N$_{\rm H} =
571: 2.7\times10^{21}$\,cm$^{-2}$\ at a position 99.5\% of the distance to
572: GX13+1, along with a smooth distribution with N$_{\rm H} =
573: 1.81\times10^{22}$\,cm$^{-2}$. In either their model or my
574: approximation, the cloud near the source dominates the variation in
575: the dust distribution. Figure~\ref{fig:xiangwd01} shows that although
576: the HRC-I data contain an obvious halo, this model is not a good fit
577: and, in fact, removing the cloud improves the fit. Although their
578: model fit can be improved somewhat with a small vertical offset
579: (corresponding to different relative flux calibrations of $\sim
580: 12\%$), the model still predicts 10-20\% more halo for $\theta <
581: 10''$\ than is observed. Table~\ref{tab:clfits} confirms this
582: conclusion, finding (in the case of WD01-type dust) a cloud near the
583: Sun, not the source, with N$_{\rm H} \sim 5\times10^{21}$\,cm$^{-2}$.
584:
585: ZDA04 described in detail how constraining dust models requires
586: combining multiwavelength data from the IR to X-rays while
587: simultaneously considering the metal abundances in the grains. Due to
588: the nature of optical/UV extinction and X-ray scattering, few sources
589: show strong signatures of dust in all of these wavebands
590: \citep{ValencicSmith07}. Nonetheless, it is possible to constrain the
591: allowed dust models by comparing the column density predicted by the
592: models to that measured using other techniques. In the case of
593: GX13+1, measurements of the column density range from
594: $2.5-4.0\times10^{22}$\,cm$^{-2}$\ \citep{CN92}. Optical measurements
595: provide only an upper limit of $2.9\times10^{22}$\,cm$^{-2}$\ based on
596: plausible but unconfirmed assumptions about the source
597: spectrum\citep{Garcia92}. The total \ion{H}{1} column density through
598: the Galaxy at the position of GX13+1 is $1.8\times10^{22}$\,cm$^{-2}$
599: \citep{DL90}, but this is misses the contribution from molecular
600: H$_2$\ that is likely to be substantial in the Galactic plane. The
601: HETG observation of GX13+1 agrees (weakly) with these results
602: \citep{Ueda04}, although it does not strongly limit it. Only Mg can
603: be directly measured ($N_{Mg} = 1.84^{+0.91}_{-0.49} \times
604: 10^{18}$\,cm$^{-2}$), equivalent to \NH $=
605: 4.8^{+2.4}_{-1.3}\times10^{22}$\,cm$^{-2}$\ assuming solar abundances.
606: The $2\sigma$\ upper limits for Si and S are equivalent to $\NH <
607: 4\times10^{22}$\,cm$^{-2}$. Of course, LMXBs have shown significant
608: variable internal absorption \citep{HG83} in X-rays, so this spectral
609: measurement sets at best an upper limit to the actual interstellar
610: component that is responsible for the halo.
611:
612: Despite these difficulties in independently measuring the total LOS
613: dust column density, we can reasonably justify excluding the value of
614: N$_{\rm H} = (1.11\pm0.06)\times10^{23}$\,cm$^{-2}$\ found in
615: Table~\ref{tab:smofits} for the ZDA04 COMP-NC-S model fit to the HRC-I
616: data. However, this was {\it only} model from ZDA04 that used
617: composite grains without bare carbon grains that had a plausible value
618: of $\chi^2_{\nu}$. Although more data are needed, this class of
619: models, along with the group of ``composite grains with bare amorphous
620: carbon'' models are clearly suspect since they do not generate an
621: X-ray halo similar to these observations. In fact, the only
622: smoothly-distributed ZDA04 composite grains model that fit with
623: $\chi^2_{\nu} < 3$\ had graphitic carbon and B star abundances
624: (COMP-GR-B). After adding a dust cloud to the model, the COMP-GR-B
625: model fit with $\chi^2_{\nu} = 2.2$\ while the next best fit
626: (excluding the unrealistic COMP-NC-S model) was COMP-GR-FG with
627: $\chi^2_{\nu} = 2.6$, a significantly worse fit.
628:
629: \section{Conclusions}
630:
631: The principal results from this analysis are:
632: \begin{enumerate}
633: \item Although challenging, HRC-I observations can be used to recover
634: the near-source region excluded by pileup in the ACIS-I detector. The
635: lack of energy resolution can be finessed if another measurement of
636: the source spectrum is available.
637: \item Fitting the source profile and background PSF independently
638: improves overall results, since calibration uncertainties in the
639: flux from the source and background objects can then be included
640: explicitly.
