A&A 443, 509-517 (2005)
DOI: 10.1051/0004-6361:20053448
V. Lebouteiller1 - Kuassivi2 - R. Ferlet1
1 - Institut d'Astrophysique de Paris,
UMR7095 CNRS, Université Pierre & Marie Curie, 98 bis boulevard Arago, 75014 Paris, France
2 -
AZimov association, 14 rue Roger Moutte, 83270 St-Cyr sur Mer, France
Received 17 May 2005 / Accepted 18 July 2005
Abstract
We present FUSE and HST/STIS measurements of the P II column density toward Galactic stars. We analyzed P II through the profile fitting of
the unsaturated 1125 and
1533 lines and derived column densities integrated along the
sightlines as well as in individual resolved components. We find that phosphorus is
not depleted along those sightlines sampling the diffuse neutral
gas. We also investigate the correlation existing between P II and O I column densities
and find that there is no differential depletion between these two specie.
Furthermore, the ratio N(P II)/N(O I) is consistent with the solar P/O value, implying that P II and O I coexist in the same gaseous phase and are likely to evolve in parallel.
We argue that phosphorus, as traced by
P II, is an excellent neutral oxygen tracer in various physical environments, except when ionization
corrections are a significant issue. Thus, P II lines (observable with FUSE, HST/STIS,
or with VLT/UVES for the
QSO sightlines) are particularly useful as a proxy for O I lines when these are saturated or blended.
Key words: ISM: abundances - ISM: atoms - ISM: clouds - Galaxy: abundances - ultraviolet: ISM
Phosphorus (
3131P) is an odd-Z element which is thought to be mainly produced in
the same massive stars that form -elements (O, Ne, Mg, Si, S, Ar, ...). Although
the nucleosynthesis of odd-Z elements is still not well understood, phosphorus seems to
be produced during the carbon and neon burning in a hydrostatic shell (Arnett 1996).
Woosley & Weaver (1995) found that no significant amount is expected to be synthesized
during the explosion phases and that the P yields should be metallicity dependent because
of the odd-even effect.
In the Galactic and extra-galactic diffuse interstellar medium (ISM), the phosphorus gaseous phase abundance is largely unknown. Apart from the early investigations with the Copernicus satellite within the solar neighborhood (Jenkins et al. 1986; Dufton et al. 1986), a few measurements are available in the distant ISM and almost none in the extragalactic ISM. Nevertheless, the ISM phosphorus abundance provides an important constraint for the dust/gas chemistry.
Dust grains are subject to dramatic chemical, morphological and structural changes during their lifetime, from the condensation of dusty cores within the outflows of evolved stars to the final destruction in shocks or the formation of new generations of stars and planetary systems. They undergo several growth (mantle accretion and molecules adsorption) and erosion (photoprocessing, sputtering) periods as they move within dense or diffuse media (Snow & Meyers 1979; Seab 1987; Turner 1991; O'Donnell & Mathis 1997). In that respect, refractory elements are expected to be the most depleted in the gas phase. On the contrary, C, N and O are less depleted onto dust grains because of their lower condensation temperatures (Field 1974; Lodders 2003). Hence, at most one third of the oxygen is bound to rocky elements (Cardelli et al. 1997). With respect to its relatively high condensation temperature, phosphorus can be expected a priori to be more heavily depleted.
Table 1: Stellar and sightline properties. Distances, reddening and spectral types are from Diplas & Savage (1994) unless otherwise noted. The typical uncertainty in the distance is 30%. The molecular hydrogen fraction along the sightline toward HD 24534 is from Rachford et al. (2002), the other sightlines have been measured by A2003.
Jenkins et al. (1986), in their ISM survey conducted with Copernicus toward
about 80 stars, found through the analysis of the far-UV absorption lines of the dominant
ionization state P II that phosphorus is not depleted along sightlines containing predominantly
warm low density neutral gas (0.1 dex after updating the solar abundances from
Asplund et al. 2004) and is depleted by
0.5 dex in cooler and denser clouds.
Using the same dataset but with a different oscillator strength for the P II
1302 transition, Dufton et al. (1986) derived phosphorus abundances
systematically larger by
0.2 dex so that the previous findings still hold. In
the cold diffuse interstellar cloud toward
Oph, the depletion of phosphorus is
dex as compared with
dex for oxygen, while in the warm diffuse
cloud along the same sightline, P is depleted by
dex as compared with
dex for O (Savage & Sembach 1996).
The first detection of the PN molecule by Turner & Bally (1987) through the 140 GHz (J=3-2) and 234 GHz (J=5-4) emission lines paved the way for extensive studies of the abundance of phosphorus in molecular clouds. It was soon recognized that the transfer of phosphorus from the gas to the solid and back to the gas phase largely involved carbon atoms via the HCP linear molecule (Turner et al. 1990). As a consequence, phosphorus is believed to mainly reside in adsorbed HCP molecules which are then released in the gas by photodesorption in warm media and readily photodissociated. This scenario would account for the lack of depletion in the warm phase.
Since the advent of the Far Ultraviolet Spectroscopic Explorer (FUSE) satellite
(Moos et al. 2000) and the STIS instrument onboard HST, it is now possible to
probe denser clouds and investigate longer sightlines. We thus revisit and extend
previous works on interstellar phosphorus abundance by presenting new P II measurements
toward Galactic sightlines obtained from FUSE and HST/STIS data.
Furthermore, in order to point out the relative behavior of phosphorus as compared to
-elements and possibly reveal a global trend toward the differential depletion
under many different physical conditions, we compare the phosphorus gaseous phase
abundance with that of oxygen. Oxygen abundance in the diffuse ISM has been extensively
studied (see e.g., Jensen et al. 2005; Cartledge et al. 2001, 2004; and
André et al. 2003 - hereafter A2003) and is used here as a reference element. The
observations and data analysis are described in Sects. 2 and 3. We present the results
in Sect. 4. Final conclusions are given in Sect. 5.
All the FUSE spectra were obtained through the large
(LWRS) aperture which results in a resolving power
(or
km s-1, FWHM). This
spectral resolution depends on the co-addition procedure used to reconstruct the total
exposure and varies with the wavelength and the detector. Hence, we did not attempt to
co-add different detectors, in order to minimize both the distortion of the resulting
Point Spread Function (PSF) and the propagation of the Fixed Pattern Noise (FPN) proper
to each detector. The detailed reduction, calibration and co-addition procedures can be
found in A2003 who reduced most of the present data in order to study the neutral oxygen
and hydrogen content along the sightlines.
Table 2: Summary of FUSE observations.
The STIS observations were taken with the far-ultraviolet MAMA detector equipped
with the E140H grating. However, three different apertures were used. The
aperture provides a resolving power
(or
km s-1, FWHM). The two others,
and
,
provide a spectral resolution of
(velocity
resolution of
km s-1, FWHM). Again, details about the data
reduction can be found in A2003.
Table 3: Log of HST/STIS observations. All the data have been obtained with the E140H grating.
Among the 10 targets we present, 8 were analyzed by A2003 with a particular concern for
the O I column density. The other targets of their sample were not observed with
STIS at the wavelength of the P II 1533 line (see next section). We
also analyzed STIS spectra of two additional targets, HD 24534 and HD 121868.
Finally, we also compiled published P II and O I measurements in the Milky Way and along a few extragalactic sightlines. These are listed in Table 6 and discussed in Sect. 4.3.
![]() |
Figure 1:
P II ![]() ![]() |
Open with DEXTER |
Because of the ionization potential of O I (13.62 eV as compared with 13.60 eV for H I) and the efficient charge exchange between O II and H I, O I is expected to be the dominant ionization state of oxygen in the diffuse neutral gas, and thus a good tracer of the neutral gas. Observations of A2003 confirm this finding. On the other hand, the ionization potentials of P I and P II (resp. 10.49 eV and 19.72 eV) suggest a priori that P II should be the dominant state of phosphorus in this gaseous phase. However a fraction of P II atoms could actually reside in a potential ionized gaseous phase where oxygen is in the form of O II and hydrogen in the form of H II. The fraction is unknown and depends on the ionizing radiation illuminating the diffuse clouds. The present study will help to identify this possible correction.
Absorption lines were analyzed assuming Voigt profiles, by using the profile fitting
program Owens. This Fortran code, developed by M. Lemoine and the FUSE
French team, is particularly suitable for simultaneous fits of far-UV spectra (Lemoine et al. 2002). A great advantage of this routine is the ability to fit different spectral
domains and various species in a single run. It was then possible to analyze
simultaneously the P II and O I absorption lines, together with Cl I, C I, and S I lines. The use of these species in a simultaneous fit allowed us to check and constrain
the radial velocity structure of the sightlines when uncertain. The P II and O I lines
we analyzed being unsaturated, the column density determination does not depend on the
b-parameter. The errors on the column densities are calculated using the
method described in Hébrard et al. (2002) and include the uncertainties on
all the free parameters such as the continuum shape and position. All the errors we
report are within 1
.
Numerous O I lines are observed in the HST/STIS + FUSE spectral
ranges, with oscillator strengths (f) spanning several orders of magnitude. However,
the main constraint on the O I column density lies in using the weak 1355.5977 Å intersystem transition (
). The total O I column densities toward
most of the present sightlines were derived by A2003 using this line. However, we have
updated these values by deriving column densities of each cloud along the sightlines and
by performing a simultaneous analysis of O I and P II lines.
A total of seven P II lines are available in the far-UV spectral domain. Thanks to the combination of datasets, the phosphorus atom content is readily explored through transitions spanning more than 2 orders of magnitude in oscillator strengths, always allowing the choice of adequate lines for a particular study. Wavelengths and f-values are from the revised compilation of atomic data by Morton (1991, 2003). Oscillator strengths of P II lines can be relatively uncertain. Indeed, significant differences are found in the few laboratory experiments that investigated P II atomic data (see the discussion of these values in Morton 2003).
In the present study, the three P II lines at 961.0412 Å (
),
963.8005 Å (
)
and 972.7791 Å (
),
observable with FUSE, are located in a region overcrowded with strong absorption
lines - mainly H I lines from the Lyman serie and H2 lines. We thus rejected these
transitions because they are blended. In addition, the two first lines are heavily
saturated, thus preventing reliable P II column determinations.
Similarly, we avoided the P II lines at 1152.8180 Å (
,
observable with FUSE), and at 1301.8743 Å (
,
observable with STIS) since they both present saturation effects most of the
time. The latter is most often blended with the extremely strong O I line at
1302.1685 Å, even at the highest STIS resolution.
The P II line at 1124.9452 Å (
,
observable with FUSE)
is the weakest transition available and is always found to lie on the linear part of the
curve of growth. When detected, it can provide an accurate measurement of the total
P II column density along a given sightline but with little information on the
velocity structure. Analyzing this line requires high-quality data and a good knowledge
of FPN for the FUSE detectors. Indeed, duplicate detectors make it easier to
discriminate FPN and absorption features in most cases. Unfortunately, in some cases this
line appears slightly blended with the broad Fe III*
1124 stellar line.
Slightly stronger the P II line at 1532.5330 Å is (
,
observable with STIS). Thanks to the STIS higher resolution, this line
allows the investigation of the detailed velocity distribution of sightlines and the
derivation of P II column densities in individual clouds. We thus used this line to
derive the P II column densities (column density of each individual cloud and integrated
column density over the sightline) and compare with the O I values. The other main
advantage of this line is that its optical depth is systematically found to be similar to
the optical depth of the O I
1356 line (see Fig. 1). This is
due to the combination of the O I column density, about 2000 times larger than the P II one, and the oscillator strength, 2000 times lower than for the P II
1533 line.
Hence, the simultaneous analysis of these two lines minimizes possible systematic errors
due to saturation and/or unresolved components. This combination further allows one to
investigate individual cloud column densities since these two lines are observed in the
same high-resolution dataset.
![]() |
Figure 2:
The P II ![]() ![]() |
Open with DEXTER |
Table 4:
Column densities integrated over the sightlines. H I and H2 measurements are from A2003.
The FUSE determination of the P II column density makes use of the
line while the
HST/STIS determination makes use of the
line. O I is analyzed through the
transition.
Another possibility is to compare the integrated column density as derived from the
unresolved profile of the P II 1125 line with FUSE, with the sum of the
P II column densities of individual clouds along the sightlines as derived with the
STIS resolved profile of the P II
1533 line (Fig. 2).
We get consistent findings within the error bars (see Table 4). This
comparison provides a twofold confirmation: 1) the relative compatibility of the
corresponding oscillator strengths, and 2) the absence of gross systematic errors when
assuming only one global velocity component to estimate integrated column densities.
We plot in Fig. 3 the integrated P II column density (as derived with the
P II
line) versus the total hydrogen content defined as
N(H
(H I
(H2). We use the values N(H I) and
N(H2) derived by A2003 through FUSE observations, available for 8 of our 10
sightlines. Column densities are integrated along sightlines and sample the diffuse ISM.
As compared with the solar abundance
(P/H)
(Asplund et al.
2004), the regression of P II vs. H
gives a consistent value
.
We can conclude that:
![]() |
Figure 3:
The P II column density integrated over each sightline versus the total
hydrogen column density defined as
![]() |
Open with DEXTER |
![]() |
Figure 4: P II total column density over each sightline is plotted versus the total O I column density. The thick line shows the solar ratio (Asplund et al. 2004) and the thin dashed line shows the regression fit. |
Open with DEXTER |
Table 5:
Column densities of the individual clouds along each sightline. No isolated component is
found in the O I 1356 and P II
1533 lines for HD 224151.
![]() |
Figure 5: P II column density of each individual component along the sightlines plotted against the corresponding O I column density. The thick line shows the solar ratio (Asplund et al. 2004) and the thin dashed line shows the regression fit. |
Open with DEXTER |
The integrated P II and O I column densities toward the 10 Galactic sightlines of our
sample were derived using the P II
and O I
lines, observable
with STIS. Values are listed in Table 4 and plotted in
Fig. 4. We find integrated O I column densities consistent within
1
error bars with those found by A2003 except toward a few sightlines for which
error bars of A2003 could have been somewhat underestimated. There is also some overlap
with past measurements (see Howk et al. 2000; Jensen et al. 2005; and Knauth et al.
2003). These past values are marginally consistent with our values within 1
error bars, except the O I column density toward HD 218915 derived by Howk et al. (2000).
However, this Howk et al. determination is also inconsistent with the other studies of
A2003 and Knauth et al. (2003).
![]() |
Figure 6:
Compiled measurements of P II and O I column densities found in the
literature (see Table 6) overplotted on our results. Empty circles represent
sightlines toward stars with distances less than 200 pc. Filled circles represent the
values of this study (results plotted in Fig. 5 toward stars between ![]() ![]() |
Open with DEXTER |
There is a clear correlation between P II and O I column densities. The derived
error-weighted mean P II/O I ratio is
-3.26+0.12-0.10 (1 error bars),
which is consistent with the solar P/O proportion
(Asplund et al. 2004).
The only exception is the sightline toward the star HD 121968 located in the halo. We
observe a single absorption component which could arise in a cloud located in the
low-density partly-ionized Galactic halo. Therefore, in such extreme conditions, P II could also exist where oxygen is ionized so that the actual P/O ratio would be
(P II)/N(O I), where
is an unknown factor which depends on the
ionization conditions within the cloud. This ionization effect is negligible for all the
other sightlines.
Does the same correlation hold if we plot the column densities of individual clouds along
the sightlines ? We selected individual components in each sightline on the basis of a
clear separation from nearby other absorptions - i.e. the wavelength shift due to the
different radial velocities of clouds must be larger than the intrinsic line widths
(convolution of the turbulent velocity and the instrumental line spread function). We
identified 17 interstellar absorbing regions along our sightlines (Table 5).
We still find a correlation over more than 1 dex in column density
(Fig. 5), with a mean P II/O I ratio of
-3.25+0.12-0.10, again
consistent with the solar P/O ratio. The only noticeable exception to this good
correlation (apart from the single component along the HD 121968 sightline, see above) is
the component #2 along the HD 104705 sightline. This sightline crosses an interarm region
in the Milky Way disk where the medium is particularly ionized (Sembach 1994). This
strongly suggests that the absorption component #2 lies in such a region where P II/O I is larger than the actual P/O ratio.
Given these correlations, the following conclusions can be put forward:
We gathered from the literature several Galactic and extragalactic sightlines toward which both P II and O I column densities have been measured (see Table 6). We have excluded two sightlines toward subdwarf O stars (Wood et al. 2004) since the P II column densities were determined through only one saturated line.
Table 6: Compilation of O I and P II measurements found in the literature. The distance we report is the target distance, it is also an upper limit for the cloud distances.
The sightlines are variable in terms of distances to the targets. We list twenty stars,
from the Local Bubble (within 100 pc) to the distant Galactic plane (with star distances
up to 5 kpc), one star in the Small Magellanic Cloud and four high-redshift damped Lyman
systems (DLAs). The sightlines also span various metallicities (from solar to
1/100 solar) and various color excesses. All these values are overplotted on our
results in Fig. 6. The correlation between P II and O I column densities
found in Sect. 4.2 still holds over more than three decades in column
densities.
We notice however that in three DLAS, the P II/O I column density ratio is significantly lower than the solar P/O value. This could be the sign of a metallicity dependence of P II/O I although the measures in other low metallicity environments give solar values (Table 6). On the contrary, Welsh et al. (2001) found a relatively high P II/O I ratio in the high velocity gas toward HD 47240, lying just behind the Monoceros Loop Supernova Remnant (SNR). This high ratio could be explained by the ionization effects of the SNR gas found by the authors.
The derived error-weighted mean P II/O I ratio of all these data points is -3.28+0.15-0.13, again consistent with the solar P/O value. Results of the previous section are thus confirmed and reinforced since P II/O I appears to be relatively homogeneous, at least within the Milky Way over nearly 5 kpc.
It is thus reasonable to propose that phosphorus, as traced by P II, can be used to trace O I column density in the diffuse ISM since there is no clear sign of differential depletion.
We investigated 10 Galactic sightlines and found that:
Acknowledgements
This work was done using the code Owens.f developed by M. Lemoine and the FUSE French team. We wish to thank Guillaume Hébrard, Jean-Michel Désert and Daniel Kunth for their help and for useful discussions.