1: \documentclass[12pt,preprint]{aastex}
2: %\documentclass[10pt,preprint2]{aastex}
3:
4: \newcommand{\etal}{{\it et al. }}
5: \newcommand{\myemail}{yaoys@jet.uah.edu}
6:
7: \shorttitle{A New Method To Resolve X-ray halos}
8: \shortauthors{Yao et al.}
9:
10: \begin{document}
11:
12: \title{
13: A New Method to Resolve X-Ray Halos around Point Sources with Chandra Data
14: and Its Application to Cygnus X-1}
15:
16: \author{Yangsen Yao\altaffilmark{1,2}, Shuang Nan Zhang\altaffilmark{1,2}, Xiaoling Zhang\altaffilmark{1,2}, Yuxin Feng\altaffilmark{1,2}}
17:
18: \altaffiltext{1}{Physics Department, University of Alabama in Huntsville,
19: Huntsville, AL 35899}
20: \altaffiltext{2}{National Space Science and Technology Center, 320 Sparkman Dr., SD50, Huntsville, AL 35805 \\
21: \hspace{0.5in}
22: yaoy@email.uah.edu, zhangsn@email.uah.edu, zhangx@email.uah.edu, fengyx@jet.uah.edu}
23:
24:
25: \begin{abstract}
26: With excellent angular resolution, good energy resolution and broad energy
27: band, the Chandra ACIS is the best instrument for studying the X-ray halos
28: around some galactic X-ray point sources caused by the dust scattering of
29: X-rays
30: in the interstellar medium. However, the direct images of bright sources
31: obtained with ACIS usually suffer from severe pile-up.
32: Making use of the fact that an isotropic image could be
33: reconstructed from its projection into any direction, we can reconstruct
34: the images of the X-ray halos from the data obtained with the
35: HETGS and/or in CC mode.
36: These data have no or less serious pile-up and enable us to
37: take full advantage
38: of the excellent angular resolution of Chandra. With the reconstructed
39: high resolution images, we can probe the X-ray halos
40: as close as 1$''$ to their associated
41: point sources. Applying this method to Cygnus X-1 observed with Chandra
42: HETGS in CC mode, we derived an energy dependent radial halo flux distribution
43: and concluded that, in a circular region (2$'$ in radius) centered at
44: the point source:
45: (1) relative to the total intensity, the
46: fractional halo intensity (FHI) is about 15\% at
47: $\sim$1~keV and drops to about 5\% at $\sim$6~keV; (2) about 50\%
48: of the halo photons are within the region of a radius less than 40$''$; and (3) the spectrum of
49: the point source is slightly distorted by the halo contamination.
50: \end{abstract}
51: \keywords{dust, extinction --- X-rays: ISM --- X-rays: binaries ---
52: X-rays: individual (Cygnus X-1)}
53:
54: \section{Introduction}
55: Small-angle scatterings between X-rays and dust grains
56: in the interstellar medium (ISM) form halos, the diffuse emission around
57: X-ray point sources. The spectrum and the intensity distribution of
58: an X-ray halo
59: depend on the properties of the ISM along the line-of-sight,
60: i.e., the density distribution of the dust,
61: the grain size of the dust and the chemical composition of the grains
62: (Overbeck 1965; Hayakawa 1970; Martin \& Sciama 1970; Mathis, Rumpl \&
63: Nordsieck 1977; Predehl \& Klose 1996).
64: Because of the ISM absorption and scattering,
65: the X-ray energy band provides advantages over other
66: wave bands in studying the interstellar dusts.
67:
68: The existence of the dust scattering X-ray halos was first discussed by
69: Overbeck (1965) and was first observationally confirmed by Rolf (1983)
70: using the data of GX339-4 with
71: %Imaging Proportional Counter (IPC/Einstein).
72: IPC/Einstein.
73: %Using the High Resolution Imager (HRI/Einstein),
74: Using HRI/Einstein,
75: Catura (1983) and Bode \etal (1985) also confirmed the existence of
76: X-ray scattering
77: halos around point sources GX 3+1, GX 9+1, GX 13+1, GX 17+2 and Cygnus X-1.
78: The recent main results on X-ray halo studies were reported by
79: Predehl \& Schmitt (1995) by
80: systematically examining 25 point sources and 4 supernova remnants with
81: ROSAT observations.
