A&A 367, 949-958 (2001)
J. Bernard Salas1,2 - S. R. Pottasch2 - D. A. Beintema1 - P. R. Wesselius1
1 - SRON Laboratory for Space Research, PO Box 800, 9700 AV Groningen, The Netherlands
2 - Kapteyn Astronomical Institute, PO Box 800, 9700 AV Groningen, The Netherlands
Received 19 July 2000 / Accepted 27 November 2000
We present the infrared spectrum of the planetary nebula NGC7027 observed with the Short Wavelength Spectrometer (SWS), on board the Infrared Space Observatory (ISO). These data allow us to derive the electron density and, together with the IUE and optical spectra, the electron temperature for several ions. The nebular composition has been determined, the evolutionary status of the central star is discussed and the element depletion in the nebula is given. We conclude that the progenitor was a C-rich star with a mass between 3 and 4 .
Key words: ISM: abundances - planetary nebulae: individual: NGC7027 - infrared: ISM: lines and bands
NGC7027 is a well known planetary nebula. It is very young, dense and has an extremely rich and highly ionized spectrum due to its hot central star. The nebula has some features that makes it very interesting. The rich spectrum has been used several times to give insight about the evolution and original status of its central star. For this purpose accurate abundances are needed, and the distance, in order to know the intrinsic luminosity, has to be accurately measured. Abundance studies have been made before, but using the near-infrared SWS-ISO data the temperature uncertainty can be avoided as well as other problems, therefore achieving an accurate measurement of the nebular abundances. These ISO advantages are presented by Pottasch & Beintema (1999) and we summarize them here:
The data were obtained in the ISO guaranteed time in spring 1997. The observations were made with the Short Wavelength Spectrometer (de Graauw et al. 1996) covering the range 2.38 to 45.2 m, using the SWS01 observing template, which provides a spectral resolution from 1000 to 2500. From all NGC7027's observations, the one used in this paper (TDT number 55800537) was chosen because of its highest signal to noise ratio.
NGC7027's infrared size is sufficiently smaller than the different sizes of the apertures used with SWS to offer no problem. A VLA radio image of NGC7027 at 14.7 GHz obtained by Roelfsema et al. (1991) clearly stays within the ISO-SWS small aperture . The other SWS apertures are larger, , and . They are all centered at almost the same position within an uncertainty of 3 . Persi et al. (1999) studied the infrared size of NGC7027 using spectral images taken with the ISOCAM-CVF, showing that it has an ellipsoidal morphology of about . This is again within the smallest ISO-SWS aperture. Persi et al. (1999) measured the spectra over an area of 25 square arcsec that with no doubt contains the whole nebula. A comparison of present fluxes with those found by Persi et al. (1999) is shown in Table 1 and demonstrate that no flux is missing in the SWS observation.
|(m)||observ.||et al. (1999)|
In present observations the Br (measured with the smallest SWS aperture), [Mg V] and [Ar III] are stronger than those of Persi et al. (1999), indicating that no flux is being lost in the present observations. Strong differences in both measurement are probably due to the poor spectral resolution of the ISOCAM-CVF in comparison with the SWS, and therefore the latter are more accurate. The extremly high flux of [Ne VI] measured by Persi et al. (1999) is wrong because it is blended with a PAH feature. No contributions from the outer parts of the nebula are expected since the gas is cold and therefore the whole flux is seen in each aperture. For all these reasons no aperture correction for the infrared lines is needed.
The IUE observations were taken at almost the same position as our observations. The aperture of the IUE is an ellipse of very similar to our smallest aperture. It therefore contains the whole nebula. The optical lines were observed with a very small aperture on the bright part of NGC7027 and represent a part of the nebula. We assume that they are representative of the entire nebula. Using the entire H flux, they are scaled to the total flux. We avoid here making any aperture corrections because these are expected to be small if they exist at all.
The data were reduced using the available tools in the inter-active analysis software package distributed by the SWS consortium. Our reduction differs from the standard pipeline in the removing of fringing in the data, and a better flatfielding. Any step in the reduction analysis which could affect the spectral resolution or the linearity of the data was avoid. Line fluxes were measured fitting Gaussians to the observed line profiles using the routine multi_fit. It allows removal of glitches or bad data interactively in each line. This was rather time consuming, but the weakest lines were found and the highest possible resolution kept at each wavelength.
The reduction didn't present any problem except in the first part of band 2c (7 to 12 m) where the up and down scans deviated significantly from each other, but matched in the rest of the band when flatfielding was applied. This doesn't affect the measured fluxes but could be important when reasonable upper limits to some lines are determined, as explained in next paragraph. This was caused by some detectors (19, 20, 21, 23 and 24) which suffered from erroneous dark current subtraction. The range 7.3 to 7.6 m was affected. These detectors were removed. A reference flat (order one) was defined using the remaining detectors and applied to the whole band. Flatfielding reduces the mismatch and then all detectors could be used to measured line intensities. It is useful to compute an upper limit to the flux for some unseen lines. For this purpose, outlying data points were removed with the sigclip tool and the spectra were rebinned (sws_rebin tool) to a standard resolution. Upper limits for the lines [Cl IV] 11.76 m, [Na VI] 8.61 m, [K VI] 5.58 and 8.83 m were calculated using a model of a synthetic Gaussian, with the width corresponding to the resolution, and a height of three times the rms (deviation from the average). These upper limits of unseen lines allow to more accurately apply the ionization correction factor for the relevant ions.
In Table 2 the fluxes measured for the
infrared (IR) lines are listed. In total around 80
lines were measured. Calibration errors are larger at
longer wavelengths but on average
amount to about 20%. Random measurement errors depend on the flux
strength: weak lines (I< 5, where I is in 10-12 ergcm-2s-1)
have estimated errors of about 20%, intermediate lines (5<I<10) of
about 10%, while the strongest lines (I>10), have normally less than 5%
uncertainty. All the lines follow this behavior with the exception of
the S III line at 33.5 m with a measured intensity of
7.12 10-12 ergcm-2s-1and an uncertainty of 19%. This part of the spectrum is the noisiest and weak
lines are not seen. Therefore, this line could be considered an intermediate
strength line in this part of the spectrum. Thus the uncertainty in the fluxes of
intermediate and strong lines is dominated
by the calibration error.
|Figure 1: SWS spectrum of NGC7027. As can be seen strong lines of high ionization stages of some ions are seen. Hydrogen lines are abundant. Also noticiable the strong continuum at longer wavelength and PAHs at 6.6, 7.7 and 8.8 m|
|Open with DEXTER|
NGC7072's spectrum is extremly rich as can be seen in
Fig. 1. The hydrogen and helium lines are abundant in the whole spectrum but
specially in the range 2.4-5.5 m. Ions with high stages of
ionization are found (i.e. Fe VII, Na VII, Ar VI)
in the nebula. Nine molecular hydrogen lines (H2) have been measured as well and
probably originate in the outer part of the nebula where the
temperature is lowest. For Ar, K and specially Ne
we have information for the most important stages of ionization
implying that the uncertainty in the ionization correction factor (ICF)
used to derive the absolute abundances will be much reduced.
Several stages of ionization of some ions were not observe. Then,
an upper limit to the flux was determined, as described in the
previous section. Also noticiable in the spectrum is the strong
continuum at the longest wavelength and the features at 3.3, 6.6, 7.7,
8.8 and 11.2 m now known as PAHs as well as a broad PAH plateau
from 15 to 20 m. In this paper we will concern ourselves only
with the emission lines. A detail study of the PAHs and the plateau can be found in the articles by Peeters et al. (2001), Hony et al. (2001) and Van Kerckhoven et al. (2000).
|(m)||(10 )||(m)||(10 )|
|2.406||H2, 1 0 Q(1)||3.73||5.580||[K VI]||<8.28|
|2.424||H2, 1 0 Q(3)||2.35||5.908||H I 6-9||6.53|
|2.495||H I 5-17||1.17||5.981||[K IV]||3.60|
|2.564||H I 5-15||2.52||6.705||[Cl V]||6.09|
|2.613||He I 4-6||1.46||6.910||H2, 0 0 S(5)||6.69|
|2.626||H I (Br)||45.9||6.947||He II 8-9||3.47|
|2.671||He I 4-6||1.45||6.985||[Ar II]||30.7|
|2.675||H I 5-13||2.24||7.317||[Na III]||13.9|
|2.758||H I 5-12||2.84||7.459||H I (Pf)||28.3|
|2.802||H2, 1 0 O(3)||2.96||7.503||H I 6-8||8.91|
|2.826||He II 7-9||5.43||7.652||[Ne VI]||96.5|
|2.905||[Al V]||1.16||7.814||[Fe VII]||2.63|
|3.004||H2, 1 0 O(4)||0.59||7.901||[Ar V]||47.1|
|3.039||H I 5-10||5.70||8.025||H2, 0 0 S(4)||2.52|
|3.092||He II 6-7||18.8||8.610||[Na VI]||<3.45|
|3.095||He II 8-11||1.66||8.761||H I 7-10||2.53|
|3.207||[Ca IV]||1.47||8.830||[K VI]||<4.17|
|3.234||H2, 1 0 O(5)||0.79||8.991||[Ar III]||122|
|3.625||H I 6-20, 6-19||0.95||9.042||[Na IV]||12.1|
|3.740||H I (Pf)||11.0||9.525||[Fe VII]||4.85|
|3.749||H I 6-17||0.78||9.665||H2, 0 0 S(3)||6.82|
|3.907||H I 6-15||1.29||9.713||He II 9-10||2.16|
|4.020||He I 4-5||1.08||10.510||[S IV]||429|
|4.022||H I 6-14||0.72||11.760||[Cl IV]||<4.16|
|4.038||He I 4-5||0.97||12.372||H I 6-7||11.3|
|4.052||H I (Br)||91.6||12.812||[Ne II]||91.6|
|4.296||He I 3-5||4.86||13.101||[Ar V]||31.0|
|4.377||H I 6-12||1.66||13.520||[Mg V]||38.7|
|4.486||[Mg IV]||148||14.321||[Ne V]||1350|
|4.529||[Ar VI]||126||15.550||[Ne III]||1090|
|4.618||[K III]||4.09||18.713||[S III]||76.0|
|4.654||H I (Pf)||17.5||19.062||H I 7-8||5.94|
|4.672||H I 6-11||2.50||21.800||[Ar III]||3.00|
|4.764||He II 7-8||9.43||24.315||[Ne V]||387|
|5.128||H I 6-10||4.51||25.888||[O IV]||715|
|5.510||H2, 0 0 S(7)||2.29||33.477||[S III]||7.12|
|5.610||[Mg V]||416||34.812||[Si II]||14.9|
In addition to the large number of infrared lines measured, also some optical and ultraviolet lines were used taken from the paper of Keyes & Aller (1990). The information on such important ions as C, N, O is mainly derived from optical and ultraviolet data.
The lines used are listed in Table 3.
To correct optical fluxes for interstellar extinction a value of
was used, which leads to an extinction of
EB-V=0.85. To calculate this the H
flux was computed
in two ways. Using the equation given by Pottasch (1984):
|Br||5 4||4.052||6.77 10-2||91.6||1.35|
|Br||6 4||2.626||4.00 10-2||45.9||1.15|
|Pf||6 5||7.459||2.09 10-2||28.3||1.35|
|Pf||7 5||4.654||1.37 10-2||17.5||1.28|
|Pf||8 5||3.740||9.20 10-3||11.0||1.20|
|1 Units in .|
|2 Units in .|
The average of these measurements leads to . The Br line observed (using no reddening correction) is the strongest and therefore has been most accurately measured. The value of deduced from this line agrees with the one derive from Eq. (1) and was finally adopted. It is worth mention here that the theoretical predictions from Hummer & Storey (1987) are given just for or 105 cm-3. Here a of 104 cm-3 is used but an cm-3 would lead to the same conclusions ( ). The optical lines were corrected for reddening using the extinction law of Fluks et al. (1994) and an extinction of EB-V=0.85. A recent publication of Wolff et al. (2000) derives an extinction of EB-V=1.10 for NGC7027 by directly measuring the central star. This difference is difficult to explain and would considerably change the values of the dereddened fluxes. One possibility is that NGC7027 shows variations in the extinction across the nebula (Wolff et al. 2000). This can be higher towards the central star region than in the nebula itself. Wolff et al. (2000) give a summary of extinction determinations by several authors: Walton et al. (1988) find an extinction EB-V=0.8 for the central star and for the rest of the nebula, Kaler & Lutz (1985) found 0.85 and Shaw & Kaler (1982) 0.94.
According to Keyes et al. (1990) the UV lines have errors less than 20%, while the optical lines have errors below 50%. Here a 15% error is assumed for the UV, and a 40% error for the optical lines. Also the N III infrared line at 57.3 m from the Long Wavelength Spectrometer (LWS, 43 to 196.7 m) has been used. It was taken from Liu et al. (1996) and was used to compare the abundance with that found using the ultraviolet line at 1750 Å.
Using the entire data set (infrared, optical and ultraviolet) the physical parameters (electron density and temperature) have been determined as a function of the ionization potential.
An electron temperature of 15500 K has been chosen to compute the density. This choice is justified in the next subsection. To determine the electron density () reliably, the ratio of lines close in energy level is needed so that the temperature hardly plays a role in populating these levels. Two ions, Ne III and Ne V fulfil this condition. In addition the for C III was calculated using ultraviolet lines (Keyes & Aller 1990). These measurements cover the ionization potential (IP) range from 23 to 97eV. In order to have information on the for ions with low IP the values given by Keenan et al. (1999) for O II, and Keyes et al. (1990) for S II, Cl III were used. All derived values are shown in Table 5.
|Ion||Ion Pot||Lines used||Observed|
|1 Values taken from Keyes & Aller (1990).|
|2 Value taken from Keenan et al. (1999).|
|1 ICF = Ionization Correction Factor.|
|a Using the predictions for the helium emission lines given by Benjamin et al. (1999).|
Table 5 does not show a trend of with IP in the range 10 to 23eV, but from 23 to 97eV a trend appears when the value of K VI of Keyes et al. (1990) is used. In this range seems to drop from a value of 75000 to 25000 cm-3. To compute abundances for each ion a is needed. Because many infrared lines are used, the is not critical, and the relative abundance of the ions do usually not strongly depends on it. was derived from the curve in Fig. 2. Most of the ions are not dependent (only three ions show dependence, Ar V, O IV and Ne V). For this reason an average cm-3 was used to compute the electron temperature (following subsection).
|Figure 2: Electron density versus ionization potential. The ratios used for each ion are indicated. Asterisks are our measurements while the triangle show value derived by Keenan and the squares gives the values by Keyes. The shape of the assumed curve is in accordance with the expected physical behaviour (see discussion Sect.7.2)|
|Open with DEXTER|
To compute the electron temperature () line ratios are needed originating from energy levels differing by several electron volts. Using optical data only two ions (N II and O III) obey this condition while using infrared and ultraviolet data, three others (O IV, Ne V and Mg V) become available as well. As can be seen from Fig. 3 these provide information on almost the whole IP range of interest. In Fig. 3 we present as a function of IP. The density used is 50000 cm-3 (see previous section). There exists a correlation between and IP, such that increases until about 70 eV and then remains at a constant value of 15500 K.
Some ions need a further discussion. O IV gives a higher temperature and this may be because the ultraviolet line at 140.0 nm is blended with Si IV. Only NeV rather strongly depends on . For this ion the average density, cm-3is used. Using cm-3 leads to a much higher of 22400 K. The Ne III ratio, 386.8 nm over 15.5 m can, in principle, be used to derived . As noticed before by Pottasch & Beintema (1999), it always gives a low temperature. In this study has been found using this ratio. This has been found in many nebulae, i.e. NGC6302, NGC6537. It is not clear why, but it may have something to do with the uncertainty in the atomic parameters and the 386.8 nm line has not been used in this paper.
|Figure 3: Electron temperature versus ionization potential. The lines used for each ion are indicated. There is a correlation of with IP, represented by the curve|
|Open with DEXTER|
In order to be able to compute abundances, values of
for each ion. They are derived from the relations
presented in Figs. 2 and 3.
are known the
relative ionic abundances can be determined. Then, applying an ICF the element
abundances are derived. Ionic abundances are normalized to
hydrogen. The equation used to determine
the ionic abundances is:
All abundance values are shown in Table 6. The ICF is unity in four cases, because all important ionization states have been measured. In most cases it is between 1.0 and 1.25, and only in three cases is it above 1.25 (S, Mg and K with ICF's of 1.27, 2 and 1.35 respectively). In Keyes & Aller (1990), the ICF is normally between 1.2 and 5, and only in a few cases is under 1.2. This shows that the present abundance determination is an important improvement. The ICF were calculated on a purely empirical basis. For a given element the ionization potential where the maximum ionic abundance is reached was taken into account. For some ions like potassium, no information on the high sixth stage of ionization was found, thus an upper limit on K5+ was derived. As can be seen from Table 6, the contribution of the K5+ ion should not exceed 1.2(-8). The contribution of this to the element abundance was constrained. For Calcium and Iron the abundance has not been determined because important contributions from other stages of ionization are expected.
Si+ gives almost the same ionic abundance as the Si2+ ion. This is suspicious because of the low electron temperature needed to excite this ion, and it probably originates mainly from the photodissociation region outside the nebula since the Si+ ion is ionized at a low potential (8.1 eV).
Comparison with the nitrogen ions that have similar ionization potentials shows that N2+ gives a major contribution, while that of N+ is not negligible but lower than other ions. Silicon should follow the same behavior, and the Si+ contribution is probably coming from the photodissociation region around NGC7027 instead of from the nebula itself. These considerations lead to lower abundances of C and Si (see Table 7).
|Element||Present||Sun1||O, B Stars2||Keyes & Aller||Middlemass|
A comparison of these abundances with those found in the Sun, O-B stars, and those of Keyes & Aller (1990) and Middlemass (1990) is shown in Table 7. The helium abundance agrees with the earlier results of Keyes & Aller (1990) and Middlemass (1990). Carbon is abundant, with a value similar to that found by Keyes & Aller (1990) but lower than found by Middlemass (1990). The model of Middlemass predicts a high abundance of all carbon ions, including C II. The abundance of oxygen is lower than in the Sun. This has also been found for many PNe in the past (Aller & Hyung 1995) and is in reasonable agreement with Keyes & Aller and Middlemass. The nitrogen, neon, magnesium, silicon and argon abundances are in agreement with the other authors. The magnesium abundance is lower than the one found in the Sun but similar to that of the O, B stars, supporting the idea that this element is depleted onto dust. The same is true for silicon, its abundance is even lower than found in O, B stars. This element is a well known component of the dust and the missing silicon is probably in this form.
In Table 8 abundance ratios are compared with the Sun and O, B stars. The ratio of N/O is almost twice as high as that found in the Sun and O, B stars, while the C/O ratio is three times high. This means that nitrogen and carbon are over-abundant in the nebula and it is likely that some oxygen has been converted into nitrogen in the course of its evolution by the CNO cycle. This is supported by the fact that the (C+N+O)/H ratio is very similar to solar indicating a similar mass of the progenitor star of the nebula. The solar O is over abundant and could explain the difference in ratios.
The derived parameters, electron density and temperature, are shown in Table 5, Figs. 2 and 3 respectively. The ionic and element abundances of NGC7027 are given in Table 6.
Ionization correction factors were needed in the past
to obtain the abundance of an element, when only one or a few stages of
ionization were observed. These were used to estimate the
contribution of these unobserved stages of ionization. These estimates
were made using models, e.g. similarities on the ionization potential of some
ions. Now, since the most important stages of
ionization of several ions have been measured it can be checked how
good the ionization correction factors,
based on similarities in the ionization potential, are. There are two
similarities introduced by Torres-Peimbert & Peimbert (1997). One is:
and as a function of IP, that is, as a function of the distance to the central star have been obtained. The behavior of as a function of IP is consistent with a simple picture of NGC7027. This simple picture relates IP and distance to the central star: high ionization stages are reached close to the star and low stages occur in the outer parts. The O3+ 140.0 nm line gives a too high temperature. This line was taken from Keyes & Aller (1990) and they claim that it is blended with Si3+. Decrease of this flux leads to a lower ratio, decreasing the temperature. In our opinion it is uncertain that the O3+ line is blended with the Si3+ line because it is a stellar line and should be formed close to the star. As the central star of NGC7027 is very weak in the ultraviolet because of dust absorption, the Si3+ line will also be weak. The relation indicated in Fig. 3 is consistent with theory.
No trend of electron density with ions of low IP has been found, although it seems that there is a correlation of the density with ions at high ionization stages (high IP), where it seems to drop. This could mean that the ejection rate of the nebula in its latest stages was low because the highest ionization stages are reached close to the nebula and they give lowest density. This tells us that the mass presently being ejected is considerably lower than at an earlier stage.
A new problem has emerged in the study of nebular abundances. For some planetary nebulae it has been found that ionic abundances deduced from recombination lines are higher than those derived from collisionally excited lines. In a recent publication by Liu et al. (2000) a deep study in this sense is carried out for the planetary nebula NGC6153. They have found that the abundance derived form optical recombination lines are a factor of 10 higher than those derive from optical, UV or infrared collisionally excited lines. This difference is significant. Liu et al. (2000) discuss density inhomogeneities and temperature fluctuations as the possible cause of such discrepancy, concluding that the latter are unimportant. Using the recombination lines measured by Keyes & Aller (1990) for NGC7027 we have derived the abundances shown in Table 9.
1 Col = Collisional excitated line, Rec = Recombination line.
2 Dereddened line intensities relative to .
These abundances have been derived by scaling those found by Liu et al. (2000). Although these lines give a somewhat higher abundance in the three cases compared, we do not regard this as significant and prefer to use the abundances found from the collisional lines. The great strength of the recombination lines in NGC6153 remains a puzzle.
By comparing the abundances obtained with the semi-analytical model of Marigo et al. (1996), the mass of NCG7027's progenitor star can be constrained. With the derived abundances of helium, carbon, nitrogen and oxygen (Table 7) we can compare NGC7027 to the model. The model predicts the chemical composition of planetary nebulae for two metallicities, Z=0.02 and Z=0.008. In the literature no clue to the metallicity of NGC7027 is found. Then the central star has to be measured and this is not trivial. Because of this, the problem can be inverted. A clue in the metallicity of the central star can be given by comparing the abundances derived with the different models of high or low metalicity and checking which one fits better. For this reason, the metallicity that better reproduces the NGC7027 abundances was used (Z=0.008).
|Figure 4: The logarithm of N/O ratio versus He/H. Abundance ratios by number, expected in PN's taken from the semi-analytical models of Marigo et al. 1996 (metallicity of Z=0.008) are shown. The numbers in the figure indicate the mass of the progenitor star in solar units. The position of NCG7027 is plotted|
|Open with DEXTER|
As can be seen in Fig. 4 the model with a metallicity of Z=0.008 constrains the mass of the star to be between 3-4 . It is worth noticing that the model overestimates the prediction of the oxygen abundance with respect to the observational values. If the ratio C/O versus He/H or C/O versus N/O (where the overabundance of oxygen in the model is canceled) is considered, this leads to the conclusion that more carbon abundance is needed to fit the model. More studies concerning the carbon abundance and improvements in these evolutionary models are therefore needed.
A reliable infrared spectrum has been obtained (Table 2). This has been complemented with optical and ultraviolet lines in order to determine the physical parameters in the nebula, as well as to determine the abundances much more accurate than before. A correlation between electron temperature and IP has been found. Ions with high IP, that are formed close to the nebula, give high temperatures (around 15000 K). Those with low IP, formed in the outer (coolest) parts give low temperature (see Fig. 3). No clear correlation of electron density with IP is found, although a tendency of density to decrease at high IP ion seems to be presented (Table 5 and Fig. 2). The relative and element abundances for 18 ions have been determined and are presented in Table 6. By comparing these abundances with the semi-analytical model of Marigo et al. (1996) at a metallicity Z=0.008 we have constrained the mass of the progenitor star to be between 3-4 solar masses.
We are grateful to Peter van Hoof for his help with the atomic parameter data (especially the effective collision strength) used in our calculations