222

Acta Polytechnica CTU Proceedings 2(1): 222–225, 2015

222

doi: 10.14311/APP.2015.02.0222

Early Super Soft Source Spectra in RS Oph

J.-U. Ness1

1XMM-Newton Science Operations Centre, ESA, PO Box 78, 28691 Villanueva de la Cañada, Madrid, Spain

Corresponding author: juness@sciops.esa.int

Abstract

Recent Swift X-ray monitoring campaigns of novae have revealed extreme levels of variability during the early super-soft-

source (SSS) phase. The first time this was observed was during the 2006 outburst of the recurrent nova RS Oph which

was also extensively covered by grating observations with XMM-Newton and Chandra. I focus here on an XMM-Newton

observation taken on day 26.1, just before Swift confirmed the start of the SSS phase, and a Chandra observation taken on

day 39.7. The first observation probes the evolution of the shock emission produced by the collision of the nova ejecta with

the stellar wind of the companion. The second observation contains bright SSS emission longwards of 15 Å while at short

wavelengths, the shock component can be seen to have hardly changed. On top of the SSS continuum, additional emission

lines are clearly seen, and I show that they are much stronger than those seen on day 26.1, indicating line pumping caused

by the SSS emission. The lightcurve on day 39.7 is highly variable on short time scales while the long-term Swift light

curve was still variable. In 2007, we have shown that brightness variations are followed by hardness variations, lagging

behind 1000 seconds. I show now that the hardness variations are owed to variations in the depth of the neutral hydrogen

column density of order 25%, particularly affecting the oxygen K-shell ionization edge at 0.5 keV.

Keywords: cataclysmic variables - recurrent novae - spectroscopy - X-rays - individual: RS Oph.

1 Introduction

The 2006 outburst of the recurrent symbiotic nova
RS Oph has attracted a large number of observers to
study the outburst in unprecedented detail and many
wavelength bands. In particular the coverage in X-rays
has considerably improved since the previous outburst
in 1985 by initiating the first high-density X-ray mon-
itoring of a nova with the X-ray Telescope (XRT) on
board Swift, starting on day 3.38 after initial explo-
sion (Bode et al. 2006a). Early shock emission orig-
inated from kinetic energy from the nova ejecta that
were dissipated in the slow, dense stellar wind of the
giant companion. The expected X-ray spectrum is that
of a collisional plasma which was confirmed by spectral
models to the low-resolution EXOSAT spectra taken
in 1985 (figure 3 in (O'Brien et al. 1992)) and the
Swift/XRT spectra (Bode et al. 2006b). The X-ray
grating spectrometers on board XMM-Newton (Reflec-
tion Grating Spectrometer RGS) and Chandra (Low-
and High Energy Transmission Grating Spectrometers
LETGS and HETGS) allow spectral lines to be resolved,
and in simultaneous RGS/HETGS spectra taken on day
13.8 after the initial explosion (2006 February 12.83),
bremsstrahlung continuum, H-like and He-like emission
lines, and numerous Fe lines could be identified. A de-
tailed analysis exploring the information from the emis-
sion lines yielded the distribution of electron tempera-

tures and abundances (Ness et al. 2009).

Further grating observations were taken, e.g., on
days 26.1, 39.7, 54.0, 66.9, and 111.7 that were guided
by continued dense Swift monitoring (see table 1 in Ness
et al. 2009 and top of figure 2 in Osborne et al. 2011).
During the XMM-Newton observation on day 26.1, a
jump in count rate was discovered by (Nelson et al.
2008) that was exclusively attributed to additional soft
emission, possibly indicating the start of the SSS phase.
The bremsstrahlung continuum has faded and become
softer, and also the ratios of H-like to He-like emis-
sion line strengths have shifted in favor of the He-like
lines, indicating a cooling trend in the shocked plasma
(Ness et al. 2009). The Swift observations of the ∼ 60
day SSS phase are described in (Osborne et al. 2011).
Between days ∼ 30 and 45, the X-ray count rate was
highly variable between ∼ 10 counts per second (cps)
and 200 cps, and then stabilized at ∼ 300 cps (figure 2 in
Osborne et al. 2011). This was a new discovery and was
later also observed in other novae and might be a gen-
eral phenomenon. During this high-amplitude phase,
a Chandra observation was taken on day 39.7, without
yet knowing about the risks of observing during times of
fainter emission. While this was the case, the soft emis-
sion was still bright enough for a well-exposed grating
spectrum that was first presented by (Ness et al. 2007).
The blackbody-like continuum contained deep absorp-

222

http://dx.doi.org/10.14311/APP.2015.02.0222


Early Super Soft Source Spectra in RS Oph

tion lines from highly ionized species such as O viii and
N vii that were blue shifted by ∼ 1200 km s−1. The line
profiles contained clear signs of emission lines in the
red wings (figure 5 of Ness et al. 2007) which could
either come from residual shock emission or are part of
P Cyg profiles. During this observation, the nova was
also highly variably on shorter time scales, and (Ness et
al. 2007) reported that the hardness variations followed
the same up- and down trends but lagged 1000 seconds
behind (see Fig. 1).

Figure 1: Reproduction of figure 8 from (Ness et al.
2007). The Chandra light curve on day 39.7, taken
during the early high-amplitude variability phase, was
also highly variable on shorter time scales. The hard-
ness ratio, shown in the bottom, evolved with the same
variability patterns but lagged 1000 seconds behind the
brightness variations. Reproduced by permission of the
AAS.

I focus here on two new aspects:
1) The emission line components on top of the SSS con-
tinuum (shown in figure 5 of Ness et al. 2007), com-
pared to the same lines before the SSS phase started,
are much stronger than expected from a cooling plasma
(Sect. 2.1).
2) Closer inspection of the brightness/hardness changes
during the day 39.7 observation show that the hardness
changes are due to changes in the depth of the O i ab-
sorption edge. Modeling shows that the overall column
density increases with decreasing brightness (Sect. 2.2).

2 Observations and Analysis

A full log of all XMM-Newton and Chandra grating
observation that were taken of the 2006 outburst of
RS Oph is given in table 1 in (Ness et al. 2009) out
of which I focus on the ones taken with XMM-Newton
between 2006 March 10, 23:04 and March 11, 02:21
(ObsID 0410180201) and with Chandra 2006 March 24,
12:25 - 15:38 (ObsID 7296). The calibrated spectra are
shown in direct comparison in Fig. 2 using the same
flux units without rescaling.

Figure 2: Comparison of a pre-SSS spectrum of
RS Oph taken on day 26.1 and an SSS spectrum taken
on day 39.7. In the top panel the entire wavelength
range is shown in logarithmic units, illustrating that at
short wavelengths, below ∼ 15 Å, the spectra are sim-
ilar, only showing a decline the Bremsstrahlung con-
tinuum component from day 26.1 to day 39.5. In the
panels below, narrow wavelength regions are shown
where the pre-SSS spectrum is added to the median
of the day 39.7 spectrum in the same flux units. While
short-wavelength lines have hardly changed, the excess
emission lines on top of the SSS continuum are much
stronger than before the start of the SSS phase.

2.1 Contributions of shock emission to
SSS continuum spectrum

In the top panel of Fig 2, the full spectra are shown
in logarithmic units to overcome the great contrast in
brightness between the two observations. At wave-
lengths <∼ 15 Å, the two spectra are similar in nature,
mainly differing in the brightness of the continuum.
The weaker continuum on day 39.7 can be explained by
the continuation of the fading trend of the shocks that
has already been established from the observations be-
tween days 13.8 and 26.1. The strengths of the emission
lines have hardly changed. At wavelengths >∼ 15 Å, the
emission lines seen on day 26.1 were outshone by the
bright continuum on day 39.7, and it is not intuitively
clear how they have evolved. In the panels below, four
lines are shown in more detail in linear units. The open
histograms in the panels below represent the day 39.7
spectrum while the colored shades are the day 26.1 spec-
trum, added to the median flux of the day 39.7 spec-
trum, taken over the narrow wavelength range selected

223



J.-U. Ness

for each panel. The Mg xii/xi lines have faded from day
26.1 and the Ne x line at 12.1 Å is about equally strong
at both times while these lines have become somewhat
narrower. The panels further below show that the lines
on top of the SSS continuum have become stronger from
day 26.1 to 39.7, depending on the strength of the con-
tinuum. The Fe xvii line at 15 Å, in the Wien tail of
the SSS continuum, is only slightly stronger while the
O viii and N vii lines, near the peak of the SSS con-
tinuum, are much stronger, defying the general trend
of fading emission from the shocks. A similar result
has been found by (Schönrich & Ness 2008) who show
in their figure 1 the evolution of the volume emission
measure for various emission lines as a function of their
peak formation temperature assuming collisional equi-
librium. The line fluxes of O vii and O viii from the
observation taken on day 39.7 yield clearly discrepant
values, orders of magnitude brighter than in observa-
tions without SSS emission, while those emission lines
that arise at shorter wavelengths are consistent with the
other observations.

2.2 Spectral changes with variability
during early SSS phase

The high-amplitude variations during the early SSS
phase are still not understood, leaving all options open
such as changes in absorption, in intrinsic brightness,
temperature, or in the rate of mass loss. A first ap-
proach to narrow down the options, I present the spec-
tra and how they changed with variability. During the
early variability phase of RS Oph, a short Chandra ob-
servation taken (day 39.7), and the spectral evolution
is illustrated in Fig. 3. The light curve, already shown
in Fig. 1, is shown in the right, turned around by 90o

clockwise to follow the vertical time axis in downward
direction. The blue dotted line is the hardness light
curve from the bottom panel of Fig. 1 that was rescaled
to fit in the same graph. Along the same vertical time
axis, a series of adjacent LETGS spectra are shown in
the central panel with wavelength along the horizontal
axis and flux encoded in a color scheme, where increas-
ing flux is represented by colors light green, yellow, or-
ange, red, dark blue to light blue. In the top panel,
two spectra are shown that have been integrated over
the time intervals marked with horizontal dashed lines
in the central panel shaded areas in the right panel.
The colors of the dashed lines and shades correspond
to the plot style of the spectra in the top, thus light
blue and orange shadings in the right correspond to the
light blue shade and orange thick line in the top, re-
spectively. The two spectra have been normalized to
coincide in the wavelength range 15 − 22 Å.

Ness et al. (2007) had extracted spectra from time
intervals of bright and faint episodes, roughly corre-

sponding to the time intervals 0.3 − 1.0 hours of elapsed
time (bright) and 1.1 − 2 hours, respectively, and found
differences in the spectral shape. I have now chosen
time intervals that are shifted by 1000 sec in order to
probe the hardness light curve rather than the inten-
sity light curve. The two normalized spectra in the
top clearly show that the variability in hardness is
purely owed to changes in the depth of the O i edge
at 22.8 Å, yielding the steeper edge 1000 sec after a low-
flux episode. Ness et al. (2007) had found changes in O i
on longer time scales between the grating observations
taken on days 39.7, 54.1, and 66.9 and explained these
by changes in the degree of ionization of oxygen. Only
neutral oxygen produces the deep edge at 22.8 Å while
ionized oxygen is more transparent to X-rays, leading to
a flatter edge. The intense soft continuum X-ray source
can ionize the neutral elements in the surroundings and
thus make them more transparent. On the other hand,
if the continuum source fades, the surrounding material
recombines, leading to an increase in the O i edge.

Figure 3: Dynamic spectrum for the Chandra obser-
vation of RS Oph taken on day 39.7 after the 2006 out-
burst. The zero-th order light curve is shown to the
right with three colored shades marking time intervals
from which the spectra in the top have been extracted.
Line labels that belong to strong transitions are in-
cluded. The central image is a brightness spectral/time
map that uses a color code from light green to light blue
representing increasing flux values.

The changes in the depth of the O i absorption edge
could be caused by changes in the degree of ionization
of circumstellar oxygen. Ness et al. (2007) argued that
the observed time lag of 1000 seconds is consistent with
the time scale for ionization/recombination in a plasma
with density ∼ 1011 cm−3. Since all circumstellar mate-
rial would experience the same changes in their degree
of ionization, this would lead to an overall change in NH.
Another possibility is a non-uniform radial distribution
of the oxygen abundance within the ejecta that could be
caused by beta decay of oxygen isotopes produced dur-
ing CNO burning. If during times of brigher and fainter
emission, different regions of the photosphere are visi-

224



Early Super Soft Source Spectra in RS Oph

ble, the difference in the oxygen abundance would only
affect the depth of the O i absorption edge.

In Fig. 4, I test these two scenarios in the top and
bottom panels, respectively. Shown is the normalized
faint/soft spectrum in orange shadings (orange line in
the top panel of Fig. 3) while the black curve repre-
sents a modification of the brighter/harder spectrum,
trying to reproduce the faint/soft spectrum. In the
top panel of Fig. 4, the black curve was computed by
dividing the bright/hard spectrum itself by the trans-
mission coefficients calculated from the warm absorp-
tion model developed by [?], and then normalized. The
faint/soft spectrum is qualitatively well reproduced as-
suming a value of NH = 6 × 1020 cm−2, indicating that
the change in hardness can be explained by an increase
of NH by this amount. The full amount of the hydrogen
column was found by [?] as NH = NH,ISM + NH,CS =
4.7 × 1021 cm−2, consisting of the well-known interstel-
lar component NH,ISM = 2.4 × 1021 cm−2 and an addi-
tional circumstellar component NH,CS. An increase by
6 × 1020 cm−2 thus corresponds to ∼ 26% of NH,CS.

Figure 4: Testing two hypotheses to explain the spec-
tral changes between hard/bright (blue shades in top
panel of Fig. 3) and soft/faint (orange thick line in
Fig. 3) spectra by the black thin lines. The normalized
soft/faint spectrum is shown as orange shadings while
the black line corresponds to two types of modifications
of the bright/faint spectrum, also normalized (see text).
Top: Overall increase of NH by 26% of NH,CS. Bot-
tom: Increase only of O i column by changes in O abun-
dance of 15%.

In the bottom panel of Fig. 4, the black curve was
computed by first dividing the bright-hard spectrum
by the transmission coefficients of an absorber with
NH = 4.7 × 1021 cm−2, assuming cosmic abundances
and then multiplying by the coefficients resulting from
the same absorption model assuming a modified oxygen
abundance. The faint/soft spectrum is well reproduced
up to ∼ 27 Å, but the black curve drops well below the

faint/soft spectrum at longer wavelengths. The bet-
ter agreement between the black line and the faint/soft
spectrum in the upper panel demonstrates that an over-
all increase of NH has occurred rather than a change in
the oxygen abundance alone.

3 Summary and Conclusions

A comparison of grating X-ray spectra before and af-
ter the start of the SSS phase shows that the emission
lines that were seen on top of the SSS continuum are not
simply a continuation of the early shock emission. Only
emission lines that arise at wavelengths were strong con-
tinuum emission is present are amplified compared to
the pre-SSS level, strongly suggesting photoexcitation
effects. The line profiles in the early SSS spectrum of
RS Oph can thus be understood as P Cyg profiles.

The early high-amplitude variability in RS Oph and
other novae still awaits an explanation. The Chan-
dra observation during a minimum of these variations
revealed that variability also occurs on shorter time
scales. The comparisons in this article show that the
hardness changes following brightness changes are con-
sistent with variations in NH of ∼ 26% which can be ex-
plained by variations in the overall degree of ionization
caused by the changes in intensity and thus variations
in the effectiveness of photoionization.

References

[1] Bode, M. F., OBrien, T. J., Davis, R. J., et al.
2006a, IAUCirc, 8675, 2

[2] Bode, M. F., OBrien, T. J., Osborne, J. P., et al.
2006b, ApJ, 652, 629 doi:10.1086/507980

[3] Nelson, T., Orio, M., Cassinelli, J. P., et al. 2008,
ApJ, 673, 1067 doi:10.1086/524054

[4] Ness, J.-U., Starrfield, S., Beardmore, A., et al.
2007, ApJ, 665, 1334 doi:10.1086/519676

[5] Ness, J.-U., Drake, J. J., Starrfield, S., et al. 2009,
AJ, 137, 3414

[6] O'Brien, T. J., Bode, M. F., & Kahn, F. D. 1992,
MNRAS, 255, 683 doi:10.1093/mnras/255.4.683

[7] Osborne, J. P., Page, K. L., Beardmore, A. P., et
al. 2011, ApJ, 727, 124
doi:10.1088/0004-637X/727/2/124

[8] Schönrich, R. A. & Ness, J.-U. 2008, in ASP
Conf. Series, 401, 291; RS Ophiuchi (2006) and
the Recurrent Nova Phenomenon, ed. A. Evans,
M. F. Bode, T. J. OBrien, & M. J. Darnley

[9] Wilms, J., Allen, A., & McCray, R. 2000, ApJ,
542, 914

225

http://dx.doi.org/10.1086/507980
http://dx.doi.org/10.1086/524054
http://dx.doi.org/10.1086/519676
http://dx.doi.org/10.1093/mnras/255.4.683
http://dx.doi.org/10.1088/0004-637X/727/2/124

	Introduction
	Observations and Analysis
	Contributions of shock emission to SSS continuum spectrum
	Spectral changes with variability during early SSS phase

	Summary and Conclusions