82:
83: In all these investigations, the analyses were performed by
84: subtracting the point source surface brightness predicted by the point
85: source flux and the point spread function
86: (PSF) of the instrument from the observed
87: surface brightness, then comparing the inferred fractional halo intensity
88: (FHI) as
89: a function of the off-axis angle to those predicted by
90: different halo models in order to deduce the halo properties.
91: However, because the differences
92: between halo
93: profiles from different spatial distributions and dust models
94: are significant only in the core of the halo (Mathis \& Lee 1991),
95: the limited angular resolutions of previous instruments
96: (1$'$ for IPC/Einstein and 25$''$ for PSPC/ROSAT)
97: prevent the previous works from probing regions
98: close to the point sources, and make it difficult
99: to study the properties of the dust grains and
100: to distinguish
101: between various dust models observationally
102: (Predehl \& Klose 1996).
103:
104: The Advanced CCD Imaging
105: Spectrometer (ACIS) aboard the Chandra X-ray Observatory,
106: with its excellent angular resolution (0.5$''$ FWHM in PSF),
107: broad energy band ($0.2-10.0$~keV) and reasonably good energy
108: resolution ($E/\Delta E=10-60$), is the most
109: promising instrument to date in the X-ray halo study.
110: However, the timed exposure (TE) mode of ACIS
111: \footnote{For the Chandra instruments (ACIS, HETG etc.)
112: and observation mode (CC mode, TE mode etc.), please refer to
113: http://cxc.harvard.edu/proposer/POG.}, which is the mode
114: with two-dimensional (2-D) image, often suffers from severe pile-up
115: because of the long exposure time (0.2 to 10 seconds) per frame
116: (pile-up is caused by two or more photons impacting one pixel or several
117: adjacent pixels in a single frame);
118: a bright X-ray point source will ``burn''
119: a hole at the source position
120: because most photons from the source are lost.
121: It is difficult to estimate the source flux and
122: the halo profile near the point source (see, e.g., Smith, Edgar
123: \& Shafer 2002) in this case.
124: Because only bright sources can generate significant
125: X-ray scattering halos, and
126: observations of bright sources with ACIS/Chandra TE mode always
127: suffer from severe pile-up,
128: the excellent spatial resolution of Chandra
129: %'s mirror and ACIS
130: has not brought the anticipated breakthrough
131: to X-ray halo study.
132:
133: The pile-up can be avoided or lessened via reducing the number of
134: photons impacting one pixel or several adjacent pixels
135: in a single frame time.
136: The Continuous Clocking (CC) mode
137: uses a very short frame time, and the transmission grating spreads
138: photons in different energies to different pixels; they can provide
139: us with pile-up free or less piled-up data.
140: However, the CC mode data only provide one-dimensional (1-D)
141: intensity distribution.
142: The grating data, though provide
143: 2-D information, are already dispersed by the
144: grating instruments.
145: In this letter,
146: we propose a new method to reconstruct the X-ray scattering halos
147: associated with X-ray point sources from
148: the CC mode data and/or the transmission grating data.
149: This method enables us to probe
150: the halo intensity distribution in a broad energy band
151: as close as 1$''$ to the point source. After testing it
152: with the MARX simulation, we applied
153: this method to Cygnus X-1 observed with
154: the Chandra High Energy Transmission
155: Grating Spectrometer (HETGS, or ACIS with HETG) in CC mode.
156:
157: \section{METHOD AND SIMULATION}
158: After reflected by the Chandra mirror, X-rays
159: will be diffracted by the transmission grating (in one dimension)
160: by an angle $\beta$ according to the
161: grating equation,
162: \begin{equation}
163: p \sin\beta = m \lambda,
164: \end{equation}
165: where $p$ is the spatial period of the grating lines, $\beta$ is the
166: dispersion angle, $m$ is the grating order
167: ($0, \pm1, \pm2, \cdots$),
168: and $\lambda$ is the photon wavelength.
169: The zeroth order image ($m=0$) is the same as the direct image
170: except for a smaller flux, because some photons are diffracted
171: to higher orders.
172: For a mono-energy source, Chandra
173: transmission grating will detect exactly the same
174: intensity
175: distribution in its higher order images
176: as in its zeroth order image, as
177: long as the source size is not too large
178: (less than 3$'$), as shown in Fig.~\ref{method}.