641: \item There is no strong signature of a dust cloud at the source in
642: the radial profile, as suggested by \citet{Xiang05}, although some
643: models include a cloud $\sim 90$\% of the distance to the source.
644: \item In agreement with SES02, the WD01 model underestimates the total
645: column density to the source, and again leads to poor fits, although
646: not so bad as to be excluded given the calibration uncertainties.
647: \item Some models from the ZDA04 paper, if not conclusively excluded,
648: are at least implausible. In general, the ZDA04 models with
649: composite grains (excepting the graphitic carbon model with B star
650: abundances) gave poor fits, while the bare carbon and silicate grain
651: models tended to fit well.
652: \end{enumerate}
653:
654: It should be noted that the relatively good fits found using the
655: simple smoothly-distributed dust model are somewhat surprising, since
656: X-ray halos probe both the largest grains whose size and composition
657: are the least constrained from observations in other wavelengths.
658: Additionally, X-ray halos are primarily observed through
659: highly-absorbed lines of sight. These probe dust in dense molecular
660: clouds that cannot be observed in the optical or UV due to the
661: extremely large extinction. Finally, all of these models assume
662: spherical grains, although recently some calculations have been done
663: on aspherical grains \citep{DraineAllaf06}\ that show the effects
664: are small ($<20\%$) for the WD01 model at $\sim 2$\,keV. Despite
665: these potential problems, a number of existing grain models agree
666: quite well with the observations, suggesting grains in dense clouds
667: (with the exception of the densest regions that take up very little
668: volume and may be optically-thick to X-rays) are not too dissimilar
669: from grains in less dense regions.
670:
671: \acknowledgments I thank Michael Juda for his substantial assistance
672: in arranging the observation and understanding the HRC-I calibration.
673: I would also like to thank Terry Gaetz and Diab Jerius for their
674: assistance in understanding the \chandra PSF, Lynne Valencic for
675: helpful discussions about dust, and Jingen Xiang for clarification of
676: his dust cloud models. Finally, my appreciation goes out to
677: Eli Dwek for many helpful discussions and for first bringing the issue
678: of X-ray scattering in dust to my attention. This work was supported
679: by the NASA Chandra observation grant GO5-6144X and by the NASA LTSA
680: grant NNG04GC80G.
681:
682: \begin{thebibliography}{}
683:
684: %\noindent Corbet,~R. 2003, ApJ, 595, 1086
685: \bibitem[Bandyopadhyay et al. (1999)]{Bandyopadhyay99}
686: Bandyopadhyay,~R.~M., Shahbaz,~T., Charles,~P.~A. \& Naylor,~T. 1999,
687: \mnras, 306, 417
688:
689: \bibitem[Chandra X-ray Center (2006)]{POG06} Chandra X-ray Center
690: Proposer's Guide, v 9.0 2006,
691: http://asc.harvard.edu/proposer/POG/html/HRC.html
692:
693: \bibitem[Charles \& Naylor (1992)]{CN92} Charles,~P.~A. \& Naylor,~T.
694: 1992, MNRAS 255, 6
695:
696: \bibitem[Dickey \& Lockman (1990)]{DL90} Dickey,~J.~M. \&
697: Lockman,~F.~J. 1990, \araa, 28, 215
698:
699: \bibitem[Donnelly, Brown \& Hole (2003)]{Donnelly03} Donnelly,~R.~H.,
700: Brown,~J.~P. \& Hole,~K.~T. 2003,
701: http://asc.harvard.edu/cal/Hrc/Documents/hrci\_qeu.ps
702:
703: \bibitem[Draine (2003)]{Draine03} Draine, B.~T.\ 2003, \apj, 598, 1026
704:
705: \bibitem[Draine \& Allaf-Akbari(2006)]{DraineAllaf06} Draine, B.~T.,
706: \& Allaf-Akbari, K.\ 2006, \apj, 652, 1318
707:
708: %\noindent Dwek,~E. 1997, ApJ, 484, 799
709:
710: \bibitem[Garcia et al. (1992)]{Garcia92} Garcia,~M.~R. et al. 1992,
711: \aj, 103, 1325
712:
713: \bibitem[Haardt et al. (1998)]{Haardt98} Haardt,~F. et al. 1998, \aap,
714: 340, 35
715:
716: \bibitem[Henke, Gullikson \& Davis (1993)]{Henke93} Henke,~B.~L.,
717: Gullikson,~E.~M. \& Davis,~J.~C. 1993, ADNDT, 54, 2
718:
719: \bibitem[Hertz \& Grindlay (1983)]{HG83} Hertz,~P. \& Grindlay,~J.
720: 1983, \apj, 275, 105
721:
722: \bibitem[Jahoda (2005)]{Jahoda05} Jahoda,~K. 2005,
723: http://astrophysics.gsfc.nasa.gov/xrays/programs/ rxte/pca/flux\_scale.pdf
724:
725: \bibitem[Landgraf et al. (2000)]{Landgraf00} Landgraf,~M.,
726: Baggaley,~W.~J., Gr\"un,~E., Kr\"ger,~H., \& Linkert,~G. 2000
727: J. Geophys. Res., 105, 10343
728:
729: \bibitem[Mathis \& Whiffen (1989)]{MW89} Mathis, J.~S.~\& Whiffen, G.\
730: 1989, \apj, 341, 808
731:
732: \bibitem[Mathis \& Lee (1991)]{ML91} Mathis, J. S. \& Lee, C. W.
733: 1991, \apj, 376, 490
734:
735: \bibitem[Mathis, Rumpl \& Nordsieck (1977)]{MRN} Mathis,~J.~S.,
736: Rumpl,~W. \& Nordsieck,~K.~H. 1977, \apj, 280, 425
737:
738: \bibitem[Murray et al.(2000)]{Murray00} Murray, S.~S., et al.\
739: 2000, \procspie, 4140, 144
740:
741: \bibitem[Murray (2000)]{MurrayGhost} Murray,~S.~S. 2000,
742: http://cxc.harvard.edu/cal/Hrc/Documents/ghosts.ps
743:
744: \bibitem[Matsuba et al.(1995)]{Matsuba95} Matsuba, E. \etal 1995,
745: PASJ, 47, 575
746:
747: %\noindent Mauche, C. W. \& Gorenstein, P. 1986, ApJ, 302, 371
748:
749: %\noindent Mauche, C. W. \& Gorenstein, P. 1989, ApJ, 336, 843
750:
751: %\noindent Overbeck, J. W. 1965, ApJ, 141, 864
752:
753: %\noindent Predehl, P. \& Schmitt, J. 1995, A \& A, 293, 889
754:
755: %\noindent Predehl, P. \etal 2000, A\&A, 357, L25
756:
757: %\noindent Rolf,~D.~P. 1983, Nature, 302, 46
758:
759: %\noindent Schnerr, R.~S. \etal 2003, A\&A, 406, 221
760:
761: \bibitem[Smith \& Dwek (1998)]{SD98} Smith,~R.~K. \& Dwek,~E. 1998,
762: \apj, 503, 831
763:
764: \bibitem[Smith, Edgar \& Shafer (2002)]{SES02} Smith,~R.~K.,
765: Edgar,~R.~J. \& Shafer,~R.~A. 2002, \apj, 581, 562
766:
767: \bibitem[van de Hulst (1957)]{vdH57} van de Hulst,~H.~C. 1957, {\it
768: Light Scattering by Small Particles} (New York: Dover)
769:
770: \bibitem[Weingartner \& Draine (2001)]{WD01} Weingartner,~J.~C. \&
771: Draine,~B.~T. 2001, \apj, 548, 296
772:
773: \bibitem[Witt, Smith \& Dwek (2001)]{WSD01} Witt,~A.~N.,
774: Smith,~R.~K. \& Dwek,~E. 20017, \apj, 50, L201
775:
776: \bibitem[Vaughan et al. (2004)]{Vaughan04} Vaughan,~S. et al. 2004,
777: \apj, 603, 5
778:
779: \bibitem[Ueda et al. (2001)]{Ueda01} Ueda,~Y. \etal 2001, ApJL, 556, L87
780:
781: \bibitem[Ueda et al. (2004)]{Ueda04} Ueda,~Y. \etal 2004, ApJ, 629, 305
782:
783: \bibitem[Valencic \& Smith (2007)]{ValencicSmith07} Valencic,~L.~A. \&
784: Smith,~R.~K.\ 2007, \apj, in press
785:
786: \bibitem[Xiang et al.(2005)]{Xiang05} Xiang, J., Zhang, S.~N.,
787: \& Yao, Y.\ 2005, \apj, 628, 769
788:
789: \bibitem[Zubko, Dwek \& Arendt (2004)]{ZDA04} Zubko,~V., Dwek,~E. \&
790: Arendt,~R. 2004, \apjs, 152, 211 (ZDA04)
791:
792: \end{thebibliography}
793:
794: \end{document}
795: