A&A 451, 157-175 (2006)
M. Eriksson1,2 - S. Johansson2 - G. M. Wahlgren2
1 - University College of Kalmar, 391 82 Kalmar, Sweden
2 - Atomic Astrophysics, Lund Observatory, Lund University, Box 43, 221 00 Lund, Sweden
Received 10 May 2005 / Accepted 4 December 2005
A detailed study of ultraviolet spectra of the symbiotic star AG Peg has been undertaken to derive the atomic excitation mechanisms and origin of formation for the lines common in symbiotic systems. More than 600 emission lines are observed in spectra from , and of which 585 are identified. Population mechanisms and origin of formation are given for a majority of those lines. Based on the understanding of the AG Peg spectra data of 19 additional symbiotic stars are investigated and differences and similarities of their spectra are discussed. Fe II fluorescence lines pumped by strong emission lines between 1000 and 2000 Å are observed in 13 of these systems. Some of the symbiotic systems belonging to the subclass symbiotic novae have more than 100 Fe II fluorescence lines in the ultraviolet wavelength region. Forbidden lines are detected for 13 of the stars, mostly from highly-ionized spectra such as Ar V, Ne V and Mg V. Further, [Mg VI] and [Mg VII] lines are observed in a symbiotic star (AG Dra) for the first time. Five of the symbiotic stars have broad white-dwarf wind profiles ( km s-1) for a few lines in their spectra. The stars with no such broad lines can be divided into two similarly sized groups, one where all lines have FWHM less than 70 km s-1 and the other where one, a few or all of the broad ( km s-1) lines of AG Peg have an enhanced broad wing (110-140 km s-1).
Key words: atomic processes - line: formation - stars: binaries: symbiotic - ultraviolet: stars
AG Peg is a symbiotic system consisting of a white dwarf and a red giant orbiting each other with a period of days (Kenyon et al. 1993). The luminosity of the white dwarf decreased by a factor of 4 during 1984-1997 (Kenyon et al. 2001). This decline can be understood as a decrease of its radius by a factor of 2 under constant temperature (Altamore & Cassatella 1997; Mürset & Nussbaumer 1994) or as an increase in the temperature of the white dwarf Schmutz 1996, which would imply a even larger decrease of its radius.
Between the two stars there is a region where the fast wind ( km s-1) (Nussbaumer et al. 1995) from the white dwarf collides with the slow wind ( km s-1) (Eriksson et al. 2004) from the red giant (Mürset et al. 1995). More recent and detailed ideas are discussed with respect to the wind structure concerning shock fronts from the wind collision region (Contini 2003) and the possibility of bipolar outflow (Yoo et al. 2002).
AG Peg belongs to a small subclass of symbiotic stars called symbiotic novae that have undergone nova events when the luminosity has increased by three to four magnitudes. Ultraviolet region line lists are available for two symbiotic novae, RR Tel (Penston et al. 1983) and V1016 Cyg (Nussbaumer & Schild 1981). Both of these systems have spectra dominated by allowed transitions from one to three times ionized elements, but also forbidden lines from more highly ionized elements have an impact on the spectrum. V1016 Cyg erupted in 1964 and RR Tel in 1944, while AG Peg erupted at least 150 years ago. An interesting open question is whether the decline of the UV flux in AG Peg during the 1980s is a natural evolutionary stage of symbiotic novae and whether, subsequently, RR Tel and V1016 Cyg will undergo a similar spectrum evolution in the future. A time analysis of the AG Peg spectrum can therefore give insights into what happens at the end of nova eruptions in symbiotic stars.
During the last 25 years the emission lines in AG Peg have been used for various diagnostics to measure parameters such as electron density (Keyes & Plavec 1980; Penston & Allen 1985) and temperature (Kenyon & Webbink 1984; Altamore & Cassatella 1997). By careful analysis of the spectra it is possible to categorize the emission lines in terms of population processes and line profiles. It is important to understand the origins of the emission lines as well as the processes that populate the corresponding upper levels when using them in any analysis. Here we explain the appearance of most of the emission lines formed at UV wavelengths in AG Peg. The main goal is to obtain an enhanced understanding of the emission lines, which would be helpful for anyone who wants to use emission lines as a diagnostic tool for symbiotic stars or related novae-like environments. Furthermore, the spectral investigation gives insight into how the UV spectum has evolved during the period 1978-1995.
Symbiotic stars are categorized in terms of stellar (S) and dusty (D) types based on their appearence at IR wavelengths (Allen 1982). The S-types have typical orbital periods of a few years and the cooler component consists of a solar-mass star on the red giant branch, while the D-types have orbital periods mostly in excess of 20 years and the cool component in those systems is typically a Mira. The small subset of D' type symbiotic stars has a cool star of spectral type G or K rather than M (Belczynski et al. 2000). The hot components are dominantly white dwarfs in both the S and D type systems. However, a few exceptions have the hot component as an accreation disk around a main sequence star (Kenyon & Garcia 1986).
Beyond the shared basic properties defining the group, the characteristics of symbiotic stars differ significantly. Besides the nova-like outbursts exhibited by the symbiotic novae, other phenomena observed in symbiotic stars are white-dwarf winds (Nussbaumer et al. 1995), recurrent-novae eruptions (Belczynski & Mikalojewska 1998), jets (Burgarella et al. 1992) and wind collision regions (Mürset et al. 1997).
By applying the understanding of the AG Peg spectrum at UV wavelengths to a set of 19 symbiotic stars we aim to detect differences and similarities among them. A detailed study of the lines showing broad white-dwarf wind lines in AG Peg is done to see how common such winds are in symbiotic systems. Fe II lines caused by photo-exitation by accidental resonance (PAR) are investigated to see whether this phenomenon is unique to a few symbiotic stars (Eriksson et al. 2003; Hartman & Johansson 2000) or more common among such systems. Also, forbidden lines of highly-ionized ions are analyzed to compare the ionization degree of low-density regions in symbiotic stars.
Table 1: Spectra used for AG Peg.
|Figure 1: The dots represent the noise level for each of the different orders in the longest exposed IUE spectra, LWP25995 and SWP47715. The lines represent the noise level in the and spectra used in this work. The y-axis gives the value of the logarithm of the noise level measured in erg cm-2 s-1 Å-1. The noise level is defined as twice the standard deviation of the pixel intensity from the mean intensity, measured in wavelength regions not affected by spectral lines.|
|Open with DEXTER|
The data recorded by the Goddard High Resolution Spectrograph (GHRS) onboard the Hubble Space Telescope (HST) during 1994 overlap in wavelength with the IUE data but were recorded at a later epoch and for only limited wavelength intervals. The greater sensitivity of the GHRS detectors makes it possible to measure intensities to a greater accuracy than can be done from the IUE spectra.
Two spectra of AG Peg have been recorded by the Far Ultraviolet Spectroscopic Explorer (FUSE) satellite, one in July 2001 and one in June 2003. The inclusion of FUSE spectra in our analysis increases the wavelength range for the detection of emission lines. However, since those spectra were obtained several years after the IUE and HST data we must be careful when incorporating these data sets into the analysis. The spectrum of AG Peg underwent obvious changes during the lifetime of the IUE (Eriksson et al. 2004) and we might expect that its development would continue in the sense that certain spectral lines would increase or decrease in intensity and width. Without contemporaneous observations in other wavelength regions, emission line intensities from the FUSE data cannot be rigorously correlated with the line intensities at longer wavelengths.
For the symbiotic stars other than AG Peg presented in this analysis only data are used. Since some of the Fe II lines as well as some of the forbidden lines are quite weak, one SWP and one LWP spectrum, observed with large aperture and with long exposure time, were selected for each system. The emission lines associated with the white-dwarf wind in AG Peg were easily saturated in the spectrum, which is why an extra SWP spectrum was retrieved from the MAST archive for some of the symbiotic systems. Table 2 lists the data used for the selected objects.
Table 2: IUE spectra used for the selected symbiotic stars.
The spectral lines can be divided into different categories depending upon their widths and profiles. However, lines that look similar in one spectrum can show a different appearance in later spectra. This means that the time history of a spectral line is also important. Some narrow lines are selectively populated by radiative line excitation and they are classified as fluorescence lines, whereas others originate from excited levels populated by collisions or recombination. We have categorized the spectral lines by their overall appearance in the following way:
Table 3: The distribution of emission lines for AG Peg, according to population processes.
Table 4: The broad emission lines in AG Peg spectra.
In IUE spectra recorded before 1986 there are 22 emission lines with a FWHM larger than 600 km s-1 (Table 4). These lines originate from abundant elements in higher ionization stages than for the narrow lines, and they are presumably formed in the hot white-dwarf wind. An analysis of some of these lines has yielded that the terminal velocity of the white-dwarf wind is 900 km s-1 (Vogel & Nussbaumer 1994). A few of the lines are ground-state transitions (resonance lines) and show P Cygni structure while others are LS-allowed transitions having an excited lower level. In a few cases the transitions are spin-forbidden (intercombination lines). After 1986, when the broad wind lines were replaced by narrow emission (Fig. 2), it is notable that the six lines with highest peak intensity in the spectra (He II 1640, C IV 1548, 1550, N IV] 1486, N V 1238, 1242) had broad line profiles in earlier spectra.
He+ has a low ionization energy compared to the other emitting ions detected in the white-dwarf
wind, which is why the He II lines can be considered as recombination lines.
The He II lines with wavelengths in the range of
and the whole Paschen series except Pa.
He II Balmer
is close in wavelength to H Ly
and therefore totally
absorbed. The He II lines detected in
spectra of AG Peg are He II Balmer
and 13 consecutive Pashen lines starting with Pa
reaching upper levels with quantum numbers n=5-17. Only the
second to the sixth lines can be detected from the noise before 1986 when the lines were broad.
|Figure 2: The development of He II Balmer . The spectrum to the left (SWP02326) was recorded in 1978 and the spectrum to the right (SWP40148) was recorded in 1990. The large change in profile width is related to the change in origin for He II emission between those dates. In 1978 most of the He II emission originated from the white-dwarf wind while in 1990 the emission instead came from the heated part of the extended red-giant atmosphere.|
|Open with DEXTER|
For a few of the white-dwarf lines (C IV 1548, 1558, N IV]1486, Si IV 1402, Si IV 1393) the development into narrow "nebular'' lines had already started in 1978, and the structure in the line profile, as shown in Fig. 3, is caused by the narrow line superimposed on the broad line. That the contribution of narrow emission lines occurred earlier for the ions of lower ionisation potential gives an idea into why the emission lines changed origin. As the white dwarf emerged at the end of the nova-like outburst the wind became less dense and more transparent for UV photons. The line intensities from the white-dwarf wind would then decrease at the same time that the material in the extended part of the red-giant atmosphere facing the white dwarf would be subjected to more of the white-dwarf UV radiation. The heated part of the red giant atmosphere became hotter and more ionised, resulting in the strong, narrow-emission lines. In recent observations of AG Peg it is clear that the O VI resonance doublet still has contributions from the white-dwarf wind (Fig. 8). Since the ionisation energy of O VI is 114 eV its presence would be expected with a temperature increase in the white-dwarf wind, contrary to what is given by Zanstra determinations (Altamore & Cassatella 1997).
The O III Bowen lines were also broader than the nebular lines in early spectra. Like the white-dwarf lines discussed earlier the O III Bowen lines also evolved into narrower lines during the 1980s. The reason that they are not included in Table 4 is that they have less than half of the FWHM of the other broad line profiles and are formed in a different way (fluorescence). The Bowen lines are discussed in Sect. 3.3.
|Figure 3: The IUE spectrum (SWP02334) of AG Peg, recorded in 1978, illustrates the difference in the profiles between lines from ions of ionisation energy near 40 eV and those of higher ionisation energy (around 70 eV). The N IV] 1486 had a contribution from a "nebular line'' already in 1978 while all flux in the N V resonance doublet came from the white-dwarf wind.|
|Open with DEXTER|
A majority of the emission lines observed in IUE spectra of AG Peg have a FWHM between 10 and 50 km s-1 and have had the same line profile although changed in intensity between 1978 and 1995. There is a wide range in ionisation energy for the narrow lines, from singly-ionised elements to more highly-ionised species such as Mg V, S IV and Fe V. Most of the emission lines correspond to LS-allowed transitions or intercombination lines but there are also a few parity-forbidden lines such as [Mg V] 2928.
The process of photo-excitation by accidental resonance (PAR) can account for as many as 218 of the 585 identified lines in spectra of AG Peg. The Fe II levels y4H11/2 and w2D3/2 are known to be pumped by C IV 1548 in at least eight symbiotic systems (Johansson 1983; Eriksson et al. 2001). The levels (5D)5p 6F9/2 and (3F)4p 4G9/2 are known to be pumped by H Ly in the symbiotic star RR Tel (Johansson & Jordan 1984; Hartman & Johansson 2000) and they are involved in laser action in gas condensations of Car (Johansson & Letokhov 2004). In an analysis of AG Peg (Eriksson et al. 2003) 29 Fe II levels were shown to be populated by the PAR mechanism, and in the present work 11 newly identified pumped channels lead to a total of 40 presumably pumped Fe II levels, of which 22 are confirmed by three or more fluorescence lines (Table 5).
Table 5: Pumped Fe II channels confirmed by 3 or more fluorescence lines.
Most of the spectral lines that pump (selectively excite) Fe II are formed in the white-dwarf wind. Since the profiles of those lines change as their origin evolves from being dominated by the white-dwarf wind to the heated part of the red-giant atmosphere (or wind) the emission from the pumped Fe II levels also changes after 1986. Fe II channels that are separated by more than 40 km s- 1 from their pumping line cannot be pumped by the narrow emission lines and therefore those lines disappeared after 1986 (Fig. 4). However, the Fe II fluorescence lines excited in channels that are close in wavelength to their pumping lines can be pumped by the narrow "nebular'' emission lines and they do appear in the spectrum throughout the lifetime of IUE. There are four lines from highly- ionised elements (Si III] 1892, O III] 1660, 1666 and O IV] 1401), which are not white-dwarf wind lines but still responsible for PAR processes in Fe II. These lines have in common that they are narrower than the other pumping lines, even compared to the FWHM after 1986, and that they, along with the corresponding fluorescence lines, did not change during the years 1978-1995.
In AG Peg H Ly activates numerous Fe II channels resulting in fluorescence lines from both primary and secondary decays of the pumped levels. Lines from the (5D)5s levels (i.e. the e6D and e4D terms) are seen in the AG Peg spectrum. Those lines are also seen in RR Tel, where they are explained as secondary decay from H Ly pumped (5D)5p levels (Hartman & Johansson 2000), and we conclude that the same explanation is valid for AG Peg. A consequence of the density decrease in the white-dwarf wind is that it becomes less dominant (less extended in the binary system) as well as less opaque so that the H II region can grow and more H Ly radiation can reach the Fe+ ions. This leads to an increase of the intensity of the H Ly pumped Fe II fluorescence lines (Fig. 5). It is interesting to note that the velocity shift is the same for the Fe II fluorescence lines but differs from the velocity shift of the lines from collisionally excited Fe II levels.
|Figure 4: The total intensity of all Fe II lines from the upper level y2D5/2, which is pumped by the C IV 1548.187 line through the channel a4F7/2-y2D5/2 at 1548.697 Å. When the broad foot of the C IV resonance lines vanished the activity in this channel stopped.|
|Open with DEXTER|
|Figure 5: The strength of the 2508.34 Fe II line representing the channel a4D7/2-(b3F) 4p 4G9/2 pumped by H Ly. As the temperature increased the H II region around the white dwarf became larger, which allowed more H Ly to reach the Fe II regions.|
|Open with DEXTER|
We observe lines from Fe II levels that are not selectively photo-excited by strong lines. The population of these levels can in general be explained by collisional or recombination excitation (see later sections). However, we also observe emission lines from 21 Fe II levels of medium excitation energy ( -8 eV), for which neither of the excitation mechanisms mentioned above is plausible. Absorption lines of Fe II at short wavelengths are observed, and they give an explanation to the population of seven of those 21 levels (Fig. 6). These are levels belonging to the LS terms y4D, z2F and x6P (except for y4D1/2 and x6P3/2), and they are photo-excited by continuum radiation from the white dwarf through the channels a4F-y4D, a4P-z2F and a6D-x6P. Only the strongest transitions between these terms are observed as absorption lines in AG Peg, probably due to the low continuum level.
|Figure 6: Grotrian diagram of Fe II showing the levels involved in continuum fluorescence. The dotted lines are the pumped channels while the solid lines represent decay channels observed in AG Peg.|
|Open with DEXTER|
Table 6: Pumped channels in ions other than Fe II.
The strong emission line 1411.94 in AG Peg is identified as the 2p3 2P3/2-2p23s 2D5/2 transition of N I but no emission is detected from the 2p23s 2D3/2 level. We suggest that the level 2p23s 2D5/2 is selectively populated by being pumped by N V 1242 in the absorption channel 2p3 2D5/2-2p23s 2D5/2. A similar case is 1409.34, which is identified as a transition from the S I level (4S)5s 3S1 having a high excitation energy (8.85 eV). We suggest the S I level is pumped by O IV] 1401. Furthermore, two pumped channels in Co II and two in Fe III have been detected in the spectrum of AG Peg (Table 8).
Radiative transitions between levels of the same parity must be formed by magnetic dipole (M1) or electric quadrupole (E2) interaction, which means many orders of magnitude smaller transition probabilities than electric dipole (E1) transitions including a parity change. Observations of M1 and E2 radiation from an astrophysical plasma require that the density of the plasma is low enough so the radiative lifetime of the metastable state is smaller than the time scale for deexcitation by collisional quenching. Hence, the lines from the parity forbidden transitions originate from low density regions, such as from a nebula surrounding the system. In the IUE spectra of AG Peg, 14 parity-forbidden emission lines have been identified, originating from highly-ionised elements such as Ne V and Fe VI (Table 7). The ionisation energies are very high for the ions emitting M1 and E2 radiation, which implies that all parity-forbidden radiation originates from a thin plasma, heated and ionized by the UV radiation from the white dwarf. There is a difference among the forbidden lines in IUE spectra before and after 1986. Before 1986 only six of the 14 observed parity-forbidden lines were present, and they originated from the ions O2+, Ne2+ and Ne3+. No forbidden lines from higher ionisation stages were observed. After 1986 the forbidden emission from O2+ ([O III] 2321.66) vanished and the emission from Ne2+ ([Ne III] 1814.63) became weaker, while eight new emission lines from Ar4+, Ne4+, Mg4+ and Fe5+ appeared in the IUE spectrum.
Table 7: Parity forbidden lines observed in AG Peg.
Emission lines from a particular ion can originate from different regions in the symbiotic system. Still, after discarding the emission lines that originated from the white-dwarf wind before 1986 as well as all fluorescence lines and parity-forbidden lines, the remaining emission lines from a specific ion have about the same velocity shifts. The widths of these remaining emission lines did not change during 1978-1995, but the relative shift between different ions was variable. In Table 3 all ions observed in AG Peg are listed with the number of corresponding emission lines that are formed by collision or recombination, excluding the white-dwarf wind emission lines. We will now give suggestions for how and where some of those lines are formed.
Four emission lines from the He I 2s 3S1-np 3P2 series
(n=5-8) are identified in the RR Tel spectrum Penston et al. 1983. In the
IUE spectra of AG Peg recorded before 1986 there are no traces of He I
emission. The broad He II emission lines from the white-dwarf wind before 1986
are presumably formed by recombination. Because of the high temperature and
strong UV flux in the white-dwarf wind there is no neutral helium there.
However, after 1986 seven He I emission lines of the series
2s 3S1-np 3P2 (
are present in the spectrum as "nebular'' lines with a mean FWHM of 35 km s-1. The explanation of the He I emission
is probably linked to the change of the He II lines from broad wind profiles to
"narrow'' lines having an average FWHM of 49 km s-1 (see Sect. 3.1).
Kenyon et al. (1993) suggest that the presence of He I is caused by heating of
the red-giant atmosphere by radiation from He+ ions in a region
around the white dwarf. If He II in the white-dwarf wind is responsible for He I
emission from the red-giant atmosphere, then the absence of He I lines before
1986 is a problem. Another suggestion is that when the white-dwarf
wind became less dense during the transition period it was not able to produce broad He II emission
enough to be observed by ,
but at the same time more continuum radiation
from the white dwarf reached the red-giant atmosphere, ionising helium to both
He+ and He2+ (Fig. 7). This implies that after 1986 both the He I and He II
emission lines are formed in the red-giant atmosphere facing the white
|Figure 7: Suggestion of the origin of broad He emission ( left) and narrow He emission ( right) before and after 1986 (see text).|
|Open with DEXTER|
It is only possible to detect the 2s 2S-2p 2P doublet transition for three species, C IV, N V, and O VI, within the wavelength regions covered by and . Atomic lithium and the ions Be+ and B2+ are too rare to be observed and for the three-electron systems heavier than oxygen the 2s-2p transition lies shortward of 900 Å. Two of the three observable doublets, C IV 1548.19, 1550.77 and N V 1238.80, 1242.78 are observed as broad (800 km s-1) emission lines that evolve into narrow emission (40 km s-1) during the 1980s as discussed in Sect. 3.1. The O VI 1031.92, 1037.614 resonance doublet is outside the range of , but is observed with (Fig. 8). Since was launched in 1999 there have been no data on the O VI resonance doublet in AG Peg prior to the dramatic change of profile of the N V and C IV resonance doublets. However, at the base of the O VI resonance lines there are wings (FWHM 800 km s-1, based on Gaussian fit of the line wings in the Q1110101 spectra) indicating that the O VI doublet evolved in the same way as the C IV and N V doublets. The peak intensity of the nebular components appears to have increased from 2001 (Q1110101) to the 2003 FUSE observation (Q1110103) while the intensity of the broad underlying wind-profile decreased slightly. This could mean that the same transfer of origin of the C IV resonance doublet at the end of 1970s to the beginning of the 1980s and the N V resonance doublet during the 1980s and beginning of 1990s (Eriksson et al. 2004) is now taking place for the O VI resonance doublet.
For the iso-electronic sequence of spectra
involving 11 electrons the 3s 2S-3p 2P doublets are reachable with
three different elements (Mg II
1854.72, 1862.79 and Si IV
1402.77), and one with
1117.98, 1128.01). The P V
resonance doublet is not detected in the FUSE data probably because of a low
abundance of phosphorus in the red-giant atmosphere. Energetically it would be
possible since the O VI resonance doublet is observed, and the ionisation
potential of O V (113.9 eV) is higher than for P IV (51.4 eV). The Si IV
resonance lines evolved in the same manner as the C IV and N V resonance lines
and are discussed in Sect. 3.1. The Mg II resonance doublet (Fig. 8) is
known to be common in the chromospheres of red-giant stars (Kondo et al. 1976) and is
probably emitted all around the red giant atmosphere. Both the Mg II and Al III
resonance doublets consist of narrow emission lines throughout 1978 to 1995 and
have probably never been formed in the white-dwarf wind because of the low
ionisation temperature required.
|Figure 8: The ns 2S-np 2P doublets observed in AG Peg. The figures represent the profile of the resonance doublets after the transition period in the middle of the 1980s. The resonance doublets corresponding to n = 3 (Si IV, Al III, Mg II) and C IV almost completely consist of narrow components, while the N V and O VI resonance doublets still have detectable contributions from the broad white-dwarf wind profile.|
|Open with DEXTER|
The spin-forbidden ns2 1S-nsnp 3P transition includes two possible intercombination lines: 1S0-3P1, which is often strong in emission in hot nebular spectra, and 1S0-3P2. Since for the latter line, it can only occur via a magnetic quadrupole transition, which has a very small transition probability. Similar to the observational restrictions for the 2s 2S-2p 2P transitions discussed in the previous subsection, the 2s2 1S-2p 3P transitions can only be observed for C III], N IV] and O V]. The C III] 1908.73 (1S0-3P1) emission line is observed as strong ( erg cm-2 s-1 Å-1) and narrow ( km s-1) in all of the IUE observations. This is probably due to the ionization energy of C III, which is too low for C2+ ions to survive in the white-dwarf wind, but the ionisation potential of C II is not too high for C2+ to be produced in the heated part of the red-giant atmosphere. At the detection limit in the longest exposed spectra at 1906.67 Å there is a weak feature ( erg cm-2 s-1 Å-1), which can possibly be identified as the E2 (1S0-3P2) transition of C III] blended with Mg IV.
The N IV] 1486.50 (1S0-3P1) emission line is a broad wind line before 1986 and is observed as a narrow emission line after 1986 as discussed in Sect. 3.1. Interestingly, the [N IV] M2 transition at 1483.32 Å is present throughout the observations but has a FWHM around 150 km s-1, which is very different from the FWHM of N IV] 1486.50 both before and after 1986. Since O V has a higher ionisation energy than N IV one would also expect the O V] 1218.34 emission line to have evolved from a broad wind profile to a narrow nebular line. However, its relative closeness to H Ly at 1215.67 Å causes absorption by the neutral hydrogen in the surrounding nebula and the O V] 1218.34 line is absent before 1986. When the temperature of the white dwarf increased the H II region expanded further out into the surrounding nebula, making it more transparent for the O V] 1218.34 emission, which after 1986 is observed as a strong emission line.
In the Mg I iso-electronic sequence of 12-electron ions the transition 3s2
1S0-3s3p 3P1 is observed in four elements (Al II], Si III],
P IV] and S V]). The S V] 1199.04 feature evolved from a broad wind
profile to a narrow nebular line during the 1980s, while Al II]
2669.95, Si III] 1892.03, and P IV] 1467.43 are
observed as narrow (FWHM 30 km s-1) throughout the time
interval 1978 to 1995. The intensity ratio I(1S0-3P1)/I(1S0-3P2) has been used in C III and N IV
as a diagnostic tool to obtain electron densities in AG Peg Nussbaumer & Schild 1979,Nussbaumer & Schild 1981.
The same transitions can be used for O V, Al II, Si III, P IV and S V to derive
complementary information about the plasma. Understanding the origin of the
1S-3P lines is therefore of great importance.
|Figure 9: The intercombination multiplet 2s22p 2P-2s2p2 4P in N III] and O IV] spectra. The fine-structure component 3/2-5/2 is over-exposed in the N III] plot. The region of the N III] multiplet has not been observed by HST (GHRS) at high resolution. The dynamic range of was not sufficient to record spectra where the weakest N III] components are observable at the same time that N III] 3/2-5/2 is not over-exposed.|
|Open with DEXTER|
The more ionised an element is the more sensitive the average energy of a configuration is to the principal quantum number, n. Excited states in highly-ionized spectra, for which all electrons have the same n as the ground configuration, are therefore at relatively low excitation energies. As an example the LS term 2s2p2 4P of C II is at 5.3 eV, while the LS term 2s2p3p 4P is at an energy of (23.1 eV), i.e. more than a factor of four higher. Because of the low energy required for exciting the (sp)k configurations they can be populated through collisions. Allowed emission lines from (sp)k, 2<k<8, dominate the contribution from C II, N III, O IV, Si II, P II, P III, S II and S IV and they are also present in C III, N IV and O V. An intercombination transition, ns2np 2P-nsnp2 4P, is specially strong in N III, O IV and S IV (Fig. 9), and is also present in the Si II spectrum. The metastable nsnp2 4P is the lowest excited term in systems with five and 13 electrons and it can therefore only decay to the ground state, which explains why the corresponding multiplets are so strong. Those multiplets can be used for electron density diagnostics (Nussbaumer & Storey 1982; Nussbaumer & Storey 1979).
Emission lines from NI, OI and SI are observed in the ultraviolet spectrum of AG Peg. Most of the emission lines from neutral elements are from allowed transitions to levels within the ground configuration, except for the ground state (Fig. 9). That the emitting gas is optically thick for transitions to the ground state indicates that the region of neutral elements is of relatively high density and low temperature (<3000 K). All of the four observed oxygen lines and five of the six observed sulphur lines correspond to transitions from np3(4S)n's 5S or 3S to the ground term np4 3P. That the 3P-5S intercombination lines of S I 1900.29 and O I 1355.60, 1358.51 are observed is interesting since their transition probabilities are 104 times smaller than for the 3P-3S transitions. This limits the electron density of the emitting region, presumably the outermost part of the surrounding nebula or the part of the red giant obscured from the white-dwarf emission.
|Figure 10: The O I multiplet 2p4 3P-2p3(4S)3s 3S. The component 3P2-3S1to the ground level is in absorption while the other two fine-structure components are in emission.|
|Open with DEXTER|
The population of some levels, from which emission is observed in AG Peg, cannot be understood by collisions or any pumping mechanisms. These include levels with very high excitation energy having no known decay channels that coincide with strong UV emission lines of other elements, and exotic configurations (doubly-excited and/or involving f-electrons) rarely populated in astrophysical plasmas. If an ion is emitting lines from doubly-excited levels or levels close to the ionization energy, recombination is the most plausible explanation. Based on this reasoning we suggest recombination as the excitation mechanism for some lines. The previous discussion of the strong 2s 2S-2p 2P transitions reveals a large abundance of C IV, N V and O VI in a hot region of AG Peg, implying that recombination to C III, N IV and O V is reasonable. Both C III and O V lines have been observed from levels rarely populated by collisions and are most certainly populated by recombination (such as 4f 3F4 in C III and 5f 3F4 in O V). Although the N V 1238.80, 1242.78 doublet is almost as strong as the C IV resonance doublet none of the observed N IV lines originate from levels that require recombination excitation. However, the N IV] 1486.50 emission line is one of the strongest lines in the wavelength range, indicating high presence of N3+ ions and the N III spectrum shows a few lines, such as N III 1387.38, with upper level 4d 2D3/2 presumably populated by recombination. Also, recombination lines of the third spectra, C III, N III and O III, were present throughout 1978-1995 and are formed in the extended red-giant atmosphere at a distance from the white dwarf sufficient to provide a temperature that gives a mixture of two and three times ionised elements.
The O V recombination lines first appeared after 1986. Before 1986, O5+ ions probably only existed in the white-dwarf wind, but after the transition period in the middle of 1980s O5+ ions were produced in the red-giant atmosphere by photo-ionisation, which explains the appearance of O V recombination lines. Four emission lines between 1669-1699 Å are identified as Mg IV lines corresponding to the multiplet (3P)3s 4P-(3P)3p 4D. These lines are the only observed Mg IV lines in the spectra of AG Peg, and the (3P)3p 4D term has an excitation energy of 75 eV. The population of these levels is likely to be due to recombination of Mg V 2p4 3P in less dense parts of the surrounding nebula since only parity forbidden lines from Mg V are observed.
Table 8: The relative line strengths of the Fe II multiplet a6D-z6D.
The lowest excited odd-parity levels in Fe II belong to the subconfiguration
(5D)4p, which is built on the lowest parent term 5D of Fe III. This
subconfiguration contains six LS terms (4P, 4D 4F, 6P,
6D and 6F) and accounts for 25 fine-structure levels. Transitions
from (5D)4p are observed in the ultraviolet region to levels within the
terms (5D)4s 6D, (5D)4s 4D, 3d7 4F and 3d7
a4P forming strong Fe II multiplets. Since transitions from all the 25 levels are observed, collisional excitation is the most likely population
process. However, the decay of the (5D)5s levels, populated by the
cascading decay of the H Ly
pumped (5D)5p levels (see Sect. 3.2.1), causes deviations
in the thermal population of (5D)4p. Thus, the observed line strengths
within the (5D)4s-(5D)4p and 3d7-(5D)4p multiplets are
different from what is expected from their gA-values, as measured by Schnabel et al. (2004) (Table 8).
Also of importance is that the transition to the ground level (5D)4s
6D9/2 is in absorption, which indicates that the optical depths also
will influence the relative line strengths.
|Figure 11: Figure demonstrating our idea of the excitation mechanism in Fe III and Fe II responsible for the observed octet transitions in Fe II. The presumed recombination channels are presented as dotted lines, the pumped channel as a double lines and the decay channels as solid lines. Only the levels involved in the process are plotted in this Grotrian diagram.|
|Open with DEXTER|
|Figure 12: The three O III 2p3p 3S-2p3d 3P transitions in spectrum lwp25995. Whit the dynamical range of IUE it is not sufficient to observe the 3S1-3P0 transition without having the 3S1-3P2 transition saturated because of the large difference of intensity.|
|Open with DEXTER|
Three emission lines ( 1926.07, 1915.62 and 1887.83) have been identified as originating from the z8P term in Fe II, which has no spin-allowed decay channels. These three lines require an explanation since the Fe II octet transitions are very rare in stellar spectra. In fact, they have only been observed in the spectrum of the sun Johansson 1977.
If Fe III is to recombine to any octet level of Fe II by electron capture the metastable septet level (6S)4s 7S3 of Fe III has to be populated. The PAR mechanism in Fe III results in population of the (6S)4p 7P levels. Furthermore, by the observed Fe III fluorescence lines at 1955.27, 1914.06 and 1895.46 Å corresponding to (6S)4s 7S-(6S)4p 7P transitions the metastable Fe III (6S)4s 7S3 level is known to be populated. In Fig. 11 our suggestion of the process leading to the Fe II octet transitions is shown.
The first known PAR process was the pumping of O III through the channel 2p2 3P2-2p3d 3P2 at 303.80 Å by He II Ly at 303.78 Å Bowen 1934,1935. In the spectra of AG Peg emission lines corresponding to 2p3p 3S-2p3d 3P and 2p3p 3D-2p3d 3P transitions are observed, which concludes that the Bowen mechanism is active in the system. The possibility of allowed primary decays from 2p3d 3P to 2p3p 3P, which form emission lines around 3430 Å, is outside the range. However, all six possible emission lines corresponding to the secondary decays 2p3s 3P-2p3p 3P are observed and hence, the 2p3p 3P-2p3d 3P is certainly active. Also, two emission lines, O III 3341.73, 3300.34 are observed as secondary decays from 2p3s 3P-2p3p 3S transitions.
It is not only the 3P2 level that is populated by PAR in AG Peg, the two other fine structure levels, 3P1 and 3P0, are also populated through channels within 2p2 3P-2p3d 3P between 303.41 Å and 303.80 Å (Eriksson et al. 2005). From the three emission lines corresponding to 2p3p 3S-2p3d 3P (Fig. 12) the relative population rate of the fine structure levels in 2p3d 3P can be estimated. The theoretical relative LS intensities for the transitions (3S1-3P2:3S1-3P1:3S1-3P0) are 100:60:20 and the observed relative intensities, which can not be measured exactly because of saturation of the strongest line, are 100: < 20: < 3. The population rate of 3P1 and 3P0 is less than for 3P2 by a factor of at least 3 and 7, respectively.
When the observed Balmer and Pashen lines of He II became narrower by a factor of 20 during the middle of the 1980s the He II Ly pumped O III lines also changed in appearance. In spectra recorded in 1981 and earlier the FWHM of the Bowen lines was 120 km s-1, which was a factor of 6 smaller than that of the white-dwarf wind lines but more than a factor of 3 broader than the emission lines from the nebula and the red-giant atmosphere. The FWHM of the O III Bowen lines then decreased and was 45 km s-1 in the early 1990s. The widths of the emission lines associated with the white-dwarf wind in early spectra decreased more rapidly during the 1980s, and the Bowen lines were actually among the broadest UV emission lines in spectra of AG Peg in the early 1990s. After 1986 the net flux of the O III Bowen lines increased by 50%. This could mean that after the He II region changed location from the white-dwarf wind to the heated part of the red-giant atmosphere more He II emission was able to reach the O2+ ions.
All emission features observed in this work are listed in Table 13. Lines longward of 1980 Å were observed with (), the lines between 1170 and 1980 Å with () and the lines below 1170 were observed with ( and ). In order to make comparison of line strengths useful for lines within the wavelength range of each instrument all peak intensities and FWHM are tabulated as measured in the same spectra, LWP25995 for 1980-3350 Å, SWP47715 for 1170-1980 Å and Q1110101 for lines below 1170 Å, which we call the reference spectra. A few emission features are present in spectra other than the reference spectra. These features are marked with "oth'' in the first column in the table and no information about intensity or FWHM are given in Table 14. The reference spectra were selected as such since they have the lowest noise level, which has the disadvantage that the strongest lines are saturated and no information about peak intensity or FWHM can be given in Table 13. The saturated lines are marked by "oe'' in the second column. In Cols. 4-7 the identifications of the observed lines are presented. For the unidentified lines those columns are left blank. Features that are considered to be blended with unknown features are marked with the superscript "bl'' on the measured intensity in the second column. In Col. 8 we have noted the subsection to which the reader can refer for our suggestion of the process responsible for the formation of the feature. For the wavelength intervals covered by the (GHRS) observations Col. 9 presents the intensities.
In Table 14 all of the emission lines absent in the reference spectra are given with the observed wavelengths (Col. 1), peak intensities (Col. 2) and widths (Col. 3). The identifications are given in Cols. 4-7 as in Table 13. Column 9 lists the spectra from which the lines were measured.
Table 15 lists all of the lines that were over-exposed in the reference spectra and is constructed in the same manner as Table 14.
Table 16 is a list of the absorption lines in AG Peg. The first column gives the observed wavelength. If the wavelength is superscripted with a; the line is measured in LWP09698, b; SWP15651 and c; SWP10454, while all other lines are measured in the reference spectra. The superscript "*'' means that the line is not measured in this work but that its presence is revealed by profile fitting of the N V and C IV resonance lines (Eriksson et al. 2004). The equivalent widths of the lines are given in Col. 2 and the identifications in Cols. 3-6. In Col. 7, IS stands for interstellar line and bl for line blended by unknown contributors.
PU Vul is one of the symbiotic stars classified as symbiotic novae Vogel & Nussbaumer 1992, just like AG Peg. The Si IV 1393 line shows a P Cygni profile with terminal velocity of 750 km s-1, which is similar to the white-dwarf wind velocity of AG Peg (Eriksson et al. 2004). We suggest, therefore, that the broad lines in PU Vul can be explained by a fast wind from the white dwarf triggered by the slow nova outburst. EG And is known to have a fast (500 km s-1) wind from the white dwarf (Vogel 1993), which can explain the broad profiles underlying the narrower nebular lines in its spectrum. Interestingly, in 1995 the C IV 1548 line displayed a P Cygni profile associated with a terminal velocity of only 114 km s-1. We have not been able to explain this absorption component but modeling of the same line in AG Peg has revealed an absorption component. This is the collision region of the white dwarf and red giant winds associated with a terminal velocity of half the white dwarf wind velocity. This should be noticed since EG And also has a region where the winds from the two stars collide (Tomov 1995). More surprising are the broad lines in T CrB and CH Cyg, since the white dwarf in these systems is known to be accreting matter from the red-giant wind (Iijima 1982; Kenyon & Garcia 1986), and therefore no white-dwarf wind is expected. The argument of matter falling into the disk surrounding the white dwarf is further strengthened here as inverted P Cygni profiles are observed in the lines C IV 1548 (both systems) and He II 1640 (only in T CrB). Even if we cannot explain how the lines are broadened to over 700 km s-1 the reason can be the infall of matter on the disk surrounding the white dwarf.
The 15 symbiotic systems with no trace of lines having FWHM greater than 400 km s-1 can be divided into two distinct categories with respect to the lines showing broad wind profiles in AG Peg. For eight of the stars all of those lines have FWHM between 40 and 70 km s-1, which is normal for most of the lines observed by in spectra of symbiotic stars. Such lines are called nebular lines and their origin being the red giant upper atmosphere irradiated by the white dwarf. For the remaining seven systems some of those emission lines have a FWHM between 110 and 140 km s-1. Common to these latter systems is that they all have underwent outbursts which could lead to an outflow of matter from the white dwarf. In Table 9 the widths of five lines associated with the white-dwarf wind in a few of the symbiotic systems is given for 20 symbiotic systems.
Table 9: The nature of lines having white dwarf wind profile in AG Peg in symbiotic stars.
In the present study ultraviolet Fe II fluorescence lines have been searched for in the symbiotic stars listed in Table 2. The presence of the Fe II lines 2507.55, 2509.10 in emission serves as an indication of H Ly pumping, while lines from the C IV pumped Fe II levels (see Table 5) serve as indicators of an Fe II region subjected to radiation from a high-ionization region. Absence of fluorescence lines in IUE data of a symbiotic star is not a sufficient reason to conclude that there is no pumping of Fe II. The lower limit of the peak intensity, for a line to be detected with the , is around erg cm-2 s-1. However, in systems where the (5D)4s-(5D)4p resonance transitions but still no Fe II fluorescence lines are observed fluorescence can be ruled out or assumed to play a negligible role. Therefore, a search for Fe II lines corresponding to the a6D-z6D and a6D-z6P multiplets, which normally form the strongest Fe II lines in the IUE wavelength domain, has been carried out. The results of this search are presented in Table 10.
Among the twenty symbiotic systems considered in this study six belong to the subgroup of symbiotic slow novae: AG Peg, HM Sge, PU Vul, RR Tel, PU Vul, V1016 Cyg and V1329 Cyg. Except for HM Sge, both H Ly and C IV 1548 pumping of Fe II occurs in these systems, as observed by Eriksson et al. (2004). Emission lines in HM Sge are, in general, one magnitude fainter than in other symbiotic novae and even the Fe II resonance lines cannot be observed in the IUE spectra. This anomaly for HM Sge could indicate that it has no significant Fe II region or that the S/N ratio of is not high enough to see the Fe II emission.
Three of the symbiotic systems (CH Cyg, CI Cyg and T CrB) are known to have accreting disks around their hot components. As pointed out by Eriksson et al. (2004) no Fe II fluorescence can be detected in those systems. However, the Fe II resonance lines are observed in CH Cyg, which implies that there might be a Fe II region but no fluorescence in symbiotic systems involving accretion onto the hot component.
It becomes more problematic when considering the results of the 11 "normal'' symbiotic stars, which do not belong to the subgroup symbiotic novae or show any sign of a disk. Four of those systems, BF Cyg, HBV475, AG Dra and KX Tra, have no detectable Fe II emission lines in the spectra, and they are in a way similar to the symbiotic nova HM Sge as regards Fe II. The systems RW Hya and R Aqr show Fe II fluorescence lines from levels pumped by high-ionization lines only. However, spectra of the systems EG And, SY Mus, Z And and RX Pup have Fe II lines from both C IV 1548 and H Ly pumped levels. Even if neither is designated as a symbiotic nova, recurrent outbursts have been observed for Z And (Tomov et al. 2003) and RX Pup (Mikolajewska et al. 1999). This could mean that the dynamics and structure of a symbiotic system change during the outbursts (both for recurrent and slow novae) so that emission from both highly-ionized regions and H I regions can reach the Fe+ ions. Then the group of symbiotic stars having numerous fluorescence lines (Eriksson et al. 2004) would be expanded to include all symbiotic stars associated with outbursts and not only symbiotic novae. A complication to this simple idea is that outbursts also have been observed in the symbiotic stars CI Cyg and AG Dra (Belczynski et al. 2000), which do not have numerous fluorescence lines. Since CI Cyg is, as discussed in the previous paragraph, a disk-system for which no Fe II fluorescence has been observed and AG Dra is known to be metal deficient, the lack of Fe II emission lines in these two systems is no surprise. Two normal symbiotic stars seem to fall outside the established groups: SY Mus, which emits Fe II fluorescence from both H Ly and C IV 1548 pumped levels, and AX Per, which is the only symbiotic star known to emit Fe II lines only from H Ly pumped levels.
Table 10: Fe II emission in symbiotic stars.
Forbidden lines are detected in spectra for 14 of the 20 selected symbiotic stars (see Table 11). [Mg V] is seen in the spectra of eight of the symbiotic stars, and is the most frequently observed among the forbidden lines. There are four systems showing forbidden lines from magnesium: EG And, RW Hya, CH Cyg and R Aqr. EG And has forbidden lines from [O II], [Ar III] and [Ne III], RW Hya from [O III] and [Ar III], CH Cyg from [O III] and [Ar III] and R Aqr from [O II] and [O III] and [Ar III]. The absence of [Ar V] and [Ne V] in these four systems indicates that the temperature in the low density region is too low for the formation of Mg4+. The system AG Dra has only forbidden lines from [Mg VI] and [Mg VII]. The relative locations of the forbidden line regions within the symbiotic stars, and thereby the radiation field they are subjected to are poorly known. Therefore, a more detailed analysis of the forbidden line regions are not possible at this point. Assuming local thermal equilibrium, the forbidden lines observed can be explained by one region of one dominant temperature in 6 of the 14 systems showing forbidden lines (see Table 11).
Table 11: Forbidden lines searched for in 20 symbiotic stars.
Table 12: Forbidden spectra in symbiotic stars.
Table 13: A sample of the AG Peg line list.
During the 1980s the spectrum of AG Peg changed remarkably. Before 1986 six forbidden lines were observable in IUE spectra: [O III] 2321.66, [Ne III] 1814.63 and four [Ne IV] lines. After 1986 the [O III] line has disappeared while six new lines from four times ionized ions (Ar4+, Ne4+ and Mg4+) and two lines from Fe5+ are observed. An explanation for the change among the parity forbidden lines can be a temperature increase in the region emitting those lines. All of the 22 emission lines observed by IUE with km s-1 before 1986 (the white-dwarf wind lines) partially or completely disappeared after 1986 and were replaced by narrow nebular lines. The transformation of the broad wind lines occurred first for the lines of lower ionization energies and later for lines of higher ionization energies. A continuous temperature increase in the white-dwarf wind during the 1980s would explain the order in which the white-dwarf wind lines disappeared, but the temperature of the white dwarf was according to Zanstra determinations constant during the same period (Altamore & Cassatella 1997). However, a temperature increase of the wind does not mean that the temperature of the white dwarf surface increased. Eriksson et al. (2004) showed that the opacity in the white-dwarf wind for two of the wind emission lines decreased, which means that more radiation from the white dwarf reaches further into the wind. A shell in the white-dwarf wind emitting, for example, C3+ emission moves outward in the white-dwarf wind until all C3+ ions in the wind are ionized to C4+.
A few of the lines which possessed wind profiles in the spectra before 1986 are shown to be pumping Fe II channels and are thereby the cause of many of the observed Fe II fluorescence lines. When the broad wind profiles disappeared in the 1980s, only the Fe II fluorescence lines corresponding to the closest coincidances remained while the other Fe II fluorescence lines vaniched.
Observations by reveal that the O VI 1031, 1037 doublet still showed a broad wind profile in 2001. The evolution of the fluorescence lines followed that of their pumping lines. As the broad wind lines vanished so did the fluorescence lines associated with channels too far apart in wavelength from the replacing nebular lines while the fluorescence lines with very close wavelength coincidences remained in the spectrum. Fe II fluorescence lines pumped by H Ly became stronger during the 1980s, indicating growth of the H II region in the system.
Many of the emission lines in spectra of AG Peg have been and surely will continue to be used for various diagnostics. When establishing diagnostics, such as for determining temperatures or densities, it is crucial to understand the nature of the lines employed.
White-dwarf winds in symbiotic stars seem to be quite rare, since broad wind profiles only could be detected in five of the 20 symbiotic systems. In the beginning we suspected that such a wind was linked to the slow nova eruption that has occurred in some of these systems. This idea must be re-considered since only two, PU Vul and AG Peg, of the five systems having broad wind profiles belong to the subclass of symbiotic novae. For seven of the other 15 symbiotic stars the FWHM of the lines show wind profiles in AG Peg are between 110-140 km s-1, while the FWHM of the same lines in the other eight symbiotic stars were 40-70 km s-1.
Except for the total lack of Fe II emission lines in the IUE spectrum of HM Sge, Fe II fluorescence lines pumped by both H Ly and C IV 1548 were observed in the symbiotic novae. Even if Fe II fluorescence lines were observed in most symbiotic stars only four of the 14 symbiotic systems with no slow nova eruption known had Fe II pumped by both H Ly and C IV. Three systems in our selection have suggested accretion disks around the hot component. The spectra of these three systems showed no signs of Fe II fluorescence. During slow nova eruptions the cool H I region and a region hot enough to create ions like C3+ simultaneously are in the line of sight with the large region with Fe+ ions, while this rarely is the case for symbiotic systems under "normal'' conditions.
All of the data presented in this paper were obtained from the Multimission Archive at the Space Telescope Science Institute (MAST). STScI is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. Support for MAST for non-HST data is provided by the NASA Office of Space Science via grant NAG-7584 and by other grants and contracts. We are grateful to the anonymous referee for a careful reading of the manuscript.
Table 13: All emission lines observed in AG Peg.
Table 14: Intensities of emission lines not detectable with LWP25995 or SWP47715.
Table 15: Emission lines saturated in LWP25995 or SWP47715.
Table 16: Aborption lines in AG Peg.