179: If we project the secondary and higher order photons
180: to a line perpendicular to the grating arm, the projected 1-D image
181: will be the same as the projection of the zeroth order image, except for
182: some broadening caused by ``mis-aligned'' grating facets of the HETG
183: in the cross-dispersion
184: direction
185: (see following discussion).
186: Because the grating only diffracts photons along
187: the direction of the grating arm, the above projection is also
188: valid for sources with continuum spectra.
189: Usually the non-zeroth order grating images and the CC mode data
190: have no or much less pile-up;
191: either of them can be used to reconstruct the original image.
192:
193: If the flux of a point source plus its X-ray halo
194: is isotropically distributed and centered at the point source as
195: $F(r)$,
196: and the projection process described above
197: can be represented by a matrix operator $M(r, d)$,
198: then the projected flux distribution $P(d)$ is
199: \begin{equation}
200: P(d) = F(r)\times M(r, d),
201: \end{equation}
202: where $r$ is the distance from the centroid source position and
203: $d$ is the distance
204: from the projection center (refer to Fig.~\ref{method}).
205: The inverse matrix of the operator $M(r,d)$ exists,
206: and the original distribution can
207: be easily resolved.
208: We used numerical integration to approach the above
209: projection process and built a matrix to approximate
210: the integration and calculate $M(r,d)^{-1}$.
211:
212: %% please put fig 1 here
213:
214: To test our method, we produced with MARX 3.0
215: simulator\footnote{http://space.mit.edu/ASC/MARX}
216: a point source plus two disk sources to mimic a point source
217: with its X-ray halo observed with Chandra/HETGS
218: in TE mode. Using the intrinsic CCD energy
219: resolution, we extracted the photons in the energy range 1.0--1.5~keV
220: and performed the test in this energy band.
221: We projected the zeroth order photons
222: along an arbitrary direction to mimic the CC mode,
223: projected the MEG photons along the MEG grating arm,
224: and then multiplied these projected flux distribution with $M(r,d)^{-1}$
225: to resolve the flux distributions of the point source plus its halo.
226: The PSF for the grating data was obtained with the same procedure,
227: from a simulated point source. It is worth noting that MARX
228: simulator takes into account the alignment blurs due to ``mis-aligned''
229: grating facets, so the broadening effects
230: mentioned before will not affect the reconstruction results.
231: For the CC mode data, the simulated point source is used directly as the PSF.
232: The halo flux
233: was obtained by subtracting the corresponding PSFs from the flux of source plus
234: halo. The reconstructed halo surface brightness distribution is
235: consistent with the simulation input, as shown in Fig.~\ref{simulation},
236: so is the FHI, with the input value of 27.7\% and
237: the recovered value of 25.5\%.
238: We therefore conclude that the proposed method is
239: feasible to resolve the intensity distribution
240: of X-ray halos associated with point sources.
241:
242: %% please put fig 2 here
243:
244: \section{APPLICATION TO CYGNUS X-1}
245: Cygnus X-1, the first dynamically determined X-ray binary system
246: to harbor a black hole,
247: had been observed seven times with Chandra by 2003 February 14.
248: It is so bright that during each of the total four observations with TE mode,
249: there was significant piled-up not only in the zeroth order image
250: but also in the dispersed grating images. The only short-frame observation
251: (1999 October 19, ObsID 107) was almost pile-up free in the grating image,
252: but the statistical quality of the data was poor.
253: The other three observations used CC mode.
254: We applied our method to the one
255: with the highest statistical quality, which was observed on 2000
256: January 12 (ObsID 1511) with effective exposure of 12.7 ks.
257:
258: We used the zeroth order data within 2$'$ from the source position.
259: For energies above 3.0~keV, the grating arms extend into the
260: 2$'$ region; therefore we must exclude the photons resolved as grating
261: (non-zeroth order) events, which inevitably include some mis-identified
262: halo photons and leave a gap in the halo intensity distribution.
263: We interpolated across the gaps with an exponential function.
264: To estimated the pile-up in the zeroth
265: order data, we calculated the source flux from the grating arm and input it to
266: PIMMS\footnote{http://cxc.harvard.edu/toolkit/pimms.jsp} and obtained about
267: 23\% pile-up.
268: To estimate the point source flux more accurately, we compared
269: the grating data of the short-frame observation on 1999 October 19 with
270: the grating data of this CC mode observation, then used the read-out streak
271: of the short-frame to infer the point source flux at different energy bands
272: (with width 0.5~keV).
273: We also input the grating spectrum of this observation
274: to MARX (scaling up the frame time by three orders of magnitude
275: and scaling down the source flux accordingly to mimic the CC mode)
276: to obtain the piled-up PSFs for different energy bands.
277:
278:
279: %% please put fig 3 here
280:
281: The reconstructed flux distributions in two energy bands
282: are shown in Fig.~\ref{cygx1}. Even though there is pile-up
283: in the core region,
284: the halo can be clearly resolved
285: down to 1$''$ from the point source in the 0.5--1.0~keV band.
286: We also fitted the reconstructed image with the simulated piled-up PSF
287: and then estimated the halo flux.
288: The results of
289: these two methods agree with each other.
290: The total FHI within 2$'$ as a function of photon
291: energy is shown in Fig.~\ref{ratio}(a); the fraction drops from about 15\%
292: around 1~keV to about 5\% around 6~keV.
293: We define the half-flux radius $R_{0.5}$
294: of the halo as the radius which encloses half of the halo photons
295: in the 2$'$ region.
296: The half-flux radius $R_{0.5}$ as a function of energy is
297: shown in Fig.~\ref{ratio}(b);
298: clearly 50\% of the halo photons
299: are concentrated within 40$''$.
300: We also investigated how the halo contaminated the point source spectrum
301: in Cygnus X-1
302: (see Fig.~\ref{ratio}(c) and Fig.~\ref{ratio}(d)).
303: The halo spectrum
304: is softer than the point source spectrum, especially in the low energy band
305: (below 3~keV). Contributing only $\sim$ 10\% to the total brightness,
306: the X-ray halo in Cygnus X-1 does not distort the original
307: point source spectrum significantly.
308:
309: %% please put fig 4 here
310:
311: The accurate Chandra PSF is important in evaluating FHI.
312: We use MARX simulator instead of the CIAO tool $mkpsf$
313: to generate the PSFs, because the PSF library grids,
314: from which $mkpsf$ interpolates to get a PSF for
315: a certain off-axis angle and
316: a certain energy, are very coarse.
317: More realistic PSFs\footnote{http://cxc.harvard.edu/chart/index.html}
318: could be generated by producing PSFs of the mirror with
319: the SAOSAC ray-tracing code
320: and then projecting
321: the simulated rays onto the detector in MARX. The MARX HRMA model
322: is a simplified version of SAOSAC and
323: is proved sufficient for this study.
324:
325: We used two previous Chandra observations
326: to make consistency check of the PSF generated by the MARX simulator;
327: one is the the calibration observation of Her X-1 on 2002 July 1,
328: the other is the long-frame Chandra/HETGS
329: observation of Cygnus X-1 on 1999 October 19.
330: Her X-1 is a ``real'' point source and almost halo free.
331: We did a quick analysis of the Her X-1 data and found that beyond
332: 15$''$ pile-up is negligible. The energy dependent
333: radial flux of Her X-1 is used it as the PSFs at off-axis
334: angle 15$''$ to 120$''$; it gave the same results as the simulated PSFs.
335: For the core region of the PSF, we projected the read-out streak of
336: the Cygnus X-1 data and reconstructed a radial profile at $\sim$arcsec region,
337: which agrees with
338: the simulated PSF in the core region very well.
339:
340: The same Cygnus X-1 data were used to double check the inferred FHI.
341: Assume the half-flux radius $R_{0.5}$ does not change between observations,
342: we calculated the halo photon intensity from photons
343: between the half-flux radius
344: $R_{0.5}$ and 2$'$.
345: The source count rate was obtained
346: from the read-out streak.
347: The derived FHI is
348: plotted in panel (a) of Fig.~\ref{ratio}. It agrees well with the result
349: given previously in this section (shown in the same plot).
350:
351: \section{CONCLUSION AND DISCUSSION}
352:
353: In this letter, we propose a new method
354: to reconstruct the image
355: of an X-ray point source with its associated X-ray scattering halo
356: from the CC mode data and/or grating data,
357: which is then used to resolve the X-ray halo.
358: With this method and the high angular resolution of the Chandra Observatory
359: we are able to probe the intensity distribution
360: of the X-ray halos as close as 1$''$ to their associated point sources.
361: This method is tested with the MARX simulation
362: and applied to Cygnus X-1.
363:
364: The derived FHI is energy dependent,
365: but does not seem to follow the $E^{-2}$ law, which was observed by
366: Predehl \& Schmitt (1995) and Smith \etal (2002).
367: The discrepancy may be due to
368: the limited region we used (within 2$'$), because of
369: the following two reasons:
370: (1) the scattering optical depth
371: $\tau_{sca} \propto E^{-2}$ was derived by integrating over all solid angles
372: (Mathis \& Lee 1991), instead of the 2$'$ region we used;
373: and (2) for
374: single-scattering the mean scattering angle is related to
375: energy as $E^{-1}$ (Mathis \& Lee 1991), which means we under-estimated
376: low energy photons more than we did for high energy photons.
377: In the low energy band, the FHI we obtained
378: are reasonably consistent with
379: the value reported by Predehl \etal (1995) (11\% at the
380: ROSAT energy range 0.1--2.4~keV).
381:
382: The existence of the halo around a point source might distort the spectrum of
383: the point source.
384: Because the cross-section of the scattering process in the ISM is
385: energy dependent, the halo spectrum is different from the point source
386: spectrum (see Fig.~\ref{ratio}(c) and \ref{ratio}(d)).
387: Therefore if the instrument is unable to resolve
388: the point source from the halo,
389: it will obtain contaminated source spectrum;
390: this is the case for X-ray instruments prior to Chandra.
391: Many of the previous measurements of the continuum X-ray spectra
392: of galactic X-ray sources with significant
393: X-ray scattering halo may suffer from this problem.
394: Despite that in Cygnus X-1 system, the X-ray scattering halo only contributes
395: about 10\% to the total brightness and does not distort the original spectrum
396: significantly, systematic studies of the X-ray halo distribution in
397: the broad band should be carried out for other
398: X-ray sources with significant X-ray scattering halos,
399: before we can draw any conclusion on the significance
400: of the halo induced distortion to their X-ray continuum spectra.
401:
402: The spatial distribution of the dust grains is very important
403: in interpreting
404: the physical properties of the grains.
405: For single scattering, numerical
406: calculations show that
407: the core region of an X-ray halo is sensitive to the spatial
408: distribution of the scattering dust along the line-of-sight;
409: grains near the observer generate halos with flatter
410: profile near the core region, whereas grains close to the source create a peak
411: toward the central source; the wings, on the other hand,
412: are not sensitive to the dust distributions
413: (Predehl \& Klose 1996; Mathis \& Lee 1991).
414: Because we can resolve the X-ray halos as close as 1$''$ to
415: the point sources, it is possible to determine
416: the grain spatial distributions.
417: Even though the grain size distribution and
418: multiple-scatterings (especially for the low energy photons)
419: may reduce the differences between halo profiles from various
420: spatial distributions, this degeneracy can be easily resolved by drawing
421: the diagnostic diagram proposed by Mathis \& Lee (1991, Fig.~8).
422: The energy dependencies of the scattering optical depth
423: $\tau_{sca} \propto E^{-2}$ and the mean scattering angle
424: $\propto E^{-1}$, also make the broad band
425: energy-dependent FHI a good diagnostic of the
426: dust grains (Mathis \& Lee 1991).
427:
428: In this letter,
429: we did not fit the energy dependent behavior of the halo with any
430: halo model to constrain the physical properties of the dust grains. However
431: with our technique, different grain spatial distributions and
432: different gain models can be well distinguished with further studies
433: of ACIS/Chandra data. We will address these issues in our next work.
434:
435: \acknowledgments
436: Y. Yao thanks Allyn Tennant for useful discussions and insightful suggestions,
437: also thanks Daniel Dewey, Bruce Draine and the anonymous referee
438: for helpful comments
439: and suggestions.
440: This work was supported in part by NASA Marshall Space
441: Flight Center under contract NCC8-200 and by NASA Long Term Space
442: Astrophysics Program under grants NAG5-7927 and NAG5-8523.
443:
444: \begin{thebibliography}{}
445: \bibitem[Bode \etal(1985)] {bod85} Bode, M. F., Priedhorsky, W. C.,
446: Norwell, G. A. \& Evans, A. 1985, ApJ, 299, 845
447: \bibitem[Catura(1983)]{cat83} Catura, R. C. 1983, ApJ, 275, 645
448: \bibitem[Hayakawa \etal(1970)]{hay70} Hayakawa, S. 1970, Progr.
449: Theor. Phys., 43, 1224
450: \bibitem[Mathis \& Lee (1991)]{mat91} Mathis, J. S. \& Lee, C.-W. 1991,
451: ApJ, 376, 490
452: \bibitem[Mathis \etal(1977)]{mat77} Mathis, J. S., Rumpl, W. \& Nordsieck,
453: K. H. 1977, ApJ, 217, 425
454: \bibitem[Martin \etal(1970)]{mar70} Martin, P. G. \& Sciama, D. W. 1970,
455: Ap. Letters, 5, 193
456: \bibitem[Overbeck(1965)]{ove65} Overbeck, J. W. 1965, ApJ, 141, 864
457: \bibitem[Predehl and Klose (1996)]{pre96} Predehl, P. \& Klose,
458: S. 1996, A\&A, 306, 283
459: \bibitem[Predehl and Schmitt (1995)]{pre95} Predehl, P. \&
460: Schmitt, J. H. M. M. 1995, A\&A, 293 889
461: \bibitem[Rolf (1983)]{rol83} Rolf, D. P. 1983, Nature, 302, 46
462: \bibitem[Smith, Edgar \& Shafer (2002)]{smi02} Smith, R. K., Edgar,
463: R. J. \& Shafer, R. A. 2002, ApJ, 581, 562
464: \end{thebibliography}
465:
466: \clearpage
467:
468: %% Use the figure environment and \plotone or \plottwo to include
469: %% figures and captions in your electronic submission.
470: \begin{figure}
471: \plotone{f1.eps}
472: \caption{
473: The projection of the photons along the grating arm and the projection
474: of the photons in
475: zeroth order image along an arbitrary direction.
476: }
477: \label{method}
478: \end{figure}
479:
480: \clearpage
481:
482: \begin{figure}
483: \plotone{f2.eps}
484: \caption{
485: The reconstruction of the intensity distribution of
486: a simulated X-ray point source with halo
487: in the energy band 1.0--1.5~keV.
488: Panel~(a): zeroth order;
489: panel~(b): MEG negative orders.
490: The curve with diamond symbols
491: and the dashed line
492: are the projected photon distribution (cts/sec).
493: The triangle symbols are for the reconstructed halo
494: distribution (cts/sec/arcsec$^2$),
495: after subtracting the PSF of the point source.
496: The solid line is the halo distribution from the zeroth order image
497: of the simulated halo (no pile-up in the simulation).
498: }
499: \label{simulation}
500: \end{figure}
501:
502: \clearpage
503:
504: \begin{figure}
505: \plotone{f3.eps}
506: \caption{
507: The flux distribution of Cygnus X-1
508: and the PSF in two energy bands. In each panel, the
509: solid line indicates the projected flux distribution (cts/sec),
510: and the other lines are for flux distributions (cts/sec/arcsec$^2$).
511: Dashed line: source with halo;
512: dotted line: PSF;
513: dash-dotted line: piled-up PSF.
514: }
515: \label{cygx1}
516: \end{figure}
517:
518: \clearpage
519:
520: \begin{figure}
521: \plotone{f4.eps}
522: \caption{
523: Panel~(a): Fractional halo intensity (FHI).
524: Triangle symbols: result of our calculation;
525: diamond symbols: test result described in Section~3.
526: The dotted line indicates the best fit
527: $I(E) = (12.2\pm 0.6) (E/{\rm 1~keV}) ^{-0.52 \pm 0.05}$.
528: Panel~(b): the half-flux radius of the halo.
529: Panel~(c) spectra (cts/s/keV/$cm^2$), from top to bottom,
530: source with halo, the ``net'' point source, halo.
531: Panel~(d): the ratio of the normalized spectra.
532: Dotted line: source with halo to ``net'' point source;
533: dashed line: halo to ``net'' point source.
534: }
535: \label{ratio}
536: \end{figure}
537:
538: \end{document}
539: %%
540:
541: