1: %\documentclass[12pt,preprint]{aastex}
2: %\voffset 1.5cm
3:
4: \documentclass{emulateapj}
5: \voffset -1.3cm
6:
7: \begin{document}
8:
9: \newcommand{\chandra}{{\it Chandra}}
10: \newcommand{\swift}{{\it Swift}}
11: \newcommand{\xmm}{{\it XMM-Newton}}
12: \newcommand{\exosat}{{\it EXOSAT}}
13: %\newcommand{\vaverage1}{$-1150\,\pm\,200$\,km\,s$^{-1}$}
14: %\newcommand{\vaveragetwo}{$-740\,\pm\,100$\,km\,s$^{-1}$}
15: \newcommand{\vaverageone}{$-1286\,\pm\,267$\,km\,s$^{-1}$}
16: \newcommand{\vaveragetwo}{$-771\,\pm\,65$\,km\,s$^{-1}$}
17:
18: \title{The SSS phase of RS\,Ophiuchi observed with \chandra\ and \xmm\ I.\\
19: Data and preliminary Modeling}
20:
21: \author{J.-U. Ness\altaffilmark{1}, S. Starrfield\altaffilmark{1}, A.P. Beardmore\altaffilmark{2}, M.F. Bode\altaffilmark{3}, J.J. Drake\altaffilmark{4}, A. Evans\altaffilmark{5}, R.D. Gehrz\altaffilmark{6}, M.R. Goad\altaffilmark{2}, R. Gonzalez-Riestra\altaffilmark{7}, P. Hauschildt\altaffilmark{8}, J. Krautter\altaffilmark{9}, T.J. O'Brien\altaffilmark{10}, J.P. Osborne\altaffilmark{2}, K.L. Page\altaffilmark{2}, R.A. Sch\"onrich\altaffilmark{11}, C.E. Woodward\altaffilmark{6}}
22:
23: \altaffiltext{1}{School of Earth and Space Exploration, Arizona
24: State University, Tempe, AZ 85287-1404, USA: [Jan-Uwe.Ness, sumner.starrfield]@asu.edu}
25: \altaffiltext{2}{Department of Physics \& Astronomy, University of Leicester, Leicester, LE1 7RH, UK}
26: \altaffiltext{3}{Astrophysics Research Institute, Liverpool John Moores University, Birkenhead, CH41 1LD, UK}
27: \altaffiltext{4}{Harvard-Smithsonian Center for Astrophysics, 60
28: Garden Street, Cambridge, MA 02138, USA}
29: \altaffiltext{5}{Astrophysics Group, Keele University, Keele, Staffordshire, ST5 5BG, UK}
30: \altaffiltext{6}{Department of Astronomy, School of Physics \& Astronomy, 116 Church Street S.E., University of Minnesota, Minneapolis, MN 55455, USA}
31: \altaffiltext{7}{XMM Science Operations Centre, ESAC, P.O. Box 50727,
32: 28080 Madrid, Spain}
33: \altaffiltext{8}{Hamburger Sternwarte, Gojenbergsweg 112, 21029
34: Hamburg, Germany}
35: \altaffiltext{9}{Landessternwarte K\"onigstuhl, 69117
36: Heidelberg, Germany}
37: \altaffiltext{10}{Jodrell Bank Observatory, School of Physics \& Astronomy, University of Manchester, Macclesfield, SK11 9DL, UK}
38: \altaffiltext{11}{Universit\"ats-Sternwarte der
39: Ludwig-Maximilians-Universit\"at,
40: Scheinerstr. 1, 81679 M\"unchen,
41: Germany}
42:
43: \journalinfo{accepted 2007, May 8}
44:
45: \begin{abstract}
46: The phase of Super-Soft-Source (SSS) emission of the sixth recorded
47: outburst of the recurrent nova RS\,Oph was observed twice with
48: \chandra\ and once with \xmm. The observations were taken on days 39.7
49: (\chandra), 54.0 (\xmm), and 66.9 (\chandra) after outburst. We confirm
50: a $\sim 35$-sec period on day 54.0 and found that it originates from
51: the SSS emission and not from the shock. We discus the bound-free
52: absorption by neutral elements in the line of sight, resonance absorption
53: lines plus self-absorbed emission line components, collisionally
54: excited emission lines from the shock, He-like intersystem lines, and
55: spectral changes during an episode of high-amplitude variability.
56: We find a decrease of the oxygen K-shell absorption edge that can
57: be explained by photoionization of oxygen. The absorption component
58: has average velocities of \vaverageone\ on day 39.7 and of
59: \vaveragetwo\ on day 66.9. The wavelengths of the emission line
60: components are consistent with their rest wavelengths as confirmed
61: by measurements of non-self absorbed He-like intersystem lines.
62: We have evidence that these lines originate from the shock rather
63: than the outer layers of the outflow and may be photoexcited in
64: addition to collisional excitations. We found collisionally excited
65: emission lines that are fading at wavelengths shorter than
66: 15\,\AA\ that originate from the radiatively cooling shock.
67: On day 39.5 we find a systematic blue shift of
68: $-526\,\pm\,114$\,km\,s$^{-1}$ from these lines.
69: We found anomalous He-like f/i ratios which indicates either high
70: densities or significant UV radiation near the plasma where the
71: emission lines are formed.
72: During the phase of strong variability the spectral
73: hardness light curve overlies the total light curve when
74: shifted by 1000\,sec. This can be explained by photoionization of
75: neutral oxygen in the line of sight if the densities of order
76: $10^{10}-10^{11}$\,cm$^{-3}$.
77: \end{abstract}
78:
79: \keywords{stars: novae, stars: individual (RS\,Oph) --- X-rays: stars}
80:
81: \section{Introduction}
82:
83: RS\,Oph is a Symbiotic Recurrent Nova (RN) that had recorded
84: outbursts in 1898, 1933, 1958, 1967, 1985, and 2006 \citep[February
85: $12.83=$\,day\,0;][]{rsophdiscovery}. In previous outbursts RS\,Oph was
86: well-studied in the optical, and in 1985 it was followed extensively in
87: ultraviolet \citep[][and references therein]{shore96}, radio
88: \citep{padin85,hje86}, and X-rays \citep[six pointings by \exosat\
89: from day 55 to day 251 of the outburst,][]{mason87,obrien92}.
90:
91: RS\,Oph is a member of a small class of Cataclysmic Variables (CVs) in
92: which a white dwarf (WD) orbits a red giant secondary (M2III).
93: The orbital period is $455.72\,\pm\,0.83$\,days and, assuming a WD near
94: the Chandrasekhar
95: Limit and a low inclination ($i<40^{o}$), the red giant mass is of order
96: 0.5\,M$_\odot$ \citep{dobrKen94}. The distance to RS\,Oph has been well
97: determined by a variety of methods as $1.6\,\pm\,0.3$\,kpc \citep{bods87}.
98: The interstellar absorbing column,
99: N$_{\rm H} = (2.4\,\pm\,0.6)\times 10^{21}$\,cm$^{-2}$, was determined from
100: H\,{\sc i} 21\,cm measurements \citep{hje86} and is consistent with the
101: visual extinction ($E(B-V) = 0.73\pm0.1$) determined from IUE observations
102: in 1985 \citep{sni87}.
103:
104: \setcounter{footnote}{1}
105: \renewcommand{\thefootnote}{\fnsymbol{footnote}}
106: Just as in Classical Novae (CNe\footnote{The companion of the WD
107: is generally a main-sequence star instead of a giant.}), accretion of
108: material from the secondary causes an explosion on the WD surface.
109: However, in RNe the rate of mass accretion and the WD mass
110: are sufficiently large for outbursts to repeat on human timescales.
111: (The other members of this class are T\,CrB, V745\,Sco, and V3890\,Sgr).
112:
113: A RN evolves analogously to a CN, the major difference between the two
114: types of outburst arises from the presence of the red giant in the
115: binary system which completely changes the environment around the WD.
116: The evolution of the X-rays in RS\,Oph can be characterized by four phases:
117: \begin{itemize}
118: \item The explosion ejects material into its surroundings and produces
119: a strong shock moving into the wind and outer atmosphere of the red
120: giant and backwards into the ejecta
121: \citep[e.g.,][]{bode06,obrien06,sokoloski06}. The strength of the
122: shock depends on the kinetic energy of the
123: ejecta and the density of the medium into which the ejecta run.
124: \item While nuclear burning continues on the WD, the bolometric
125: luminosity is approximately constant \citep{gallcode74}.
126: Depending on the opacity due to electron scattering within the
127: expanding shell, the peak of the observed spectrum gradually shifts
128: from low energies (optical and UV) to soft X-rays \citep{gallcode74}.
129: In CNe this happens after a few weeks to months
130: \citep[e.g., V4743\,Sgr;][]{v4743}, and the spectrum resembles that of
131: the class of Super Soft X-ray Binary Sources \citep[SSS][]{kahab}. This
132: phase is therefore called the SSS phase.
133: \item After the end of nuclear burning on the WD, the SSS emission
134: decreases and RS\,Oph enters a phase in which a recombining plasma,
135: exhibiting emission lines from radiatively excited states, appears.
136: \item The final stage is that of the ejected material radiatively
137: cooling. Collisional bound-bound excitations are balanced by
138: radiative de-excitations, giving rise to emission lines (also called
139: the coronal approximation). In this way kinetic energy is effectively
140: converted into radiation.
141: \end{itemize}
142:
143: The first few weeks of the evolution of the X-ray emitting blast wave
144: in RS\,Oph were studied by \cite{sokoloski06} using the Rossi X-ray
145: Timing Explorer (RXTE) and \cite{bode06} using \swift. The shock wave was
146: also extensively studied at radio wavelengths by Very Long Baseline
147: Interferometry (VLBI) and MERLIN imaging observations
148: \citep{obrien06}. From these observations, those obtained with the
149: Hubble Space Telescope \citep{bode07} and with ground-based infrared
150: interferometers \citep[VLTI+AMBER][]{Chesneau07}, asymmetries and multiple
151: emission components have clearly been established that show that there is
152: jet-like ejection in addition to shell-like ejection of material.
153: In the 2006 outburst \chandra\ obtained one, and
154: \xmm\ obtained two observations of this phase
155: \citep[][Ness in prep.]{rsoph_iau1,drake06,2006IAUC.8695....2G}.
156:
157: The SSS phase commenced about 30 days into the outburst \citep{atel770}.
158: Early in the evolution, the energy output of a nova explosion
159: happens primarily at optical and UV wavelengths, because the high-energy
160: radiation produced by nuclear burning on the surface of the WD
161: will be scattered within the surrounding shell, leaving the atmosphere
162: as lower-energy radiation. When the surrounding shell becomes thinner
163: as a consequence of the expansion, the optical brightness decreases
164: while the X-ray brightness increases and originates from deeper within
165: the outflow. X-ray observations during this phase allow studies of the
166: plasma regions at the radial distance from the WD within the outflow
167: where the optical depth $\tau\approx 1$. The spectrum is a hot stellar
168: atmosphere whose peak energy and shape are dominated by the temperature
169: and by
170: interstellar and circumstellar absorption. In CNe, additional absorption
171: lines have been observed \citep[e.g., V4743\,Sgr,][]{v4743} which
172: originate from highly ionized species and are shifted and broadened
173: according to the dynamics of the observed plasma.
174:
175: X-ray observations of the 1985 outburst taken with \exosat\ after day 55
176: \citep{mason87} were modeled by \cite{obrien92} assuming the emission
177: to originate from the shocked wind. A direct comparison of the
178: shapes of the light curves taken with \exosat\ and with \swift\
179: in 2006 \citep{osborne06} show a remarkable resemblance, implying that
180: the spectra obtained in 1985 (at least in the case of the LE of
181: \exosat) were actually dominated by the SSS emission \citep{ness_suzaku}.
182:
183: We present three grating observations taken during the SSS phase.
184: Two observations were taken with \chandra\ \citep{rsoph_iau2} and one with
185: \xmm. We describe our method of extraction of the X-ray light curves and
186: X-ray spectra in Sects.~\ref{lcsect} and \ref{spec}, respectively. We
187: present our spectral analysis in Sect.~\ref{anal}, focusing on the
188: broad-band continuum spectrum in Sect.~\ref{continuum}, the resonance
189: absorption- and emission lines in Sect.~\ref{pcygsect}, emission lines
190: originating from the shock in Sect.~\ref{elinesect}, He-like
191: intersystem lines in Sect.~\ref{hesect}, and the spectral changes
192: during brightness variations detected during the first observation
193: in Sect.~\ref{variability}.
194: We discuss various possible interpretations in Sect.~\ref{disc}
195: and summarize our conclusions in Sect.~\ref{concl}.
196:
197: \section{Observations}
198:
199: \begin{table}[!ht]
200: \begin{flushleft}
201: \renewcommand{\arraystretch}{1.1}
202: \caption{\label{tab1}Grating observations during SSS phase}
203: %\vspace{-.1cm}
204: \begin{tabular}{lrllrl}
205: \multicolumn{2}{l}{2006 Date \hfill day$^{[a]}$}& Mission & Grating & ObsID & exp. time\\
206: start--stop & & & /detector & & (net)\\
207: \hline
208: March 24, 12:25:22 & 39.69 & \chandra & LETGS & 7296 & 10.0\,ksec\\
209: \multicolumn{2}{l}{--March 24, 15:38:20 \hfill 39.82} & & /HRC & &\\
210: April 07, 21:04:52 & 54.05 & {\it XMM} & \multicolumn{2}{l}{RGS1\hfill 0410180301} & 9.8\,ksec\\
211: \multicolumn{2}{l}{--April 08, 02:20:04 \hfill 54.27} & & RGS2& & 18.6\,ksec\\
212: April 20, 17:23:48 & 66.89 & \chandra & LETGS & 7297 & 6.5\,ksec\\
213: \multicolumn{2}{l}{--April 20, 20:27:57 \hfill 67.02} & & /HRC & &\\
214: \hline
215: \end{tabular}
216:
217: $^{[a]}$after outburst (2006 Feb 12.83UT)
218: \renewcommand{\arraystretch}{1}
219: \end{flushleft}
220: \end{table}
221:
222: During the SSS phase that started after day 26 (2006 March 10), three
223: grating observations were carried out by \chandra\ \citep{chandra} and \xmm\
224: \citep{xmm}. We summarize the dates of each observation, the respective days
225: after the outburst, instrumental setup, observation identification numbers
226: (ObsID), and the net exposure times in Table~\ref{tab1}. The first and the
227: last observations were
228: taken by \chandra\ with net exposure times of 10.0\,ksec and 6.5\,ksec,
229: respectively. As a consequence of the brightness of the source, telemetry
230: saturation (full buffer all the time) occurred, which reduced the effective
231: exposure times compared to the on-source times. Provisions were taken to
232: ensure that these
233: reduced spectra were correctly calibrated. We used the LETGS/HRC combination
234: \citep[Low Energy Transmission Grating/High Resolution Camera:][]{letg,hrc},
235: which is an imaging dispersion spectrometer. Higher dispersion orders are
236: not filtered out, but in our case contamination of the first-order spectrum
237: by higher-order photons is negligible because the SSS spectrum is
238: sufficiently narrow as to prevent an overlap of the dispersion orders.
239:
240: The second grating observation during the SSS phase was taken with \xmm\
241: (XMM) on day 54 after outburst (2006 April 7). XMM carries five X-ray
242: instruments, three low-resolution CCD detectors and two gratings
243: \citep[Reflection Grating Spectrometers RGS1 and RGS2:][]{rgs}. Due to
244: the unexpected brightness of the target, the allocated telemetry
245: rates were not sufficient to handle the large event rate. All instruments
246: were affected by telemetry saturation. This implies that the effective
247: exposure times were shorter than the on-target times. For the RGS2, the
248: net exposure time was 18.6\,ksec but only 9.8\,ksec in the RGS1 (total
249: on-target time was 18.9\,ksec). Since the telemetry losses do not depend
250: on the energy of the photons, they have no effect on the extracted
251: spectrum, except for the shorter net exposure time. However, the light
252: curves were strongly affected, since telemetry losses cannot be recovered.
253: The X-rays dispersed into the spectrum by reflecting off the RGS are recorded
254: with a strip of CCD detectors, and the wavelength range is 5--38\,\AA.
255: Unfortunately, two chips within these CCD arrays failed early in the mission,
256: and those portions of the dispersed spectrum that should be recorded by
257: these chips are lost. The wavelengths affected by the chip failure range
258: from 10.5--13.8\,\AA\ in the RGS1 and 20--24\,\AA\ in the RGS2. Since the
259: peak emission of our SSS spectrum ranges between 15\,\AA\ and $\sim
260: 30$\,\AA\ (see Sect.~\ref{spec} and Fig.~\ref{cmp_sss_rgs}), only the RGS1
261: gives us sufficient spectral information, and we concentrate primarily on
262: the RGS1 for spectroscopy. We use the RGS2 for light curve analyses.
263: Since the source spectrum has its peak in the middle of the RGS2 chip gap
264: (Fig.~\ref{cmp_sss_rgs}), the RGS2 collects fewer photons and is less
265: affected by telemetry losses. This may be the only example where the chip
266: gap is actually of use, providing us with the only useful light curve of
267: this observation.
268:
269: We present our extraction of the light curves in Sect.~\ref{lcsect} and that
270: of the spectra in Sect.~\ref{spec}. We used standard tools provided by the
271: mission-specific software packages SAS (Science Analsis Software, version
272: 7.0) and CIAO (Chandra Interactive Analysis of Observations, version 3.3.0.1).
273:
274: \subsection{Extraction and analysis of light curves}
275: \label{lcsect}
276:
277: \begin{figure}[!ht]
278: \resizebox{\hsize}{!}{\includegraphics{f1}}
279: \caption{\label{lc}Light curves of the \chandra\ LETG observations, days 39.6 and
280: 66.9, in time bins of 50\,sec. The light curves have been extracted from the
281: dispersed photons covering a wavelength range $\sim 6-50$\,\AA. For the first
282: observation we extracted separate spectra from the time intervals marked with
283: different shadings (see Fig.~\ref{tres}).}
284: \end{figure}
285: %
286: \chandra:
287: For the extraction of \chandra\ light curves we used the CIAO tool
288: {\tt lightcurve}, which extracts all photons within previously defined
289: source and background extraction regions on the detector. As source
290: extraction regions we chose two
291: polygons around the streaks of dispersed photons including both dispersion
292: directions and we used only the middle chip (i.e., $\lambda <50$\,\AA). The
293: background was extracted from adjacent regions and subtracted.
294:
295: \begin{figure}[!ht]
296: \resizebox{\hsize}{!}{\includegraphics{f2}}
297: \caption{\label{lc_rgs}RGS2 light curve (day 54.0) in 45.947-sec time bins
298: (10 times readout time, top) and periodogram (bottom).
299: A $\sim 35$-sec periodicity is found in
300: the RGS2 light curve that confirms observations of a similar period with
301: \swift\ reported by \cite{atel770}.}
302: \end{figure}
303: %
304: \xmm:
305: For the extraction of \xmm\ RGS2 light curve we used the SAS tool {\tt evselect},
306: which extracts all dispersed photons within standard source- and background
307: extraction regions. We extracted the RGS2 light
308: curve in four different bin sizes in time, all chosen to be multiples of
309: the readout time (4.5947\,sec) in order to search for periodicity.
310:
311: In Fig.~\ref{lc} we show the \chandra\ light curves in 50-sec time bins in units
312: of ksec after the start of each observation. In the observation of day 39.7,
313: the count rate increases rapidly by a factor of two within only
314: 1000\,seconds, remaining at a high count rate for about 3000\,seconds, then dips
315: down (a little bit lower than the start count rate), only to rise again with
316: an even steeper slope towards the end of the observation. In Fig.~\ref{lc_rgs}
317: we show the RGS2 light curve extract for day 54.0, which is much flatter than
318: that on day 39.7, but is not as flat as on day 66.9. For the latter light
319: curve we investigated the possibility of rapid variability, however, a
320: simple model assuming a constant emission level reproduces the measured
321: light curve with a reduced $\chi^2=0.94$, thus models including any
322: kind of variability cannot improve the fit by more than 68.3\%.
323:
324: We computed a power spectrum for all light curves, and for days 39.7 and
325: 66.9 we find no significant periodic variations. In order to search for
326: periods shorter than 50\,sec, we also extracted the \chandra\ light curves
327: in smaller time bins, but did not detect any periods longer than 2\,sec.
328: In the bottom panel of Fig.~\ref{lc_rgs} we show the result from the
329: period search of the RGS2 light curve. In this observation we find a
330: 34.8-sec period that is consistent with the $\sim 35$-sec period reported
331: by \cite{atel770}. We also found a period of $\sim 5.3$\,sec, which is
332: the beat period of the instrument readout time (4.5947\,sec) and the
333: 34.8-sec period. We extracted the light curves from each chip and found
334: this period only on those chips that recorded the SSS spectrum.
335:
336: Our re-analysis of the \swift\ light curves obtained on days $\approx 39.7$,
337: 54.0, and 66.9 indicated that the $\sim 35$\,sec periodicity is only
338: present on day 54.0. \cite{atel801} reported that this periodic
339: variability was not always present, disappearing entirely after day 63 …
340: (2006, April 17).
341:
342: %\begin{figure*}[!ht]
343: %\resizebox{\hsize}{!}{\rotatebox{270}{\includegraphics{f3}}}
344: %\caption{\label{swiftlc}\swift\ XRT light curve taken from \cite{osborne06}
345: %with the light curves of the grating observations added in the insets.}
346: %\end{figure*}
347:
348: We compared the emission levels in the grating spectra with those
349: of \swift\ at the same times. The first observation was taken during
350: the phase of highly variable soft X-ray flux reported in
351: \cite{atel764,atel770}. This phase lasted from days 29 to 46
352: \citep{atel801} and is thus coincident with the highly variable light
353: curve taken with \chandra\ on day 39.7. Since \swift\ light curves
354: are never continuous over more than at most 2\,ksec due to its orbit,
355: it is the \chandra\ light curve that demonstrates that the source was
356: actually variable on time scales shorter than 0.1 days.
357:
358: Close inspection of the \swift\ monitoring light curve
359: \citep{osborne06} revealed that the
360: first \chandra\ observation (day 39.7) was taken at a time when
361: the \swift\ count rate was close to one of the minima during the
362: variability phase, while the second \chandra\ observation
363: (day 66.9) was taken during the decline shortly after the peak
364: emission level had been reached. This explains the higher
365: \chandra\ count rate on day 66.9 compared to day 39.7.
366: The \xmm\ observation was carried out during a phase of higher emission.
367: We compared the relative emission levels between days 39.7,
368: 54.0 and 66.7 by computing the photon fluxes by integration of the
369: grating spectra (see Sect.~\ref{spec}). Within the uncertainties
370: of the flux variations on day 39.7, the changes are consistent
371: with the changes of the corresponding \swift\ count rates.
372:
373: \subsection{Extraction of spectra}
374: \label{spec}
375:
376: Grating spectra are extracted on an equidistant wavelength grid (in units
377: of \AA), and we use wavelength units throughout this paper. We extracted
378: the spectra with the \chandra\ CIAO tool {\tt tgextract} for the LETGS
379: observations and the {\it XMM} SAS tool {\tt rgsproc} for the RGS
380: observation. These routines place standardized extraction regions over the
381: dispersed photons on the detector and are optimized to maximize the ratio
382: of collected source counts to the included background. The mirror point
383: spread function is translated into an instrumental line profile that is
384: approximately Lorentzian for the RGS and more Gaussian-like for \chandra\
385: LETG (both FWHM$\sim 0.055$\,\AA\ at all wavelengths, corresponding to
386: 500--1100\,km\,s$^{-1}$ from 30\,\AA\ to 10\,\AA, respectively).
387: Velocities below these values that contribute to line broadening are
388: difficult to determine. However, velocities from line shifts can be
389: determined accurately (see Table~\ref{tab_profiles} in
390: Sect.~\ref{pcygsect}). Using standard tools to calculate the effective
391: areas (in cm$^2$) and then dividing a count rate by the effective area
392: at the corresponding wavelength results in a photon flux that is
393: sufficiently independent of the instrument (RGS vs. LETGS). Thus, the
394: photon flux spectra from the RGS1 and the LETGS can be compared
395: (see Fig.~\ref{cmp_sss_rgs}).
396:
397: \chandra:
398: The dispersed photons are recorded in two streaks in opposite directions
399: from the zero-th order, delivering two independent spectra. We co-added
400: these two spectra for best signal to noise. In addition, independent
401: analyses of the spectra can be carried out for consistency checks. In
402: Fig.~\ref{cmp_sss_rgs} we present the two \chandra\ observations (light
403: and dark shadings) along with the {\it XMM} observation (solid line).
404:
405: \xmm:
406: RGS data were processed with the SAS task {\tt rgsproc} in SASv7.0
407: up to the creation of the merged, filtered event file. This file was
408: manipulated to correct for pileup. Given the high flux of the source,
409: spectra were extracted
410: from the full field of view instead of using standard extraction regions,
411: and no background subtraction was applied. The separation
412: of spectral orders was accomplished by using the energy resolution of the CCDs.
413: Pileup occurs when two or more events arrive at the same
414: (or neighboring) pixel during the same readout frame. These photons are
415: registered as a single event with an energy that is the sum of the energies of the
416: individual events. In our case, pileup results in events with CCD-measured
417: energies $n\times E$ (where $n=1,2,3,4$) while the photons are dispersed
418: according to their individual wavelengths. Due to this pile-up, counts occur in
419: the energy-dispersion plane in regions normally associated with higher orders.
420: There is no ambiguity between pile-up and higher order dispersion
421: due to the narrow spectral range of the super-soft continuum spectrum.
422: We were thus able to reconstruct the true spectrum before pile-up by adding those
423: events in the energy-dispersion plane back into the first order spectrum that
424: resulted from pile-up in the first-order spectrum. The rates of the new
425: first-order spectra are about 30\,percent higher than before this correction.
426:
427: \begin{figure}[!ht]
428: \resizebox{\hsize}{!}{\includegraphics{f3}}
429:
430: \caption{\label{cmp_sss_rgs}
431: Comparison of all three grating spectra in units of photon fluxes obtained by
432: dividing count rates by effective areas. The \chandra\ spectra are shown with
433: shading and the RGS1 spectrum is a thick solid histogram line.}
434: \end{figure}
435:
436: For a qualitative comparison of the spectra from the different missions we
437: converted all three count rate spectra into photon flux spectra. We do not
438: correct for the redistribution matrix, so the line profiles will still
439: depend on the individual instrumental point spread function (PSF).
440: However, the shape of the continuum is not affected by the instrumental
441: PSF, as it is very small compared to the width of the continuum.
442:
443: In Fig.~\ref{cmp_sss_rgs} we compare the photon flux spectra of the three
444: grating observations. The integrated fluxes relative to each other are
445: consistent with the relative count rates \swift\ measured at the respective
446: times. All spectra show continuum emission over the same wavelength range
447: with similar shapes. At 22.83\,\AA\ there is a strong absorption edge in all
448: spectra that is clearly non-instrumental and originates from O\,{\sc i}
449: (K-shell ionization). Also, an expected narrow $1s-2p$ absorption line
450: at 23.5\,\AA\ \citep{paerels01_abs} from atomic oxygen can be identified
451: in all three spectra. At $\sim 29$\,\AA\, there are emission line features
452: in all spectra that we attribute to N\,{\sc vi} (see Sect.~\ref{hesect}).
453: Strong absorption
454: lines can be identified in all three spectra (see Sect.~\ref{pcygsect}).
455:
456: \section{Spectral Analysis}
457: \label{anal}
458:
459: \subsection{Continuum emission and broad-band absorption}
460: \label{continuum}
461:
462: In order to understand the cause of the brightness changes we
463: computed a series of blackbody models and found reasonable
464: agreement with the measured spectra for temperature ranges
465: $(630-830)\times10^3$\,K on day 39.7,
466: $(650-710)\times10^3$\,K on day 54.0, and $(590-720)\times10^3$\,K
467: on day 66.9. The bolometric luminosities, $\log(L_{\rm bol})$, of
468: these models are between 37.3 and 38.5 for day 39.7, 38.5 and 38.9
469: for day 54.0, and 38.2 and 39.3 for day 66.9. We used a standard
470: model for interstellar plus circumstellar absorption. The value
471: of hydrogen column density N$_{\rm H}$ was greater than the
472: interstellar value (N$_{\rm H}=2.4\times10^{21}$\,cm$^{-1}$)
473: for all models (more below).
474: However, the values of reduced $\chi^2$ ($\chi^2_{\rm red}$) are
475: greater than 20 for all models, and no secure conclusions can be
476: drawn from these numbers. In particular, we don't claim from these
477: models that Super-Eddington luminosities occurred. With these
478: limitations we are not able to explain the
479: brightness changes in terms of temperature or luminosity.
480:
481: As a first approach to characterize the shape of the continuum, we
482: computed hardness ratios HR=(H-S)/(H+S) from the three spectra with
483: H and S denoting the fluxes extracted from within the wavelength ranges
484: 15--23\,\AA\ and 23--30\,\AA, respectively. We found values of
485: -0.06, +0.12, and +0.03 with 4\%, 1\%, and 8\% uncertainties for
486: days 39.7, 54.04, and 66.9, respectively. Changes in hardness generally
487: imply a change in temperature, but changes in broad-band absorption
488: (bound-free transitions) by elements in the line of sight can also lead
489: to changes in spectral hardness. We estimated the optical depth of the
490: O\,{\sc i} absorption edge at 22.83\,\AA\ from the intensities of
491: the continuum,
492: $\tau=\ln({\rm cont}_{\lambda<22.8}/{\rm cont}_{\lambda>22.8})$,
493: and found $\tau=0.9,\ 0.5,$ and $0.7$ for days
494: 39.7, 54.0, and 66.9, respectively. This implies that the larger
495: hardness ratio on day 54 was caused by reduced O\,{\sc i} within
496: the material in the line of sight.
497:
498: In order to investigate this effect closer, we use the blackbody
499: models from above as continua and apply an absorption model that
500: accounts for variations of the neutral oxygen abundance. We optimized
501: the blackbody parameters temperature and bolometric luminosity
502: simultaneously with the parameters of the absorption model.
503:
504: The continuum spectrum produced by the WD atmosphere has to pass
505: through the circumstellar and the interstellar material in the line
506: of sight. The former may be partially ionized while the latter consists of
507: neutral elements of solar composition. The bound-free absorption imposed
508: by this material leads to element-specific absorption edges that are
509: detectable with the LETGS \citep[e.g.,][]{paerels01_abs}.
510: The depth of these edges depends on the
511: elemental composition and absorption cross sections, e.g. for K-shell
512: absorption by neutral oxygen. Absorption edges from higher ionization stages
513: also depend on the fractional number density of the ion. We parameterize
514: bound-free absorption in the line of sight by the column density of neutral
515: hydrogen, N$_{\rm H}$. While the column density of the interstellar material
516: is assumed constant in all observations, it may vary with time in the
517: circumstellar material. We, therefore, split the absorption into two
518: terms, one term with fixed N$_{\rm H}$ and one with variable N$_{\rm H}$,
519: both with solar composition \citep{grev}. We computed the total
520: transmission coefficients as implemented in the software package PINTofALE
521: \citep{pintofale} from the single parameter N$_{\rm H}$. Using this model we
522: account only for absorption from the neutral component of the absorbing
523: column while absorption by ionized elements is included only for helium.
524:
525: \begin{figure*}[!ht]
526: \resizebox{\hsize}{!}{\includegraphics{f4}}
527: \caption{\label{bbb}Three grating spectra of RS\,Oph in
528: chronological order from top to bottom with a continuum
529: model and two different absorption models applied to each
530: case. The standard absorption models (black solid line) assume
531: solar abundances for the neutral elements in the line of
532: sight. For the second models (grey dashed) we allowed the abundance
533: of neutral oxygen to vary. For all observations, both models
534: have large $\chi^2_{\rm red}$, however, the assumption of
535: reduced neutral oxygen leads to an improvement of the
536: model.
537: }
538: \end{figure*}
539:
540: In Fig.~\ref{bbb} we present the three spectra from the
541: gratings (in photon flux units) in comparison with our best-fit
542: continuum plus absorption models. First, we optimized the
543: blackbody parameters and the value of N$_{\rm H}$ that
544: represents the neutral component of the circumstellar
545: material with the abundances
546: fixed at solar \citep{grev}. In the upper right corner
547: of each panel we give the value of $\chi^2_{\rm red}$.
548: These models are identical to the ones described above,
549: and are formally unacceptable. The blackbody continuum
550: seems to represent the rough observed spectral shape.
551: However, the high resolution of the
552: gratings clearly exposes the shortcomings of blackbody
553: models with many spectral features not being reproduced.
554:
555: Next, we varied the abundance of the oxygen content within
556: the circumstellar component, while the interstellar oxygen
557: abundance remained fixed at solar. We fit the oxygen
558: abundance simultaneously with the blackbody parameters and
559: N$_{\rm H}$ and include the best-fit models with grey lines in
560: Fig.~\ref{bbb}. Although, these models are also formally
561: unacceptable, an improvement in $\chi^2_{\rm red}$ is
562: apparent. The resulting best-fit abundances suggest that
563: after day 54 all neutral oxygen in the circumstellar material
564: has disappeared. We note that for day 66.9 the optical depth
565: around the edge was higher than on day 54.0, suggesting that the
566: neutral oxygen abundance has increased again. The model for
567: day 66.9 does not show this increase, but the peak is not
568: well modeled. The blackbody temperatures of these models are
569: $(540-760)\times10^3$\,K on day 39.7,
570: $(520-580)\times10^3$\,K on day 54.0, and $(440-590)\times10^3$\,K
571: on day 66.9. The bolometric luminosities, $\log(L_{\rm bol})$, of
572: these models are between 37.6 and 39.2 for day 39.7, 39.3 and 40.0
573: for day 54.0, and 38.9 and 40.5 for day 66.9. Comparison with the
574: parameters of the models with the oxygen abundance fixed at solar
575: shows that the introduction of the oxygen abundance as a free
576: parameter introduces considerable additional uncertainty in the
577: blackbody parameters.
578:
579: The simplest explanation for the reduction of neutral
580: oxygen is photoionization of the circumstellar material
581: by the radiation field. We tested this hypothesis and computed
582: models with reductions of other elements that would also
583: have to be ionized (nitrogen and carbon with their
584: K-shell absorption edges at 30.25\,\AA\ and 43.63\,\AA,
585: respectively). We were not able to find a model that
586: reproduced the measured spectrum better than the
587: one with only oxygen reduced. Next, we carefully inspected the
588: grating spectra for evidence of any absorption edges at the
589: ionization energies of O\,{\sc vii} (16.77\,\AA),
590: O\,{\sc viii} (14.23\,\AA), N\,{\sc vi} (22.46\,\AA), and
591: N\,{\sc vii} (18.59\,\AA). In all three spectra, there is no
592: evidence for these absorption edges. Since the
593: continuum level at these wavelengths is sufficiently high, and
594: the instruments are sensitive enough to detect these edges,
595: we conclude that no absorption edges from H-like and He-like
596: ions are present. This indicates that the material that
597: produces the absorption lines from high ionization stages
598: (see next section) contributes little bound-free
599: absorption and resides deeper within the outflow.
600:
601: \cite{paerels01_abs} pointed out that the detailed structure
602: of the spectrum around the O edge is extremely complex,
603: consisting of absorption lines and -edges for various ions
604: and molecular species. For example, we studied the O\,{\sc i}
605: $1s-2p$ absorption line at 23.62\,\AA. On days 39.7 and 66.9
606: we found this line at the same wavelength of 23.51\,\AA\ and
607: with the same optical depth at line center of $\tau_c=1.4$.
608: This similarity suggests that this line was formed in a
609: non-changing medium (interstellar medium). The shift of this
610: line is similar to shifts 30-50\,m\AA\ found by
611: \cite{juett04} for various features from singly and doubly
612: ionized oxygen
613: in HETG spectra of seven X-ray binaries which are consistent
614: with discrepancies between theoretical calculations and
615: laboratory measurements. It is thus likely that the
616: 23.51-\AA\ line is the O\,{\sc i} $1s-2p$ line at the
617: rest wavelength. We also found evidence for the $1s-2p$
618: absorption lines of O\,{\sc ii} to O\,{\sc v}, expected
619: at 23.3\,\AA, 23.11\,\AA, 22.78\,\AA, and 22.33\,\AA,
620: respectively (see also Table~\ref{alines}), which indicates
621: that we are not only dealing with atomic oxygen.
622:
623: \cite{juett04} have developed detailed models for the
624: wavelength range around the oxygen K-shell absorption edge.
625: These models allow determinations of the ratios of ionization
626: stages and oxygen column densities and lead to a more accurate
627: bound-free absorption model needed for further analyses.
628: Similar analyses have been carried out by \cite{page03},
629: however, in no other source is the oxygen edge as deep as
630: in RS\,Oph.
631:
632: Alternatively to ionization, \cite{paerels01_abs} propose
633: that grain formation could reduce the O edge. In the
634: event of dust formation in the outer regions
635: of RS\,Oph, oxygen could have been locked up in grains.
636:
637: \subsection{Modeling of resonance absorption and emission lines}
638: \label{pcygsect}
639:
640: \begin{figure*}[!ht]
641: \resizebox{\hsize}{!}{\includegraphics{f5}}
642: \caption{\label{profiles}Modeling of resonance absorption- and
643: emission lines detected on days 39.7 (left panels) and 66.9
644: (right panels) for four prominent transitions (labeled in bottom
645: right). The models were calculated as described in Sect.~\ref{pcygsect}
646: and the model parameters are shown in Table~\ref{tab_profiles}.
647: The dashed lines mark models where the optical depth at line
648: center, $\tau_c$, was fixed at unity.}
649: \end{figure*}
650:
651: The \chandra\ and \xmm\ grating spectra are the only spectra
652: that allow us to identify and study absorption- and emission
653: lines.
654: In Fig.~\ref{profiles} we show small portions of the \chandra\
655: count spectra of days 39.7 and 66.9, focused on the wavelength
656: ranges of four prominent lines, projected on a velocity scale
657: (assuming the rest wavelengths $\lambda_0$
658: given in the legends). We modeled these spectral regions with three
659: components, a continuum (rescaled blackbody), an absorption line
660: component, and an emission line component. We include the emission
661: line component in order to determine if we are seeing
662: P-Cygni profiles \citep{rsoph_iau2}.
663: We restrict the model to the narrow spectral region around each line
664: \begin{eqnarray}
665: \label{pcygmodel}
666: M(\lambda)&=&[ \{C\times B_\lambda(T)+E\times G_e(\lambda,\lambda_e,\sigma_e)\}\\ \nonumber
667: & &\ \times \{1-A\times G_a(\lambda,\lambda_a,\sigma_a)\} ]\\
668: & & \times T_{\rm ISM}(\lambda)\times T_{\rm CS}(\lambda) \nonumber
669: \end{eqnarray}
670: with $C\times B_\lambda(T)$ being the continuum (blackbody model) and
671: $A\times G_a(\lambda,\lambda_a,\sigma_a)$ and $E\times
672: G_e(\lambda,\lambda_e,\sigma_e)$ being the absorption- and emission line
673: components, respectively, both modeled as Gaussians. In an iterative process
674: we optimized the normalization $C$ of the blackbody continuum at the fixed
675: temperature $T=6.2\times 10^5\,{\rm K}$ (based on the blackbody parameters
676: found in Sect.~\ref{continuum}) and the central wavelengths
677: ($\lambda_a$ and $\lambda_e$), line widths ($\sigma_a$ and $\sigma_e$),
678: and normalizations ($A$ and $E$) of the absorption and emission line
679: components, respectively. Finally, we corrected the model for interstellar
680: and circumstellar absorption using the same absorption model described
681: in Sect.~\ref{continuum} (without the reduction of neutral oxygen)
682: with $T_{\rm ISM}(\lambda)$ and $T_{\rm CS}(\lambda)$ the transmission
683: coefficients for interstellar and circumstellar material
684: with N$_{\rm H}({\rm ISM})=2.4\times10^{21}$\,cm$^{-2}$ and
685: N$_{\rm H}({\rm CS})=2.3\times10^{21}$\,cm$^{-2}$, respectively
686: (equivalent to an effective H\,{\sc i} column density of
687: $4.7\times10^{21}$\,cm$^{-2}$).
688:
689: For a given transition with the rest wavelength $\lambda_0$, the parameters
690: $\lambda_a$ and $\lambda_e$ can be converted to velocities from line shifts
691: ($v_s$) for the absorption- and emission line components, respectively.
692: This velocity represents the expansion velocity at the radial distance
693: within the outflow where the optical depth is unity.
694: The line widths $\sigma_a$ and $\sigma_e$ are
695: representative of the velocity distribution of the observed
696: regions which depend on the velocity- and density profile within the
697: outflow, however, we have not accounted for the contributions of the
698: instrument profile and can therefore not use these numbers. We derive
699: emission line fluxes from the parameter $E$ and optical depths at line
700: center,
701: $\tau_c=\ln \{M(\lambda_{\rm min})/{\rm CONT}(\lambda_{\rm min})\}$ at
702: the wavelength where the model finds its minimum, $\lambda_{\rm min}$.
703: ${\rm CONT}(\lambda_{\rm min})=C\times B(\lambda_{\rm min},T)
704: \times T_{\rm ISM}(\lambda_{\rm min})\times T_{\rm CS}(\lambda_{\rm min})$
705: is the continuum flux at $\lambda_{\rm min}$. Note, that the emission
706: line fluxes represent the intrinsic line fluxes, including
707: self-absorption, corrected for interstellar plus circumstellar
708: absorption. Integration over the entire absorption line profile
709: allows us to estimate absorbing column densities, $xn_e$ for each line.
710: We use the oscillator strengths given in Table~\ref{tab_profiles}.
711: The emission line fluxes and column densities depend strongly on the
712: choice of N$_{\rm H}({\rm ISM})+$N$_{\rm H}({\rm CS}$), especially
713: at long wavelengths.
714:
715: We calculated uncertainty ranges for the model parameters $\lambda_{a,e}$,
716: $\sigma_{a,e}$, $A$, and $E$ separately by
717: calculating a grid of $\chi^2_{\rm red}$ over a given range of
718: the respective parameters. For each grid point all other
719: parameters were optimized, respecting the boundary condition of the
720: individual parameter of interest to be fixed at
721: the given grid value. The 1-$\sigma$ uncertainty ranges were then obtained
722: by interpolating the grid until $\chi^2_{\rm red}$ had increased by unity.
723: If the curve was found to be too flat so that $\chi^2_{\rm red}$ never reached
724: values higher than one above the minimum for a sensible range of values,
725: then lower or upper limits were calculated. Next, the uncertainties of
726: $\lambda_{a,e}$ and $\sigma_{a,e}$ were translated into the respective
727: uncertainties in velocity, and those of $A$ and $E$ into the uncertainties
728: of $\tau_c$ and emission line fluxes, respectively. For $\tau_c$ we only
729: considered variations of $A$, but we allowed all other parameters to be
730: optimized before a value of $\chi^2_{\rm red}$ was calculated for a given
731: $\tau_c$. The uncertainties of the column densities
732: were obtained by calculating a grid of values of $\chi^2_{\rm red}$ from
733: a range of pairs of $A$ and $\sigma_a$ that represent a grid of
734: values of $xn_0$ and probing the range of $\chi^2_{\rm red}+1$.
735:
736: The uncertainties of $\tau_c$ and $xn_0$ do not explore the full
737: uncertainty range as variations of
738: the central wavelength of the emission line component can dramatically
739: change those values. In particular if saturation is reached, we loose
740: all constraints. This is a shortcoming of our parameterized model that
741: cannot deal with saturation self-consistently.
742: For now, the given uncertainty ranges represent the minimal
743: uncertainties within which no changes can be considered significant.
744: We have also not included uncertainties from the amount of interstellar
745: and circumstellar absorption.
746:
747: In Fig.~\ref{profiles} we compare the count spectra with the models
748: $M(\lambda)$
749: after iteration of $C$, $E$, $A$, $\lambda_e$, $\sigma_e$, $\lambda_a$,
750: and $\sigma_a$ (Eq.~\ref{pcygmodel}). The relevant
751: spectral ranges are well fitted, and the values of $\chi^2_{\rm red}$
752: are below 1.2 for day 39.7 and below 3.0 for day 66.9.
753: In Table~\ref{tab_profiles} we list the line-shift
754: velocities, optical depths at line center, and emission line fluxes,
755: derived from the model parameters with their uncertainty ranges for
756: the absorption- and emission line components.
757:
758: The optical depths in the line centers are unity within the uncertainties
759: for all lines in both observations, which is consistent with our
760: expectation that we are observing regions in the outflow
761: where $\tau=1$. In Fig.~\ref{profiles} we add a dashed lines denoting
762: the respective models where $\tau=1$ was enforced during the
763: fit. For N\,{\sc vi} (day 66.9) we were not able to constrain any of
764: the parameters because the absorption trough is quite flat. An optical
765: depth of $\tau=1$ leads only to an acceptable fit
766: if the absorption line is assumed to be very narrow and
767: leads to an increase in $\chi^2$ in only one spectral bin (see top
768: right panel of Fig.~\ref{profiles}).
769:
770: The column densities could not be sufficiently constrained to assess
771: whether they varied significantly from day 39.7 to 66.9. Only
772: N\,{\sc vii}, and possibly N\,{\sc vi}, exhibited a decrease of column
773: density within the given errors. However, we
774: caution that uncertainties from the emission line component and the
775: continuum as well the amount of interstellar+circumstellar absorption
776: are not included in these error estimates.
777:
778: Within each observation the absorption line velocities are
779: consistent with \vaverageone\ for day 39.7 and \vaveragetwo\ for
780: day 66.9 (weighted averages). This marks a clear reduction in the
781: measured expansion velocity. The line widths, although dominated by
782: the instrumental line profile, have systematically but not significantly
783: decreased. We tested models where we fixed all line widths at the
784: instrumental resolution ($\sigma_{\rm gauss}=0.03$\,\AA\ at all
785: wavelengths) and found acceptable fits except for N\,{\sc vii}.
786:
787: The potential of this
788: approach over global approaches is that we can concentrate on single
789: lines that are least affected by blending or other uncertainties that
790: are more difficult to disentangle in complex atmosphere models. We
791: found optical depths at line centers of unity.
792: The second reliable result at this stage is the decrease of the
793: characteristic velocity consistently measured for five different
794: absorption lines.
795: Although the velocities derived from these absorption lines are very
796: similar to those expected at these epochs from the simple shock model as
797: applied to the Swift X-ray data \citep{bode06}, similar effects in
798: these lines observed in X-rays in classical novae are commonly ascribed
799: to observing material deeper into the envelope in which there is a
800: velocity gradient (Hubble flow). Further, the origin of the emission
801: line components has to be determined before further steps can be
802: taken. This requires a reduction of the uncertainties which we plan
803: to achieve by applying sensible boundary conditions.
804:
805: \begin{table*}[!ht]
806: \begin{flushleft}
807: \renewcommand{\arraystretch}{1.1}
808: \caption{\label{tab_profiles}Results from profile fitting illustrated in
809: Fig.~\ref{profiles}}
810: {\small
811: \begin{tabular}{p{.1cm}rr|rr|rr|rr|rr}
812: \hline
813: ion & \multicolumn{2}{c}{O\,{\sc vii}} & \multicolumn{2}{c}{O\,{\sc viii}}
814: & \multicolumn{2}{c}{N\,{\sc vi}} & \multicolumn{2}{c}{N\,{\sc vii}}
815: & \multicolumn{2}{c}{Fe\,{\sc xvii}}\\
816: $\lambda_0$ & \multicolumn{2}{c}{21.60\,\AA} & \multicolumn{2}{c}{18.97\,\AA}
817: & \multicolumn{2}{c}{28.78\,\AA} & \multicolumn{2}{c}{24.78\,\AA}
818: & \multicolumn{2}{c}{15.01\,\AA}\\
819: $f^{[a]}$ & \multicolumn{2}{c}{0.68} & \multicolumn{2}{c}{0.83}
820: & \multicolumn{2}{c}{0.66} & \multicolumn{2}{c}{0.83}
821: & \multicolumn{2}{c}{2.52}\\
822:
823: & day 39.7 & day 66.9 & day 39.7 & day 66.9 & day 39.7 & day 66.9 & day 39.7 & day 66.9& day 39.7 & day 66.9\\
824: \multicolumn{9}{l}{\bf Absorption Component}\\
825: $^{[b]}v_{\rm s}$ & $-1143^{+939}_{-317}$ & $-685^{+101}_{-104}$ & $-1264^{+477}_{-254}$ & $-772^{+91}_{-87}$
826: & $-1365^{+907}_{-227}$ & -- & $-1098^{+444}_{-116}$ & $-701^{+407}_{-78}$
827: & $-1556^{+661}_{-426}$ & $-834^{+439}_{-428}$\\
828:
829: %$^bv_{\rm w}$ & $579^{+702}_{-330}$ & $453^{+161}_{-102}$ & $723^{+536}_{-289}$ & $652^{+159}_{-117}$
830: % & $372^{+920}_{-213}$ & -- & $615^{+548}_{-169}$ & $401^{+367}_{-84}$
831: % & $475^{+736}_{-475}$ & $487^{+724}_{-487}$\\
832: %
833: $^{[c]}\tau_c$ & $0.79^{+0.83}_{-0.46}$ & $1.23^{+0.94}_{-0.45}$ & $1.04^{+0.46}_{-0.85}$ & $1.37^{+0.92}_{-0.44}$
834: & $0.72^{+0.97}_{-0.40}$ & $<3.0$ & $1.08^{+0.45}_{-0.40}$ & $1.01^{+0.49}_{-0.24}$
835: & $0.87^{+1.93}_{-0.87}$ & $0.79^{+1.81}_{-0.60}$\\
836:
837: $^{[d]}xn_0$ & $3.38^{+1.04}_{-0.96}$ & $3.17^{+0.61}_{-0.43}$ & $4.53^{+0.59}_{-0.69}$ & $4.53^{+0.44}_{-0.50}$
838: & $1.88^{+0.47}_{-0.54}$ & $<1.3$ & $3.25^{+0.35}_{-0.40}$ & $2.42^{+0.24}_{-0.30}$
839: & $1.23^{+0.81}_{-0.66}$ & $1.25^{+0.54}_{-0.57}$\\
840: \multicolumn{9}{l}{\bf Emission Component}\\
841: $^{[b]}v_{\rm s}$ & $633^{+316}_{-371}$ & -- & $544^{+402}_{-1334}$ & --
842: & $329^{+250}_{-1839}$ & -- & $559^{+593}_{-735}$ & $3^{+583}_{-335}$
843: & $363^{+383}_{-1969}$ & --\\
844:
845: %$^bv_{\rm w}$ & $483^{+2163}_{-263}$ & -- & $835^{+\infty}_{-835}$ & --
846: % & $442^{+561}_{-240}$ & -- & $423^{+\infty}_{-248}$ & --
847: % & $637^{+2552}_{-338}$ & --\\
848: $^{[e]}$flux & $3.78^{+3.5}_{-2.0}$ & -- & $2.20^{+3.0}_{-0.9}$ & --
849: & $31.9^{+18}_{-15}$ & -- & $3.55^{+7.52}_{-2.56}$ & $2.87^{+14}_{-2.9}$
850: & $0.13^{+0.13}_{-0.05}$ & --\\
851: %\hline
852: %$^f2v_{\rm exp}$ & 3567 & 1632 & 4213 & 2124 & 3054 & 2117 & 3357 & 1799 & 3834 & 1536\\
853: \hline
854: \end{tabular}
855: }
856:
857: {\scriptsize
858: $^{[a]}$oscillator strength \ \ \ \ \ \ $^{[b]}$ velocity from line shift (km\,s$^{-1}$)\hfill
859: $^{[c]}$optical depth at line center\\
860: $^{[d]}$column density (10$^{16}$cm$^{-2}$)\hfill
861: $^{[e]}$emission line flux ($10^{-10}$erg\,cm$^{-2}$sec$^{-1}$)\\
862: $^{[d]}$uncertainties of $xn_0$ were calculated at fixed wavelengths and
863: with fixed emission line parameters and are underestimated\\
864: }
865: \renewcommand{\arraystretch}{1}
866: \end{flushleft}
867: \end{table*}
868:
869: For the observation taken on day 39.7, we first assumed only absorption
870: lines, but an additional emission line component was necessary in
871: order to arrive at acceptable fits. Their nominal (line-shift)
872: velocities range from $\sim +300$ to $+600$\,km\,s$^{-1}$, but the
873: uncertainty ranges are large and include the possibility that the
874: emission lines could be at their rest wavelengths.
875:
876: We scanned the \chandra\ spectrum of day 39.7 for all absorption lines
877: that appear not to be blended and summarize the wavelengths and central
878: optical depths $\tau_c$ below the continuum in Table~\ref{alines}. We
879: attempted identifications and found lines from N, O, Fe, S, Si, Ca, and
880: Ar using the collisional database APED\footnote{Astrophysical Plasma
881: Emission Database v1.3 \citep{smith01}}. The identifications are based
882: on the assumption that a strong collisional excitation probability
883: implies a strong radiative excitation probability. We searched for
884: candidates at wavelengths that imply blue shifts of
885: $\sim 0.1$\,\AA, as this wavelength shift is found for
886: the well-identified lines used for Fig.~\ref{profiles}.
887:
888:
889: \begin{table}[!ht]
890: \begin{flushleft}
891: \renewcommand{\arraystretch}{1.1}
892: \caption{\label{alines}Strongest absorption lines found in ObsID 7296 (day 39.7)}
893: %\vspace{-.3cm}
894: \begin{tabular}{cccr||cccr}
895: $\lambda^{[a]}$ & $\tau_c$ & $\lambda_0^{[a]}$ & ID$^{[b]}$ & $\lambda^{[a]}$ & $\tau_c$ & $\lambda_0^{[a]}$ & ID$^{[b]}$\\
896: \hline
897: 14.93 & 0.86 & 15.01 & Fe\,{\sc xvii} & 23.51 & 1.41 & 23.62 & O\,{\sc i}\\
898: 15.77 & 0.41 & 15.87 & Fe\,{\sc xviii}? & 23.66 & 0.35 & 23.78 & N\,{\sc vi}\\
899: 15.94 & 0.92 & 16.00 & O\,{\sc viii} & 24.50 & 0.24 & 24.51 & S\,{\sc xiv}?\\
900: 16.71 & 0.43 & 16.78 & Fe\,{\sc xvii} & 24.69 & 0.93 & 24.78 & N\,{\sc vii}\\
901: 16.99 & 0.51 & 17.05 & Fe\,{\sc xvii} & 25.61 & 0.57 & 25.68 & Ar\,{\sc xvi}?\\
902: 17.69 & 0.48 & 17.77 & O\,{\sc vii}? & 25.76 & 0.25 & 25.84 & Ar\,{\sc xv}?\\
903: 18.88 & 1.00 & 18.97 & O\,{\sc viii} & 25.91 & 0.21 & 26.00 & Ar\,{\sc xv}?\\
904: 19.84 & 0.75 & 19.83 & N\,{\sc vii} & 26.07 & 0.51 & 26.12 & N\,{\sc vi} \\
905: 20.63 & 0.43 & 20.68 & S\,{\sc xiv}? & 26.82 & 1.06 & 26.99 & C\,{\sc vi}?\\
906: 20.84 & 0.44 & 20.91 & N\,{\sc vii} & 27.30 & 0.42 & 27.47 & Ar\,{\sc xiv}?\\
907: 21.52 & 0.83 & 21.60 & O\,{\sc vii} & 27.52 & 0.79 & 27.64 & Ar\,{\sc xiv}?\\
908: 22.25 & 0.56 & 22.33 & O\,{\sc v} & 27.83 & 0.49 & 27.90 & Si\,{\sc xii}?\\
909: 22.64 & 1.18 & 22.78 & O\,{\sc iv} & 28.00 & 0.55 & 28.11 & Fe\,{\sc xx}?\\
910: 22.93 & 1.06 & 22.97 & Ca\,{\sc xvi}? & 28.65 & 0.70 & 28.78 & N\,{\sc vi}\\
911: 23.16 & 0.31 & 23.11 & O\,{\sc iii} & 29.80 & 0.66 & 29.91 & Fe\,{\sc xvii}?\\
912: \multicolumn{2}{r}{\bf or} & 23.29 & N\,{\sc vi} & 30.35 & 1.59 & 30.47 & Ca\,{\sc xi}?\\
913: 23.36 & 0.25 & 23.30 & O\,{\sc ii} & 31.29 & 0.94 & 31.32 & Fe\,{\sc xviii}?\\
914: \hline
915: \end{tabular}
916:
917: $^{[a]}$in \AA\hfill $^{[b]}$Uncertain identifications are marked with '?'
918: \renewcommand{\arraystretch}{1}
919: \end{flushleft}
920: \end{table}
921:
922:
923: \subsection{Emission lines from the shock}
924: \label{elinesect}
925:
926: \begin{figure}[!ht]
927: \resizebox{\hsize}{!}{\includegraphics{f6}}
928: \caption{\label{cmp_sss_lines}Comparison of the \chandra\ spectra taken
929: on days 39.7 (dark) and 66.9 (light) focusing on the high-energy emission
930: lines. These lines are independent of the SSS spectrum and are fading.}
931: \end{figure}
932:
933: Shortwards of 14\,\AA, we found emission lines at wavelengths where
934: \chandra\ and \xmm\ measured shock-induced emission lines before
935: the SSS spectrum appeared
936: \citep[][Ness in prep.]{rsoph_iau1,2006IAUC.8695....2G}.
937: In Fig.~\ref{cmp_sss_lines} we show the count rate spectra of this
938: spectral range for the \chandra\ observations taken on days 39.7 and
939: 66.9. In contrast to the SSS spectra on these days, the latter spectrum
940: exhibits fainter emission lines than during the day-39.7 observation.
941: These lines must be collisionally excited, due to the absence of any
942: ionizing radiation of sufficient energies, and we interpret them as
943: residual emission from the shock.
944:
945: We measured wavelengths and emission line fluxes using our line
946: fitting tool Cora \citep{newi02}. The wavelengths can be converted
947: to velocities that characterize the dynamics of the plasma that
948: produces these lines. For day
949: 39.7 we found a small systematic line shift from all lines that
950: translate to a weighted average plus variance of
951: $-526\,\pm\,114$\,km\,s$^{-1}$. For the other two observations,
952: the lines were not strong enough to estimate reliable uncertainties
953: of the line positions.
954:
955: \begin{table*}[!ht]
956: \begin{flushleft}
957: \renewcommand{\arraystretch}{1.1}
958: \caption{\label{elines}Evolution of emission line fluxes.}
959: \begin{tabular}{lrccccc|ccc}
960: & & & & flux$^{[c]}$ & flux$^{[c]}$& flux$^{[c]}$ & ratio & ratio & ratio\\
961: \multicolumn{2}{l}{Ion \hfill $\lambda_0^{[a]}$\small{(\AA)}}& TC$^{[b1]}$ & TC$^{[b2]}$ & day 39.6 & day 54.0 & 66.9 & 54/39 & 66/54 &66/39\\
962: \hline
963: \multicolumn{7}{l|}{\bf Lines from shock ('hard lines')}&&\\
964: Si\,{\sc xiv} & 6.18 & 0.841 & 0.897 & $112.1\,\pm\,10.6$ & $ 47.2\,\pm\,12.8$ & $22.1\,\pm\,6.3$ & 0.42 & 0.47 & 0.20 \\
965: Si\,{\sc xiii} & 6.65 & 0.809 & 0.876 & $336.4\,\pm\,55.5$ & $158.3\,\pm\,68.2$ & $50.9\,\pm\,33.1$ & 0.47 & 0.32 & 0.15 \\
966: Mg\,{\sc xii} & 8.42 & 0.693 & 0.793 & $196.8\,\pm\,12.3$ & $106.7\,\pm\,7.9 $ & $40.5\,\pm\,7.2$ & 0.54 & 0.38 & 0.21 \\
967: Mg\,{\sc xi} & 9.20 & 0.626 & 0.744 & $386.3\,\pm\,38.1$ & $240.2\,\pm\,26.5$ & $133.3\,\pm\,27.7$ & 0.62 & 0.56 & 0.35 \\
968: Ne\,{\sc x} & 12.13 & 0.390 & 0.556 & $285.5\,\pm\,11.8$ & $105.7\,\pm\,9.3 $ & $95.2\,\pm\,8.3$ & 0.37 & 0.90 & 0.33 \\
969: Fe\,{\sc xvii} & 12.26 & 0.380 & 0.547 & $ 70.2\,\pm\,11.5$ & $ 17.4\,\pm\,8.1 $ & $26.5\,\pm\,8.6$ & 0.25 & 1.52 & 0.38 \\
970: Ne\,{\sc ix} & 13.44 & 0.290 & 0.463 & $138.1\,\pm\,13.8$ & $ 42.0\,\pm\,5.5 $ & $34.1\,\pm\,11.4$ & 0.30 & 0.81 & 0.25 \\
971: Ne\,{\sc ix}$^{[i]}$ & 13.55 & 0.282 & 0.455 & $ 66.7\,\pm\,8.0 $ & $ 28.8\,\pm\,4.1 $ & $38.1\,\pm\,8.1$ & 0.43 & 1.32 & 0.57 \\
972: Ne\,{\sc ix}$^{[i]}$ & 13.69 & 0.273 & 0.445 & $104.4\,\pm\,9.9 $ & $ 36.4\,\pm\,3.8 $ & $52.5\,\pm\,8.4$ & 0.35 & 1.45 & 0.50 \\
973: Fe\,{\sc xviii} & 14.20 & 0.239 & 0.410 & $ 48.7\,\pm\,6.6 $ & -- & $20.7\,\pm\,9.1$ & & & 0.42\\
974: \hline
975: \multicolumn{7}{l|}{\bf Lines on top of the SSS spectrum ('soft lines')}&&\\
976: Fe\,{\sc xvii} & 15.01 & 0.225 & 0.403 & $170^{+170}_{-70}$ & -- & --&&&\\
977: O\,{\sc viii} & 18.97 & 0.087 & 0.226 & $848^{+1140}_{-130}$ & -- & --&&&\\
978: O\,{\sc vii} & 21.60 & 0.031 & 0.121 & $361^{+340}_{-190}$ & -- & --&&&\\
979: O\,{\sc vii}$^{[i]}$ & 21.80 & 0.028 & 0.115 & $381.8\,\pm\,24.6$ & -- & --&&&\\
980: O\,{\sc vii}$^{[i]}$ & 22.10 & 0.025 & 0.106 & $418.2\,\pm\,24.6$ & -- & --&&&\\
981: N\,{\sc vii} & 24.78 & 0.051 & 0.181 & $400^{+850}_{-290}$ & -- & $321^{+1380}_{-320}$&&&\\
982: N\,{\sc vi} & 28.78 & 0.010 & 0.072 & $325^{+180}_{-150}$ & -- & $247^{+920}_{-250}$ &&&\\
983: N\,{\sc vi}$^{[i]}$ & 29.10 & 0.009 & 0.066 & $282.1\,\pm\,19.1$ & -- & $443.9\,\pm\,31.9$&&&\\
984: N\,{\sc vi}$^{[i]}$ & 29.54 & 0.007 & 0.059 & $788.5\,\pm\,27.3$ & -- & $397.7\,\pm\,32.6$&&&\\
985: \hline
986: \end{tabular}
987:
988: $^{[a]}$rest wavelengths\ \ \ $^{[b1], [b2]}$transmission coefficients
989: assuming $^{[b1]}$N$_{\rm H}=4.7\times10^{21}$\,cm$^{-2}$
990: and $^{[b2]}2.4\times10^{21}$\,cm$^{-2}$\\
991: $^{[c]}10^{-14}$\,erg\,cm$^{-2}$\,s$^{-1}$, not corrected for interstellar absorption\ \
992: $^{[i]}$intersystem line
993: \renewcommand{\arraystretch}{1}
994: \end{flushleft}
995: \end{table*}
996:
997: In Table~\ref{elines} we list the emission line fluxes for the strongest
998: lines. We have not corrected the measured fluxes for the effects of
999: interstellar and/or circumstellar absorption. As shown in
1000: Sect.~\ref{continuum}, our models of interstellar and circumstellar
1001: absorption are not accurate enough to derive reliable corrections. In
1002: the first four columns we list the line identifications, wavelengths,
1003: and transmission coefficients that would have to be applied under the
1004: assumptions of
1005: N$_{\rm H}({\rm ISM})+$N$_{\rm H}({\rm CS})=4.7\times10^{21}$\,cm$^{-2}$
1006: and N$_{\rm H}({\rm CS})=0$ (i.e., only interstellar absorption with
1007: N$_{\rm H}({\rm ISM})=2.4\times10^{21}$\,cm$^{-2}$) assuming
1008: solar abundances. As soon as a more accurate absorption model is
1009: available, the fluxes listed in Table~\ref{elines} can be converted
1010: to unabsorbed fluxes at the source. However,
1011: if the lines originate from the shock, they may only be affected by
1012: interstellar absorption and not by circumstellar material.
1013: In the top part of Table~\ref{elines} we list the line fluxes
1014: originating from the shock and in the bottom part we include the
1015: emission line fluxes measured as part of the models described in
1016: Sect.~\ref{pcygsect} for days 39.7, 54.0, and 66.9. We reference we
1017: label these lines 'hard lines' and 'soft lines', respectively.
1018: For the soft lines we reversed the correction for interstellar
1019: absorption that has been applied with the terms
1020: $T_{\rm ISM}(\lambda)\times T_{\rm CS}(\lambda)$ in Eq.~\ref{pcygmodel}
1021: for comparison with the unabsorbed hard lines. We also include the line
1022: fluxes for He-like intersystem lines measured in Sect.~\ref{hesect}.
1023: In the last three columns we list the changes in the line fluxes
1024: between days 54.0 and 39.9, 66.9 and 54.0, and 66.9 and 39.7,
1025: respectively.
1026:
1027: The fluxes are all of the same order of magnitude which implies that
1028: they originate from the same plasma. In that case the soft
1029: lines considered as part of the absorption lines in
1030: Sect.~\ref{pcygsect} are actually coming from the
1031: shock. This would also explain why these lines disappeared by day 66.9,
1032: which earlier had been interpreted as disappearance of P Cygni profiles
1033: \citep{rsoph_iau2}. However, correcting for interstellar plus
1034: circumstellar absorption will result in considerable enhancements
1035: of the fluxes of the soft lines because of their longer wavelengths,
1036: but that depends critically on the value of
1037: N$_{\rm H}$(ISM)$+$N$_{\rm H}$(CS) and the applied absorption model.
1038: The resulting higher fluxes in the soft lines
1039: can be explained by photoexcitation in addition
1040: to collisional excitations. If the lines originate from the shock,
1041: only interstellar absorption has to be accounted for, leading to line
1042: fluxes that suggest a mixture of collisional
1043: excitations and photoexcitations. Otherwise, circumstellar absorption
1044: has to be added, and stronger line fluxes would have to be assumed
1045: that could be explained by a smaller distance between the
1046: radiation source and the plasma emitting the soft lines.
1047:
1048: \subsection{He-like triplets}
1049: \label{hesect}
1050:
1051: The most prominent transitions in the range 5\,\AA --45\,\AA\ are H-like
1052: and He-like lines of elements with nuclear charge $Z=6$ (carbon) to $Z=14$
1053: (silicon). The $n=2-1$ transitions of He-like ions consist of three strong
1054: lines, a resonance line (r: 1s2p\,$^1$P$_1$) an intercombination line
1055: (i: 1s2p\,$^3$P$_1$), and a forbidden line (f: 1s2s\,$^3$S$_1$), all
1056: decaying into the ground state \citep[1s$^2\,^1$S$_0$, see,
1057: e.g.,][]{ness_CS}. These lines are important for two reasons. First,
1058: the measured wavelengths of the intercombination- and forbidden lines
1059: can constrain the exact location of the emission line components in
1060: the models described in Sect.~\ref{pcygsect} because they are not
1061: absorbed. Second, the flux ratio f/i of the He-like triplets are
1062: important density diagnostics \citep[e.g.,][]{gj69,denspaper}. In
1063: low-density environments with the presence of strong UV fields, the
1064: f/i ratios can also be used to measure the distance between the X-ray
1065: plasma and the UV source \citep[applied, e.g., by][to the winds of
1066: O stars immersed
1067: in the UV radiation field from the stellar surface]{waldcas01}.
1068:
1069: We used our line fitting tool Cora \citep{newi02} to determine the
1070: line positions and line fluxes of the He-like lines of O\,{\sc vii},
1071: $\lambda_0=$(21.6, 21.8, 22.1)\,\AA, and N\,{\sc vi}
1072: $\lambda_0=$(28.78, 29.1, 29.54)\,\AA,
1073: using Gaussian line profiles. We varied the line
1074: positions and found that the i and f lines are not significantly
1075: shifted from their rest wavelengths (upper limit
1076: $\sim 600$\,\,km\,s$^{-1}$). More refined models in
1077: Sect.~\ref{pcygsect} can therefore be calculated with the boundary
1078: condition that the emission line components are not shifted.
1079:
1080: Next, we measured the line fluxes with variable line strengths and
1081: fixed wavelengths and line widths on top of a constant continuum.
1082: For O\,{\sc vii} we used a constant value of 9000\,counts per \AA\
1083: and for N\,{\sc vi} where the continuum is not as flat as for
1084: O\,{\sc vii}, we used a linear continuum with a slope of -40
1085: counts per \AA. In Fig.~\ref{o7} we show the spectra and
1086: our models for day 39.7.
1087:
1088: We considered the possibility that the lines shortwards of the
1089: forbidden lines ($\sim 22$\,\AA\ and $\sim 29.4$\,\AA) are
1090: blue-shifted counterparts of the forbidden lines. In that case
1091: the intercombination lines should also have blue-shifted companions
1092: at the same velocity. For O\,{\sc vii} a model with an additional
1093: component,
1094: shifted by $-1763.5$\,km\,s$^{-1}$, results in a good representation of
1095: the data (top panel of Fig.~\ref{o7}). For N\,{\sc vi} the forbidden
1096: line clearly shows a second component at a different velocity, but there
1097: is no evidence for
1098: a blue-shifted component of the intercombination line at this
1099: velocity. No transitions from other ions are known for these
1100: wavelengths that would be strong enough, but these lines could be
1101: satellite lines that appear in a photoionized recombining plasma.
1102: For our further analysis we only use the fluxes measured at the rest
1103: wavelengths.
1104:
1105: We include the measured fluxes for the i- and f lines in
1106: Table~\ref{elines}. In a collisional plasma the sum of the fluxes
1107: in the i- and f- lines is expected to be of the same order
1108: as that in the resonance line (r), while in a plasma where
1109: photoexcitation takes place, the ratio $G=$(f+i)/r should be less
1110: than 1. If this ratio is greater than one, this indicates a purely
1111: recombining plasma or a photoionized plasma with high column densities
1112: suppressing resonant diffusion \citep{coupe04,godet04}. From the numbers
1113: in Table~\ref{elines} we compute $G=2.2\,\pm\,1.2$ for O\,{\sc vii}
1114: and $G=3.3\,\pm\,2.0$ for N\,{\sc vi}, both greater than one, but
1115: with large uncertainties.
1116:
1117: At low densities the ratio of $R=$f/i is expected to be 3.9 for
1118: O\,{\sc vii} and 4.9 for N\,{\sc vi} (low-density limit $R_0$),
1119: while in a high-density plasma, this ratio
1120: is reduced by collisional excitations from the upper level of
1121: the f-line into the upper level of the i-line, followed by
1122: radiative de-excitations from there into the ground. We measured
1123: f/i ratios of $1.1\,\pm\,0.1$ and $2.8\,\pm\,0.2$, which correspond to
1124: densities of $\log(n_e)=11$ and $\log(n_e)=10.8$ for O\,{\sc vii} and
1125: N\,{\sc vi}, respectively. We used the summarized parameterized
1126: form given by, e.g., equation 2 in \cite{denspaper} that has been
1127: derived by \cite{gj69} with the critical densities
1128: $\log(N_c)=10.5$ and $\log(N_c)=9.7$ for O\,{\sc vii} and
1129: N\,{\sc vi}, respectively. As a first approach, we neglected the
1130: radiation term $\phi/\phi_c$. For comparison, \cite{nye07} measured
1131: densities of order $10^5-10^{10}$\,cm$^{-3}$ from
1132: various infrared lines between days 64 and 84, and our values
1133: are rather high compared to that. However, the IR and the X-ray
1134: emission may originate from different places in the flow,
1135: especially if the X-ray lines originate from the shock (see
1136: previous section).
1137:
1138: We next study the possibility that UV radiation fields lead
1139: to the low f/i ratios. \cite{ness_CS} have given a method that
1140: converts measured UV fluxes at the wavelengths $\sim 1630$\,\AA\
1141: for O\,{\sc vii} and at $\sim 1900$\,\AA\ for N\,{\sc vi} into
1142: the parameter $\phi/\phi_c$ that we have neglected above
1143: \citep[see also][]{blum72,waldcas01}. This method includes a
1144: dilution factor which can be converted into the distance between
1145: the plasma and the UV radiation source, assuming a distance to
1146: RS\,Oph of 1.6\,kpc. We examined the simultaneous UVOT Grism data
1147: taken with \swift, but the data are not calibrated yet, and the
1148: wavelength range around 1630\,\AA\ is not covered.
1149: We, therefore, inspected the archive of the International
1150: Ultraviolet Explorer satellite (IUE), which contains a number of
1151: observations of RS\,Oph taken after the outburst on 1985, January 28
1152: \citep[e.g.][]{shore96}. We extracted an observation taken 1985,
1153: Feb. 26, 23:47:00UT ($\sim 30$ days after outburst; ObsID SWP25327)
1154: and estimated flux levels for the continuum of
1155: $\sim 2\times 10^{-13}$\,erg\,cm$^{-2}$\,sec$^{-1}$ at 1630\,\AA\
1156: and 1900\,\AA. Emission lines near these wavelengths from
1157: He\,{\sc ii} (1640\,\AA), Si\,{\sc iii}] (1892\,\AA), and C\,{\sc iii}]
1158: \citep[1908\,\AA;][]{shore96} can increase the relevant flux to no more than
1159: $10^{-12}$\,erg\,cm$^{-2}$\,sec$^{-1}$. The former flux level
1160: leads to a distance between the UV source and the immersed
1161: plasma of $\sim 4\times 10^{11}$\,cm (6\,R$_\odot$) for O\,{\sc vii}
1162: and $\sim 3\times 10^{12}$\,cm (40\,R$_\odot$) for N\,{\sc vi},
1163: while the latter flux leads to $\sim 9\times 10^{11}$\,cm
1164: (13\,R$_\odot$) and $\sim 10^{13}$\,cm (180\,R$_\odot$) for
1165: O\,{\sc vii} and N\,{\sc vi}, respectively.
1166: These numbers are quite uncertain because we assume that the UV
1167: emission level was similar during the 1985 and 2006 outbursts.
1168: However, we can conclude that if the plasma emitting
1169: these lines is further away from the UV radiation source,
1170: than 200\,R$_\odot$, then the density must be higher than
1171: 10$^{10}$\,cm$^{-3}$. This distance is small compared to the
1172: extent of the expanding shell on day 39.7. With an expansion
1173: velocity of 1100\,km\,s$^{-1}$ the shell has reached a radius of
1174: $3.8\times10^{14}$\,cm (5400\,R$_\odot$).
1175:
1176: \begin{figure}[!ht]
1177: \resizebox{\hsize}{!}{\includegraphics{f7a}}
1178:
1179: \resizebox{\hsize}{!}{\includegraphics{f7b}}
1180: \caption{\label{o7}Spectral region around the He-like triplets of
1181: O\,{\sc vii} (top) and N\,{\sc vi} (bottom) on day 39.7. The respective
1182: resonance lines (r) at 21.6 and 28.78\,\AA\ are seen in absorption and
1183: emission while, the intercombination- (i) and forbidden (f) lines
1184: are only seen in emission, giving us the likely wavelengths of
1185: all emission lines. Gaussian fits to the lines
1186: reveal no line shifts (upper limit $\sim 600$\,km\,s$^{-1}$).
1187: Additional lines are seen shortwards of the forbidden lines,
1188: that could come from a velocity component with $-1763$\,km\,s$^{-1}$
1189: and $-1237$\,km\,s$^{-1}$ for O\,{\sc vii} and N\,{\sc vi},
1190: respectively, but no counterpart in the N\,{\sc vi} intercombination
1191: line is detected.}
1192: \end{figure}
1193:
1194: \subsection{Spectral Variability from High- to Low-flux phases}
1195: \label{variability}
1196:
1197: \begin{figure}[!ht]
1198: \resizebox{\hsize}{!}{\includegraphics{f8}}
1199: \caption{\label{hr}{\bf Top}: \chandra\ light curve extracted from the
1200: dispersed photons in the wavelength range 15--30\,\AA\ (day 39.7).
1201: {\bf Bottom}: evolution of hardness ratio (in steps of 200\,sec)
1202: defined by the energy ranges given in the legend. The grey curve
1203: is the light curve from the top panel, shifted by 1000\,sec on the
1204: time axis and rescaled with a factor 0.012.}
1205: \end{figure}
1206:
1207: We now focus on spectral changes on shorter time
1208: scales. During the observation taken on day 39.7, variations
1209: in brightness were detected which were not seen in the other two
1210: observations. In Fig.~\ref{hr} we show the evolution of the observed
1211: flux integrated over the wavelength range 15--30\,\AA\ (top panel) in
1212: comparison to the hardness ratio HR (bottom panel) with the same
1213: definitions as used in Sect.~\ref{spec} (repeated in the legend).
1214: The rescaled light curve, marked
1215: with the grey curve in the bottom panel of Fig.~\ref{hr} indicates that
1216: the hardness ratio has the same shape but is retarded by 1000\,sec.
1217:
1218: In order to study the spectral changes in more detail, we extracted
1219: spectra from the two different time intervals marked in dark and light
1220: shadings in Fig.~\ref{lc}, representing high-flux
1221: and low-flux emission. In Fig.~\ref{tres} we compare these spectra
1222: with the dark shading being the high-flux spectrum
1223: and the light shading the low-flux spectrum. In the bottom two panels we
1224: show the cumulative distribution of counts and the ratio spectrum,
1225: respectively.
1226:
1227: The cumulative distribution traces the spectral shape regardless of the
1228: different intensities and can be used for a Kolmogorov-Smirnov 2-sample
1229: test. At $\sim 27$\,\AA\ we find the largest difference of 0.03, which
1230: has to be compared to the maximally allowed difference of 0.007 if the
1231: two spectra were to be considered identical within 95\,percent
1232: probability. The spectra are therefore different in their shapes in
1233: addition the obvious difference of intensity. Inspection of the
1234: cumulative curve reveals that the high-flux spectrum is slightly softer.
1235:
1236: The ratio spectrum shows up to a factor of two higher emission in the
1237: high-flux spectrum. The absorption lines of O\,{\sc viii} at 19\,\AA\
1238: and N\,{\sc vii} at 25\,\AA\ as well as the O\,{\sc i} absorption edge
1239: at 22.83\,\AA\ are deeper relative to the continuum in the high-flux
1240: spectrum. This indicates that the brightness changes involve more than
1241: changes in the brightness of the continuum. In contrast,
1242: shortwards of 15\,\AA\ as well as near 29.5\,\AA\ (the wavelength of
1243: the N\,{\sc vi} forbidden line) the two spectra are identical.
1244:
1245: \begin{figure}[!ht]
1246: \resizebox{\hsize}{!}{\includegraphics{f9}}
1247: \caption{\label{tres}Comparison of spectra extracted from high-flux
1248: and low-flux intervals in ObsID 7296. {\bf Top}: direct comparison of
1249: count rate spectra. {\bf Middle}: Cumulative count distribution.
1250: {\bf Bottom}: Ratio of spectra.}
1251: \end{figure}
1252:
1253: In Fig.~\ref{lshape1_7296_lh} we compare the high-flux and low-flux spectra
1254: around the absorption lines in velocity space. While the high-flux phase
1255: provides additional continuum emission, the emission level
1256: in the troughs of the absorption lines is unchanged. There is no detectable
1257: difference in blue shift of the absorption troughs.
1258:
1259: \begin{figure}[!ht]
1260: \resizebox{\hsize}{!}{\includegraphics{f10}}
1261: \caption{\label{lshape1_7296_lh}Details of \chandra\ spectra of day
1262: 39.7 for high-flux (solid) and low-flux (dashed) phases, plotted in
1263: velocity space for the lines indicated in the legends of each panel.
1264: The absorption lines are blue shifted by the same amount in both spectra.}
1265: \end{figure}
1266:
1267: We repeated our model from Sect.~\ref{pcygsect} for these two spectra. While
1268: the optical depths at line center were around unity for both spectra, the
1269: line column densities were significantly lower in the low-flux observation.
1270: For example, for O\,{\sc viii} we found a column density of
1271: $6.33^{+1.58}_{-1.51}$ in the high-flux spectrum but only
1272: $2.93^{+1.92}_{-1.16}$ in the low-flux spectrum. The line shift velocities
1273: were the same in the models to both spectra.
1274:
1275: \section{Discussion}
1276: \label{disc}
1277:
1278: In Sect.~\ref{anal} we have presented a number of issues that can
1279: be addressed with the three grating spectra of RS\,Oph. The shape of the
1280: spectrum cannot be approximated by a blackbody model. The dominant
1281: feature, the O\,{\sc i} absorption edge, changes with time. The
1282: quantitative analysis of these changes is complicated by various
1283: overlapping absorption processes. For example, the
1284: ionization edges and 1s-2p transitions from oxygen in low ionization
1285: stages introduce significant structure to the observed spectrum.
1286: These low ionization stages can occur in both the circumstellar and
1287: the interstellar material in the line of sight. We thus need a more
1288: refined model accounting for more processes than only K-shell
1289: ionization of neutral elements in order to understand the effects
1290: that the interstellar plus circumstellar material have on the observed
1291: spectra.
1292:
1293: These uncertainties affect the interpretation of the origin and
1294: production mechanism of the soft emission lines listed in the bottom
1295: part of Table~\ref{elines}. While the hard lines at short wavelengths
1296: can only be collisionally excited and originate from the shock, the
1297: soft lines can in addition be photoexcited because of the presence
1298: of the bright SSS spectrum. The contribution from photoexcitations
1299: depends on the distance between the hot WD atmosphere that provides
1300: the radiation and the plasma that emits the soft lines. If they
1301: originate from the shock, this distance will be greater than if
1302: they are part of the hot WD atmosphere. In the latter case
1303: the soft lines should be much stronger than in the former case.
1304: However, the amount of assumed extinction will be enhanced by
1305: circumstellar material in addition to the interstellar medium,
1306: leading to lower line fluxes measured at Earth.
1307: In order to solve the ambiguity from these competing processes,
1308: an improved absorption model for interstellar and circumstellar
1309: absorption is required to derive accurate unabsorbed lines fluxes
1310: for each case. The hard lines can be used to constrain the mean
1311: emission measure distribution for a collisional plasma as developed by
1312: \cite{ness_vel} for the Classical Nova V382\,Vel. This will be part
1313: of an upcoming paper (Ness et al. in prep). If the soft lines originate
1314: from the shock, they will show up with an excess in the emission
1315: measure distribution, reflecting the contributions from
1316: photoexcitations. The amount of excess depends on the distance between
1317: the plasma and the hot WD atmosphere, which has to be of the same
1318: order as that between atmosphere and the shocked plasma.
1319: We note that the disappearance of the soft emission lines between
1320: days 39.7 and 66.9 is consistent with the fading of the lines produced
1321: in the shock. The only soft lines measurable on day 66.9 were also
1322: weaker than on day 39.7, even though the SSS spectrum was brighter.
1323: We thus favor a scenario in which the soft lines originate from the
1324: shock and are partially photoexcited, while being absorbed only
1325: by interstellar material.
1326:
1327: The optical depths at line center and the
1328: velocities derived from the absorption lines (Sect.~\ref{pcygsect})
1329: are not affected by the uncertainties in interstellar and/or
1330: circumstellar absorption. The reduction of the velocities is due
1331: to observing deeper into the flow at the time of
1332: the later observation.
1333:
1334: Our absorption line model can be improved beyond the results
1335: presented in Table~\ref{tab_profiles}. First, we need to
1336: include the instrumental line profile in order to explore the
1337: line widths. From our conclusion in Sect.~\ref{hesect} we can
1338: define the boundary condition that the emission line components
1339: are at their rest wavelengths, which leads to a reduction of
1340: the uncertainties of the other parameters. Our above conclusion
1341: that the soft emission lines likely originate from the shock has
1342: to be accounted for by not allowing self absorption. We repeated
1343: our models without self absorption and found similar results,
1344: such that our results are robust.
1345:
1346: The issue of the appearance of P Cygni type profiles on day
1347: 39.7 cannot fully be answered at this stage. The shape of the
1348: absorption lines is not typical for P Cygni profiles, which usually
1349: show a sharp blue edge which represents the terminal velocity
1350: of the wind. Our finding of symmetric absorption lines indicates
1351: that we are not observing the entire velocity profile but
1352: only a fraction of it. Those plasma regions that reside at the
1353: terminal velocity (in velocity space) could be optically thin and
1354: thus contribute little to the absorption lines. In the same
1355: way, this plasma may also be too thin to produce measurable emission
1356: lines, that would be scattered into our direction from the
1357: photoexcited plasma moving perpendicularly to the line of sight,
1358: further supporting our above conclusion that the soft lines
1359: originate from the shock. Also, the He-like resonance lines are
1360: expected to be significantly stronger than the intersystem lines
1361: if photoexcitation of the circumstellar material was dominant.
1362:
1363: The high degree of variability during the first week of the SSS phase
1364: (including our observation on day 39.7) is a new discovery that has
1365: not been observed in any other nova. However, we have never had the
1366: same frequency of observations of the same event. In one case, V4743\,Sgr
1367: \citep{v4743}, a decline of X-ray brightness occurred, which
1368: may be a similar phenomenon. Although this decay is only seen during
1369: the first of four observations
1370: taken during the SSS phase, it is not known when the SSS phase
1371: started, and thus how early in the SSS evolution the decline was
1372: observed. Variability was also observed in V1494\,Aql, V4743\,Sgr,
1373: and V832\,Vel, but the amplitude was much lower and the variations
1374: were periodic \citep[e.g.,][]{v4743,drake03}.
1375:
1376: In our \chandra\ observation on day 39.7 we discovered the hardness
1377: ratio light curve lagging behind the total light curve by 1000\,s.
1378: The hardness ratio is commonly used as a temperature indicator.
1379: If the temperature varies, then this leads to an
1380: increase in brightness before the spectral hardness increases.
1381: Temperature variations could occur periodically, but we did not
1382: detect any periods in the observation on day 39.7. With
1383: the existing observations, we cannot exclude the presence of any
1384: periods longer than three hours or shorter than a few days. However,
1385: photoionization of oxygen could also lead to the higher hardness
1386: ratios, as ionized oxygen allows more hard emission to pass through.
1387: In this case the material in the line of sight responds
1388: to the brightness of the continuum emission by adjusting the
1389: degree of ionization within 1000 s. This timescale would
1390: be the same for ionization and recombination and implies
1391: a density of the absorbing material of $\sim 10^{10}-10{11}$\,cm$^{-3}$.
1392: This is consistent with the densities derived from the He-like
1393: triplets, and the changes in hardness can therefore be explained
1394: by variations in the degree of ionization of the circumstellar
1395: material. However, this does not explain the origin of the
1396: brightness changes.
1397:
1398: The spectra extracted from two different time intervals during the
1399: episode of high-amplitude variations show complex changes. We found
1400: deeper absorption lines and a deeper oxygen edge during times of
1401: brighter continuum emission. This shows that the brightness
1402: variations are correlated to the absorption behavior of the
1403: expanding shell and the surrounding material and each spectrum
1404: must be treated independently.
1405:
1406: As an example, we applied our model from Sect.~\ref{pcygsect}
1407: to the two spectra presented in Sect.~\ref{variability} and found
1408: higher column densities during the high-flux phase while the
1409: velocities were the same for both spectra.
1410: Higher column densities arise when we either look deeper into the
1411: outflow (longer path length) or when the density is higher (or
1412: both). The high-flux phases can have their origin in a
1413: reduction of the opacity, allowing us to view deeper into the
1414: outflow where more emission is produced, the densities are higher,
1415: and the temperatures are higher.
1416:
1417: Most of our results are based on preliminary models, and we are
1418: improving these models. The first steps will
1419: have to improve the atmosphere and the model for interstellar and
1420: circumstellar absorption. The stellar atmosphere has to be modeled as
1421: an expanding atmosphere with PHOENIX
1422: \citep[e.g.][]{petz05}. The O edge has not been explored much, and
1423: new models have to be developed \citep[see, e.g.,][]{paerels01_abs}.
1424: We will further improve our model that fits the absorption and emission
1425: lines simultaneously (Sch\"onrich et al. in prep). With improved models
1426: in place, the changes of temperatures, luminosities and absorption
1427: behavior during the phase of extreme variability can be determined.
1428: A separate analysis of the emission lines and continuum produced in
1429: the shocked plasma is under way (Ness et al. in prep), and more
1430: detailed hydrodynamic models are being developed for comparison with
1431: the data \citep[see e.g.][]{vaytet07}. Finally, the
1432: \swift\ spectra can be used to constrain the evolution between the
1433: grating observations. The model parameters obtained from grating
1434: spectra on days 37.9, 54.0, and 66.9 can be interpolated for days
1435: on which \swift\ spectra are available. The spectral models obtained
1436: from the interpolated parameters have to agree with the respective
1437: measured \swift\ spectra.
1438:
1439: \section{Summary and Conclusions}
1440: \label{concl}
1441:
1442: The X-ray grating spectra of the SSS phase are complex and contain
1443: a great deal of information. The objective of this paper is to provide
1444: an overview of the physical information that can be obtained from
1445: X-ray data. We have found:
1446:
1447: \begin{itemize}
1448: \item The brightness changes between the three observations are
1449: consistent with the variations measured from the \swift\
1450: observations at the same epochs.
1451: \item Short-period oscillations ($\sim$ 35\,sec), first detected in some
1452: of the \swift\ observations, are confirmed. We discovered that this
1453: period resides in the plasma that emits the SSS spectrum.
1454: \item The episode of high-amplitude variations detected with \swift\
1455: that occurs on time scales of days is accompanied by variations on
1456: time scales of hours.
1457: \item The depth of the O\,{\sc i} absorption edge is variable, yielding
1458: the lowest optical depth on day 54.0. One possible explanation is
1459: photoionization of the absorbing material in the line of sight.
1460: \item The absorption lines are shifted by \vaverageone\ on day 39.7 and
1461: \vaveragetwo\ on day 66.9.
1462: \item Declining emission from the shock is observed in the form
1463: of collisionally excited emission lines shortwards of the SSS continuum.
1464: These lines are slightly blue-shifted.
1465: \item Superimposed on the SSS spectrum are emission lines
1466: whose origin could either be from the shocked plasma or from photoexcited
1467: material in the outer regions of the outflow.
1468: \item Intersystem lines from He-like ions, are found
1469: at their rest wavelengths. The ratio (i+f)/r (r the
1470: 1s-2p resonance line) is greater than one, indicating contributions from
1471: recombination into excited states and small contributions from
1472: resonance scattering. The ratio f/i indicates either densities in
1473: excess of $10^{11}$\,cm$^{-3}$ or a short distance between the
1474: plasma containing the He-like ions and a source for UV radiation.
1475: \item The brightness changes during the phase of variability induce
1476: changes in spectral hardness, retarded by 1000\,sec. Absorption
1477: lines and the O edge are deeper during high-state phases, and
1478: if the density of the absorbing material is of order
1479: $10^{10}-10^{11}$\,cm$^{-3}$, then the hardness changes can
1480: exclusively be explained by ionization. In that case the increases
1481: and reduction of brightness occur underneath the absorbing
1482: material, possibly in the regions where nuclear burning takes
1483: place.
1484:
1485: \end{itemize}
1486:
1487: \acknowledgments
1488:
1489: JUN gratefully acknowledges support provided by NASA through \chandra\ Postdoctoral
1490: Fellowship grant PF5-60039 awarded by the \chandra\ X-ray Center, which is operated by
1491: the Smithsonian Astrophysical Observatory for NASA under contract NAS8-03060.
1492: SS received partial support from NSF and NASA grants to ASU.
1493: PHH was supported in part by DFG grant HA 3457/2 and by the P\^ole
1494: Scientifique de Mod\'elisation Num\'erique at ENS-Lyon.
1495: KLP, JPO, APB, and MRG acknowledge support by the Particle Physics and
1496: Astronomy Research Council. MFB is grateful to the UK
1497: PPARC for the provision of a Senior Fellowship.
1498: RS acknowledges support from Stiftung Maximilianeum,
1499: Studienstiftung des deutschen Volkes, and the Max-Weber-Programm.
1500: We are grateful to Harvey Tananbaum and the Chandra
1501: Observatory for a generous allotment of Directors Discretionary.
1502: Some of the observations have been obtained with XMM-Newton, an ESA science
1503: mission with instruments and contributions directly funded by ESA Member States and NASA.
1504: Some preliminary calculations carried out for this project were performed at the
1505: H\"ochstleistungs Rechenzentrum Nord (HLRN), and at the National Energy
1506: Research Supercomputer Center (NERSC), supported by the U.S. DOE, and
1507: at the computer clusters of the Hamburger Sternwarte, supported by the DFG
1508: and the State of Hamburg. We thank all these institutions for a generous
1509: allocation of computer time. We thank an anonymous referee for
1510: instructive comments.
1511:
1512: \bibliographystyle{apj}
1513: \bibliography{rsoph,cn,astron,jn}
1514:
1515: \end{document}
1516:
1517: