Edward L. Fitzpatrick
Princeton University Observatory, Princeton, NJ 08544
dex for Si and
0.6 dex for the other elements.
Throughout the GTO period, beginning with observations in Cycle 0, I
have been working in collaboration with Lyman Spitzer, Jr. at Princeton
University on a multi-year spectroscopic investigation of the
interstellar medium (ISM) using data obtained with the Goddard
High Resolution Spectrograph. This program was based originally on
GTO time, and has been expanded most recently by GO time during Cycles
5 and 6. Our general goals are straightforward (to explain, if not to
achieve!): (1) to determine the physical properties of individual
interstellar clouds in the Galactic disk and halo and (2) to use these
new data to test and expand our understanding of the processes which
determine the composition, ionization state, temperature and pressure,
and kinematics of the ISM. In particular, we concentrate our attention
on three areas: the elemental abundances and depletions in neutral H I
gas, the physical conditions (particularly 
and 
) in the
clouds, and the relationship between highly-ionized gas and the low
ionization clouds.
We are pursuing these goals through detailed analyses of the
interstellar gas along the sightlines to a selected set of target stars
located in the Galactic disk and lower halo, accumulating statistics on
the (sometimes) many individual clouds found along each sightline.
Analyses have been completed already for lines-of-sight towards three
Galactic halo stars, HD 93521 (Spitzer & Fitzpatrick 1992, 1993,
hereafter Paper I), HD 149881 (Spitzer & Fitzpatrick 1995), and HD
215733 (Fitzpatrick & Spitzer 1996, hereafter Paper IV), as well as
towards one Galactic disk star, HD 68273 (
Vel; Fitzpatrick
& Spitzer 1994, hereafter Paper II). The halo program, which is the
subject of this short report, will be augmented in the future by
analyses of the lines of sight towards the stars HD 119608, HD 167756,
and HD 219188, as well as some additional observations of HD 93521 and
HD 149881. These halo targets are all early-type stars, with typical V
magnitudes of 
7--7.5 mag and z-distances of 
1--2 kpc.
For each target star, the primary data used in this program are GHRS
measurements of UV interstellar absorption features for a large number of
distinct species and covering several different ionization
stages, including: C I, C I
, C I
, C II, C II
, C IV, N
V, O I, Mg I, Mg II, Si II, Si II
, S I, S II, S III, Cr II, Mn II,
Fe II, and Zn II. The data are obtained using the GHRS in its
echelle modes (ECH-A and ECH-B) and with the stellar image placed in
the Small Science Aperture, yielding a spectral resolution of
3.5
. Observations of the halo star HD 149881 were made during
Cycle 3 when the ECH-A mode was unavailable; measurements of spectral
features at 
Å were obtained using the lower
resolution G160M observing mode. This star will be re-observed using
ECH-A during Cycle 5. The echelle observations of the halo targets
typically have signal-to-noise ratios in the range 20--30.
Figure: GHRS, Optical, and Radio, Interstellar Data for the Halo Star HD 215733
When possible, the analysis of the GHRS spectra is supplemented
with ground-based data. Danly et al. (1992) carried out H I 21-cm
emission line observations for a large number of stars in the Galactic
halo, including all the stars in our GHRS program, and these data
have been made available to us. The high signal-to-noise 21-cm spectra
were obtained using a 20' beam and have a spectral resolution of 1
.
Particularly important is that they have been corrected for
sidelobe contamination by 21-cm emission from gas in the Galactic
disk.
Recently, I have begun a project with Dan Welty (University of Chicago)
to obtain very high resolution measurements of the Ca II K and Na I
D1 and D2 lines towards all the halo stars in the GHRS
observing program (except the southern hemisphere star HD 167756),
using the Coudé feed telescope at Kitt Peak National Observatory.
This very efficient instrument can produce beautiful high
signal-to-noise (
100) spectra of Ca II K with a spectral
resolution of 
1.4 km s
in about 10 hours for these
relatively faint halo targets. Na I observations are about 30%
faster. These data were included in the analysis of the star HD 215733
(Paper IV) and will be incorporated in future studies of HD 119608 and
HD 219188, as well as in planned re-examinations of HD 93521 and HD
149881.
Figure 1 shows examples of the three types of data described above for the halo star HD 215733 (filled circles). Superimposed on the data are the final component models for this star, described below.
Our analysis is based on the component fitting technique. The choice of this technique as our analysis tool, and indeed the entire focus of this program on individual interstellar clouds, is a direct result of the unprecedented ability of GHRS data to resolve much of the structure present in the profiles of the myriad of important interstellar absorption features present in the UV region.
This method assumes that the often-complex interstellar absorption
profiles seen towards many stars (e.g., see Fig. 1!) result from the
superposition of a set of k simple absorption components, each of
which arises in a spatially distinct region, which we refer to here as
clouds. An interstellar feature for a species X
can thus be
described by the three sets of parameters 
(X
), 
(X
),
and 
(X
) which are, respectively, the column density of gas of
species X
, the two-dimensional velocity dispersion, and the
velocity centroid (in any desired reference frame) for each component
k. The velocity dispersions 
are the products of the kinetic
temperature of the gas and any macroscopic turbulence in the clouds.
We determine the values of 
, 
, and 
(which we refer to
as the component model) by (1) computing theoretical line
profiles based on initial guesses of 
, 
, and 
, (2)
comparing the theoretical profiles (computed assuming the Voigt
profile) with the observed profiles, and (3) iterating with revised
estimates of the component model (using the Marquardt method) until the
best fit to a particular feature or set of features is achieved. The
final component model for the halo star HD 215733 is illustrated in
Figure 1 as the solid lines through the data points. The component
structure towards this star is the most complex that we have examined,
with 23 individual clouds detected.
A straightforward application of the column densities measured for the gas in neutral H I clouds is a determination of the gas-phase abundances of the heavy elements relative to the standard solar system abundances. Such a study can provide information on both the intrinsic abundances of the elements in the ISM and on the composition of dust grains in the clouds (through the usual assumption that material ``missing'' from the gas is locked up in interstellar dust grains). For halo sightlines, these measurements may shed light on the origins and dynamical history of the high-z gas.
Ideally, we would like to express the elemental abundances for each
cloud relative to H and, at least initially, with respect to the solar
system abundance framework. However, in practice, we spurn H and adopt
the element S (and the S
column densities) to serve as the
reference species. This is because the individual H I 21-cm component
column densities are sometimes poorly determined and, more importantly,
because the 20' beam of the 21-cm measurements samples a much larger
spatial region than the very narrow ``beam'' of the absorption line
measurements. It is not necessarily reasonable to assume that the H I
column densities will be identical on these two different scales.
Along diffuse sightlines, such as those towards the halo stars, the
gas-phase abundance of S has been found to be close to the solar system
abundance level, suggesting that very little S is incorporated in dust
grains and that the ISM abundance of S is close to that of the solar
system.
Using S as the reference, we define the relative gas-phase abundance for element X in a cloud as
For the solar system abundances (X/S)
, we average the values
for meteorites and the solar photosphere listed by Grevesse & Anders
(1989) and we adopt log (Fe/S)
= +0.28 (see the discussion in
Paper I). Equation 1 assumes that essentially all of S and element X
in the gaseous state are in the singly-ionized form, as is expected in
H I gas for the elements under consideration here. The computation of
[X/S] is relatively insensitive to small amounts of S and X in other
ionization states since they tend to cancel out (see, e.g., Cardelli,
Sembach, & Savage 1995 and Sembach & Savage 1996). The quantity
[X/S] is often referred to as 
, the logarithmic
``depletion'' of X, when apparent deficiencies in the gas-phase
abundances are attributed to the incorporation of some fraction of
species X into dust grains.
Figure: Interstellar Gas-Phase Abundances Towards Three Halo Stars
Combining the results from Papers I (HD 93521), III (HD 149881), and IV
(HD 215733), we have determined relative gas-phase abundances for some
25 clouds along halo lines of sight. We summarize these data
graphically in Figure 2, where we plot [X/S] versus the absolute values
of 
for the elements Si and Fe (in the upper panel) and Cr,
Mn, and Ti (in the lower panel). The gas-phase abundances towards HD
149881 were computed originally relative to the element Zn; in Figure
2, we have corrected these values to the S scale (see Paper IV). The
values of [Fe/S] are indicated in the lower panel by the small x's to
facilitate comparison between the two panels.
Figure 2 shows that all the relative gas-phase abundances measured for
these elements are negative, indicating that the elemental abundances
in the gas are subsolar. We note the velocity dependence of
[X/S] seen at low values of 
, i.e., the most negative values
of [X/S] tend to be at small 
. This dependence, often
referred to as the Routly-Spitzer effect, is interpreted as arising
from the destruction of dust grains by the processes which accelerate
interstellar clouds, and indicates that at least some of the
differences between the observed gas-phase abundances and the solar
system abundances arise from depletion of the elements onto dust
grains.
Two important aspects of Figure 2 are that the gas-phase abundances of Fe, Si, Cr, Mn, and Ti towards some of the clouds in these directions are the largest that have ever been detected for these elements --- although they still do not reach the levels of the solar system abundances---and that there appear to be well-determined limits on the maximum values of [X/S]. Dotted lines are shown in the top panel of Figure 2 at [X/S] = -0.2 and -0.6, which represent the limits observed for Si and Fe, respectively. Upper limits for Cr, Mn and Ti are in the range -0.5 to -0.7.
In Papers I, II, and III, we interpreted the observed gas-phase
abundances exclusively as the result of depletions onto dust grains.
This would imply that, while some dust grain destruction must have
occurred to yield the relatively large values of [X/S], even in the
most extreme cases significant amounts of the elements are
still locked into grains (e.g., 37% of the Si and 75% of the Fe).
Further, it was argued that the inferred dust-phase abundances
could be understood in terms of very robust silicate grain cores,
perhaps the remnants of original silicate stardust injected into the
ISM. The composition of these cores could be mainly Si and Fe or,
depending on assumptions about the oscillator strengths of the
important Mg II 
1239, 1240 lines (see Sofia et al.
1994), could be a mixture of Si, Fe, and Mg (see, e.g., the paper by
Ken Sembach & Blair Savage in this volume, p. 
). This notion of
nearly-indestructible stardust silicates is attractive because it is
not clear that a formation channel for silicates exists in the ISM and
the strength of the 9.7 
m band seems to suggest that most
interstellar silicon is locked in silicate material.
In Paper IV we looked more critically at the issue of depletions in the halo gas and came to a different possible interpretation of Figure 2. In summary, given the fact that it has long been known that dust grains are relatively easy to destroy (e.g., Draine & Salpeter 1979), we find it surprising that neither we, nor others examining halo gas (e.g., Cardelli et al. 1995, Sembach & Savage 1996, Jenkins & Wallerstein 1996) have observed any clouds in which less than 75% of Fe, for example, is still locked into grains. We suggest instead that the limiting gas-phase abundances seen in Figure 2 are not a signature of depletion onto dust grains, but rather represent the true subsolar abundances of Si, Fe, Cr, Mn, and Ti in the halo clouds. Since most of the halo gas probably originated in the disk, we thus suggest that the general interstellar abundances of these elements are subsolar.
This seemingly extreme suggestion receives some support from recent
results on interstellar O which, based on GHRS measurements of
the O I 
1355] line, strongly suggest that the abundance of O
in the ISM is about 0.3 dex below the solar system value (Meyer et al.
1994). In addition, recent surface abundance analyses of main sequence
B stars are yielding metal abundances generally at the level of a
factor-of-2 below solar---but with a wide scatter in the results.
The surface abundances of such stars ought to reflect the abundances of
the ISM at a much more recent epoch than the those of the Sun.
The scatter seen among the B star abundance analyses may certainly indicate an intrinsic range in the surface abundances---arising perhaps from inhomogeneity in the ISM---but undoubtedly also arises from uncertainties in the analyses. Gies & Lambert (1992) for example, find an overabundance of Fe (by 0.2 dex) relative to the solar system scale, but the results of Cuhna & Lambert (1994) show that a wide range in Fe abundance can arise depending on which Fe lines are chosen for analysis. Recently Derck Massa (Applied Research Corporation) and I have been comparing low-resolution IUE energy distributions of lightly reddened main sequence B stars with the ATLAS9 model atmospheres available from R.L. Kurucz. This analysis essentially averages the results from thousands of individual spectral lines and, thus, should be insensitive to uncertainties in atomic parameters or radiation transfer peculiarities that may affect some individual transitions. We find that the models reproduce the observations extremely well, but only for significantly sub-solar metallicities. The mean metallicity of a sample of 70 stars, for example is -0.5 dex (Fitzpatrick & Massa, in preparation). This general metallicity likely reflects the abundance of Fe most closely, since Fe provides much of the UV opacity in the B stars, and is close to the interstellar Fe abundance proposed above.
The intrinsic abundances of the elements are important parameters in many areas of study of the ISM. In the future we plan to continue our examination of the gas-phase abundances in halo gas (as well as in the disk) with the general goal of defining better the properties of the gas and dust in the ISM and with a particular goal of testing the tantalizing suggestion of subsolar interstellar abundances that has come from the existing data.
The study of the elemental abundances in the halo clouds described above represents only one application of the data obtained for our general study of interstellar clouds but demonstrates well that, with its unique combination of high spectroscopic resolution and high photometric precision, the GHRS has the potential to make significant contributions to our understanding of the ISM. The expanded capabilities to be available in the future with the Space Telescope Imaging Spectrograph (STIS) can only enhance the possibilities for exciting discoveries in the field of ISM research.
This research has been supported by Rice University subcontract SC-437-5-16591 with Princeton University, supported in turn by NASA grant NAG5-1626 to Rice.
Cardelli, J.A., Sembach, K.R., & Savage, B.D. 1995, ApJ, 440, 241
Cuhna, K. & Lambert, D.L. 1994, ApJ, 426, 170
Danly, L., Lockman, F.J., Meade, M.R., & Savage, B.D. 1992, ApJS, 81, 125
Draine, B.T. & Salpeter, E.E. 1979, ApJ, 231, 438
Fitzpatrick, E. L. & Spitzer, L. 1994, ApJ, 427, 232 (Paper II)
Fitzpatrick, E. L. & Spitzer, L. 1996, in preparation (Paper IV)
Gies, D.R. & Lambert, D.L. 1992, ApJ, 387, 673
Grevesse, N. & Anders, E. 1989 in Cosmic Abundances of Matter, ed. C.J. Waddington (New York: AIP), 1
Jenkins, E.B. & Wallerstein, G. 1996, ApJ, in press
Meyer, D.M., Jura, M., Hawkins, I., & Cardelli, J.A. 1994, ApJ, 437, L59
Sembach, K.R. & Savage, B.D. 1996, ApJ, in press
Sofia, U.J., Cardelli, J.A., & Savage, B.D. 1994, ApJ, 430, 650
Spitzer, L. & Fitzpatrick, E. L. 1993, ApJ, 409, 299 (Paper I)
Spitzer, L. & Fitzpatrick, E. L. 1995, ApJ, 445, 196 (Paper III)
