Skip to main content

Full text of "The Properties of the local Interstellar Medium and the Interaction of the Stellar Winds of epsilon Indi and lambda Andromedae with the Interstellar Environment"

See other formats


NASA-CR-203138 

The Astrophysic al Journal, 470:1157-1171,1996October20 

C) 1996. The American Astronomical Society, All rights reserved. Printed in U.S.A. 






SS^^^ 



//^ 



c^ 



THE PROPERTIES OF THE LOCAL INTERSTELLAR MEDIUM AND THE INTERACTION 

OF THE STELLAR WINDS OF e INDI AND X ANDROMEDAE WITH THE 

INTERSTELLAR ENVIRONMENT^ 

Brian E. Wood 

JILA, University of Colorado, Boulder, CO 80309-0440; wood@inarmot.colorado.edu 

William R. Alexander 

1250 Buffalo Creek Road, Huntington, WV 25704-9608; walexander@ashland.e-mail.com 

AND 

Jeffrey L. Linsky 

JILA, University of Colorado, Boulder, CO 80309-0440; jlinsky@jila.colorado.edu 

Received 1995 November 27; accepted 1996 May 6 

ABSTRACT 

We present new observations of the Lya lines of c Indi (K5 V) and A Andromedae (G8 IV III + ?). 
These data were obtained by the Goddard High Resolution Spectrograph (GHRS) on the Hubble Space 
Telescope. Analysis of the interstellar H i and D i absorption lines reveals that the velocities and tem- 
peratures inferred from the H i lines are inconsistent with the parameters inferred from the D i lines, 
unless the H i absorption is assumed to be produced by two absorption components. 

One absorption component is produced by interstellar material. For both lines of sight observed, the 
velocity of this component is consistent with the velocity predicted by the local flow vector. For the c 
Ind data, the large velocity separation between the stellar emission and the interstellar absorption allows 
us to measure the H i column density independent of the shape of the intrinsic stellar Lya profile. This 
approach permits us to quote an accurate column density and to assess its uncertainty with far more 
confidence than in previous analyses, for which the errors were dominated by uncertainties in the 
assumed stellar profiles. For the short {d = 3.46 pc) line of sight to e Ind, the H i column density is 
found to be log Nhi= 18.0 + 0.1, which implies an average density for the local interstellar medium 
(LISM) of «H , = 0.094 ± 0.022 cm"^ For the much longer {d = 23 pc) line of sight to A And, we estimate 
the H I column density to be log N^i = 18.45 ±0.15, which corresponds to an average density of «„, = 
0.041+0.014 cm ^ The D/H ratios we measure from the data are (1.6 + 0.4) x 10"' and 
(1.7 + 0.5) X 10 * for e Ind and A And, respectively. These values are consistent with those measured 
from observations of Capella, Procyon, and a Cen. We measure LISM temperatures of T = 8500 ± 500 
K and T = 11,500 ± 500 K from the e Ind and A And data, respectively. The i And temperature is 
significantly higher than temperatures previously measured from GHRS data, which leads us to specu- 
late that the H i and D i absorption lines may be broadened by multiple ISM components with dilTerent 
velocities. The results of our /I And analysis should be considered as tentative, until GHRS observations 
of the much narrower Mg ii and/or Fe ii absorption lines can be obtained. 

We believe that hot hydrogen surrounding e Ind and A And is responsible for the second H i absorp- 
tion component, although we consider this conclusion to be tentative in the case of A And. These 
" hydrogen walls " are produced by the interaction of the winds of these stars with the surrounding inter- 
stellar material. An anologous solar hydrogen wall has been predicted by recent models of the helio- 
spheric interface region and confirmed by GHRS observations of a Cen. The column densities we 
measure for the second components are log N^j = 14.2 + 0.2 and log Nh, = 14.8 + 0.2 for c Ind and A 
And, respectively, and the temperatures are 100,000 ± 20,000 K and 62,000 ± 18,000 K. These tem- 
peratures are too hot for the solar hydrogen wall, and for e Ind the velocity of the second component is 
clearly inconsistent with the solar hydrogen wall. Thus, for these components we assume a stellar origin, 
in which the higher temperatures are a consequence of higher interstellar wind velocities in the stellar 
rest frames. Because the heliospheric models demonstrate the importance of the solar wind in the forma- 
tion of the solar hydrogen wall, our detection of anologous structure around e Ind and perhaps A And 
may constitute a first detection of solar-like winds around dwarf and subgiant stars. 
Subject headings: ISM: abundances — ISM: kinematics and dynamics — 

stars: individual (c Indi, X Andromedae) — stars: mass loss — ultraviolet: stars 

' Based on observations with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science Institute, which is operated by the 
Association of Universities for Research in Astromomy Inc., under NASA Contract NAS 5-26555. 



1157 



1158 



WOOD, ALEXANDER, & LINSKY 



Vol. 470 



1. INTRODUCTION 

The Goddard High Resolution Spectrograph (GHRS) 
aboard the Hubble Space Telescope (HST) is an invaluable 
tool in the study of the local interstellar medium (LISM) 
because it is the first instrument capable of fully resolving 
narrow interstellar absorption features in the ultraviolet. 
The first study of the LISM using GHRS observations was 
that of Linsky et al. (1993, hereafter Paper I), who deter- 
mined the properties of the LISM along the 12.5 pc line of 
sight to Capella. The focus of their analysis was on the 
abundance ratio of deuterium to hydrogen (i.e., the D/H 
ratio). This ratio can be used to estimate the primordial 
D/H ratio, which is a very important quantity for cosmol- 
ogy because it is the most sensitive diagnostic available for 
estimating the baryon density of the universe (see, e.g.. 
Walker et al. 1991). The D/H ratio for the Capella Hne of 
sight was found to be l.65t^°il x 10"^ (Paper I). An 
analysis of more recent GHRS observations of Capella led 
Linsky et al. (1995, hereafter Paper II) to slightly alter this 
value to 1.60 i!:":!* x 10 '. The errors quoted here are a 
combination of the random errors and the estimated sys- 
tematic errors. 

In Paper II, GHRS observations of ISM absorption lines 
seen along the line of sight toward Procyon [d = 3.5 pc) 
were also analyzed. Unfortunately, the echelle-A grating 
was not available at the time of these observations, and it 
was therefore necessary to use the lower resolution G160M 
grating. This complicated the analysis because the ISM 
lines were not fully resolved. Another complicating factor 
was a second ISM absorption component, which is separat- 
ed by only 2.6 km s " ' from the component associated with 
the local cloud. An estimate of D/H = (1.4 ± 0.2) x 10"* 
was given in Paper II, which is consistent with the Capella 
result. It was also shown that a resonable alteration of the 
assumed stellar Lya hne yields a D/H value even closer to 
the Capella result. Lemoine et al. (1995) found three ISM 
absorption components in their analysis of the ISM along 
the line of sight toward the white dwarf G191-B2B, which is 
only 7° away from Capella. One of the three components 
detected toward G191-B2B matched the absorption com- 
ponent detected for the Capella line of sight very well in 
terms of velocity, temperature, and column density. 
However, the value of D/H for this component derived by 
Lemoine et al. (1995) on the basis of their G160M obser- 
vations of Lya, D/H= 1.4:!;^:^ x 10"', is slightly lower 
than the value derived in Paper II. 

In addition to measuring the local D/H ratio, these 
analyses also sought to investigate the properties of the 
LISM. Using measurements of optical Ca ii LISM absorp- 
tion hnes observed for many lines of sight, Lallement & 
Berlin (1992) determined two different flow vectors for local 
interstellar material. For lines of sight in the general direc- 
tion of the Galactic center, the flow vector points toward 
/ = 184?5 and b = -20?5 with a velocity of v^ = 29 A km 
s" S whereas the flow vector for Hnes of sight in roughly the 
anti-Galactic center direction points toward / = 186?1 and 
h = - 16?4 with a velocity of v^ = 25.7 km s '. These two 
vectors are referred to as the V^ and F^g vectors, respec- 
tively, and the intersteUar clouds they represent are the G 
and AG clouds. The latter is also called the Local Inter- 
stellar Cloud (LIC). Since the V^ and F^g vectors are very 
similar, it is debatable whether the G and AG clouds are 
really separate entities. Nevertheless, there is evidence that 



LISM material in the Galactic center direction is moving 
shghtly faster than LISM material in the opposite direction. 
Other ISM absorption components have been observed 
toward very nearby stars that are not consistent with either 
the Vq or the F^g vectors. For example, besides the afore- 
mentioned second component seen in the Procyon data, 
Lallement et al. (1994) observed a second ISM absorption 
component for the even shorter 2.7 pc line of sight toward 
Sirius. 

The Ulysses satellite measured a velocity of 26 + 1 km 
s~^ for interstellar He i atoms in the solar system (Witte et 
al. 1993). This velocity agrees much better with the modulus 
of the Fag vector (25.7 km s"') than the modulus of the Fq 
vector (29.4 km s '), suggesting that the former represents 
the true local cloud. The GHRS observations of Capella, 
Procyon, G191-B2B, Sirius, and other stars (see, e.g., Gry et 
al. 1995; Ferlet et al. 1995; Lallement et al. 1995) have also 
established the F^g vector as the flow vector that best char- 
acterizes the interstellar material in the immediate vicinity 
of the Sun. As summarized by Lallement et al. (1995), nearly 
all of the lines of sight observed by the GHRS so far show 
interstellar absorption at the velocity predicted by the F^o 
vector, and almost none shows absorption at the velocity 
predicted by the Fg vector. The exception is the 1.3 pc line 
of sight to a Cen, which ironicafly is the nearest star system 
to the Sun and therefore is the line of sight that one would 
think would best represent the properties of the ISM in the 
immediate solar vicinity. The interstellar material detected 
along the a Cen line of sight appears to be entirely in the G 
cloud. If the Fag vector really does best represent the local 
cloud, it is very curious that no absorption is seen at the 
velocity predicted by this vector. Lallement et al. (1995) 
interpreted this to mean that the boundary between the G 
and AG clouds is very nearby for the Sun a Cen line of 
sight. 

In a thorough analysis of GHRS echelle observations of 
intersteUar Mg ii, Fe ii, H i, and D i absorption lines seen 
toward a. Cen, Linsky & Wood (1996, hereafter Paper III) 
found that the temperature and abundances of the material 
along the a Cen hne of sight are diiferent from those seen 
toward Procyon and Capella. The temperature and non- 
thermal velocity found for the Capella line of sight are 
T = 7000 + 900 K and ^ = 1.6 + 0.6 km s" ', respectively 
(Paper II). Similar values were found for the Procyon line of 
sight (r = 6900 ±380 K and ^ = 1.21 ± 0.27 km s '; 
Paper II). These quantities include estimates of systematic 
errors. Although the nonthermal velocities detected toward 
a. Cen (^ = 1.20 ± 0.25 km s"') are similar to the Capella 
and Procyon line-of-sight values, the temperature reported 
in Paper III is significantly lower (T = 5400 ± 500 K). 

The quantity used in Paper III to demonstrate the differ- 
ence between the LISM abundances along the a Cen line of 
sight and the abundances observed toward Capella and 
Procyon is the D i/Mg ii ratio. This ratio is 4.1 + 0.4 and 
5.1 + 0.7 toward Capella and Procyon, respectively (Paper 
II), but toward a Cen the ratio is much lower (1.2 + 0.2; 
Paper III). The D i/P'e ii ratio shows a similar behavior. 
This proves that abundances vary significantly in the LISM 
over distance scales of only a few parsecs. In order to deter- 
mine whether it is D/H, Mg/H (and Fe/H), or both that are 
varying, it is necessary to measure accurately the H i 
column densities for all these stars to calculate the absolute 
abundances. Unfortunately, this proved to be impossible for 
the a Cen line of sight. 



No. 2, 1996 



e IND AND X AND 



1159 



The analysis of the a Cen line of sight in Paper III pro- 
vides a dramatic illustration of the difficulties in measuring 
interstellar H i column densities from the very broad, satu- 
rated absorption features present in Lya lines. The profiles 
of the Mg II, Fe ii, and D i absorption fines show no evi- 
dence for more than one absorption component, but when 
the H I Lya absorption line was fitted with only one com- 
ponent, it was found that the velocity and temperature mea- 
sured from this line were inconsistent with the other fines. 
The only way to resolve this discrepancy was to add a 
second absorption component to the H i line, one with a 
column density too low to be detected in the other lines. 
Although the single-component fits to the H i absorption 
observed toward a Cen A and B were shown to be unique 
solutions, the two-component fits were not. By modifying 
the assumed stellar Lya profiles used for a Cen A and B in a 
reasonable way, we found acceptable fits to the observed 
profiles with very different values for the interstellar H i 
column density, leading to a very large uncertainty in this 
quantity (log Nh, = 17.8 ± 0.3). 

The additional absorption component was found to have 
a very interesting interpretation. Recent models of the hefio- 
spheric interface region (shown schematically in Fig. 1), 
where the outflowing solar wind collides with the inter- 
stellar wind, predict a region of heated, compressed, and 
decelerated H i just outside the hefiopause (see, e.g., 
Baranov & Malama 1995; Pauls, Zank, & Wilfiams 1995). 



This region, shown as a shaded area in Figure 1, has been 
referred to as a " hydrogen wall." The temperature, column 
density, and flow velocity of the hydrogen wall predicted by 
the models agree nicely with the parameters of the second 
H I absorption component detected toward a Cen, which 
suggests that the hydrogen wall has been detected, thereby 
supporting the models of the theorists. After taking into 
account the efi"ects that a similar hydrogen wall around a 
Cen might have on the analysis, the temperature and 
column density of the solar hydrogen wall along the a Cen 
line of sight were estimated to be T = 29,000 ± 5000 K and 
log Nh I = 14.74 + 0.24, respectively (Paper III). 

With all this in mind, we now present our analysis of the 
H I and D i absorption features detected in the Lya lines of e 
Indi and X Andromedae. Besides investigating the proper- 
ties of the LISM with these data, we also hope to learn more 
about the solar hydrogen wall and/or detect analogous 
hydrogen walls around e Ind and X And. 

2. GHRS OBSERVATIONS OF e rWDI AND X ANDROMEDAE 

The two targets we have studied with the GHRS were 
previously observed with the Copernicus and lUE satellites 
in order to analyze the properties of the LISM along these 
fines of sight (McCfintock et al. 1978; Baliunas & Dupree 
1979; Murthy et al. 1990). The K5 V star e Ind ( = HD 
209100) is located only 3.46 pc from the sun at Galactic 
coordinates / = 336° and b = —48° (Gliese & Jahreiss 




Fig. 1. — Schematic illustration of how the solar wind {thin solid lines) interacts with the plasma component of the interstellar wind {dotted lines). The solar 
wind is decelerated from supersonic speeds to subsonic speeds at the termination shock. The bow shock is a standing shock wave through which the 
supersonic interstellar wind passes. The heliopause is the contact surface separating the plasma flows of the solar and insterstellar winds. Some neutral 
hydrogen in the interstellar wind is deflected around the heliopause with the protons, but some also penetrates into the heliosphere. The shaded area is a 
region of heated, compressed, and decelerated H I that has been referred to as a "hydrogen wall." The hydrogen wall is roughly 150 AU from the Sun in the 
upwind direction according to recent models of the heliosphere (Baranov & Malama 1995; Pauls et al. 1995). 



1160 



WOOD, ALEXANDER, & LINSKY 



Vol. 470 



TABLE 1 
Summary of GHRS Observations 



Target 


Grating 


Aperture and 
Substep Pattern 


Spectral Range 

(A) 


Spectral Resolution 
(km s ') 


Exposure Time 
(s) 


Date 


Start Time 
(UT) 


/I And 

£lnd 


EA-46 
EA-46 


SSA 9 
SSA 9 


1212-1219 
1212 1219 


3.57 
3.57 


3318 
3291 


1994 Aug 31 

1995 Mar 18 


15:18 

2:42 



1991). The k And ( = HD 222107) star system is an RS CVn 
binary system about 23 pc away (Jenkins 1952) with a 20.52 
day orbital period (Batten, Fletcher, & Mann 1978), consist- 
ing of a G8 IV III star and a companion star of unknown 
spectral type (Herbst 1973; Strassmeier et al. 1988). The 
Galacticcoordinatesof A And are/ = 110"and/5= —15". 

Table 1 summarizes the relevant information about our 
observations of the Lya lines of e Ind and X And. Images of 
the Pt-Ne calibration lamp were taken prior to both 
observations and were used to calibrate the wavelengths of 
the spectra. The spectra displayed in Figure 2 show very 
broad absorption lines centered near 1215.6 A that are due 
to interstellar H i. Located about —0.33 A from the H i 
absorption lines are the narrower D i absorption lines. The 
emission feature at 1215.74 A in the € Ind spectrum is due to 
geocoronal H i. In our analysis, we corrected for this emis- 
sion by fitting a Gaussian to the feature and then subtrac- 
ting the Gaussian from the data. 

The cores of the H i absorption lines should have zero 
flux because of the interstellar absorption, but our spectra 
show positive flux there, which is caused by scattered light 
(see Papers 1 and III). Because the scattered light level is 
constant in the saturated line cores, we correct for this flux 




1214 



1215 1216 

Wavelength 



1217 



Fig. 2. — GHRS observations of the Lya lines of f. Ind and X And, 
showing broad interstellar H i absorption at 1215.6 A and narrow D I 
absorption near 1215.3 A. The emission feature at 1215.74 A in the e Ind 
spectrum is geocoronal emission. 



by simply subtracting the mean flux level observed in the 
line cores (1.2 x 10 '^ and 3.1 x 10"'^ ergs cm"^ s ' A ' 
for 6 Ind and X And, respectively) from a spectral region 
(1215.2-1215.85 A and 1215.2-1216.1 Afor e Ind and / And, 
respectively) encompassing the absorption. In all figures 
except Figure 2, the spectra displayed will be the observed 
spectra after the corrections for the scattered light and geo- 
coronal emission have been applied. 

3. SINGLE-COMPONENT FITS TO THE DATA 

The first step in analyzing the line profiles is to obtain 
initial estimates for the intrinsic stellar Lya emission lines. 
Because e Ind has a spectral type (K5 V) similar to a Cen B 
(Kl V) and both are inactive stars, we adopt the Lya profile 
used for a Cen B in Paper III (profile model IB, to be 
specific) as our first estimate for the line profile of e Ind. We 
normalize the fluxes of this profile so that the model fits the 
observed profile in the far wings of the Lya fine. 

Following a procedure similar to that used in Paper III, 
we use a -f^ minimization technique to determine the best 
single-component fit to the D i and H i absorption lines. 
The atomic data necessary for the computation are taken 
from Morton (1991). Throughout this paper, we correct 
for instrumental broadening using a Gaussian with 
FWHM = 3.7 pixels as the instrumental profile (Gifliland 
1994). Because the D i and H i absorption features are both 
fine-structure doublets, we include both components of the 
doublets in our computations. For a Cen A and B, the D i 
and H I absorption lines are well separated, and it was not 
difficult to use a polynomial fit to estimate what the line 
profile would look like in absence of any D i absorption. 
Therefore, in Paper III the D i and H i lines were analyzed 
separately. We cannot do this for either e Ind or ). And 
because we do not believe that polynomial fits can accu- 
rately estimate the "continuum" upon which the D i 
absorption is superposed (see Fig. 2). Therefore, in this 
paper we must fit the D i and H i lines simultaneously. 

The initial fit to the e Ind data using the a Cen B profile 
model was not very good, so we used the residuals of that fit 
to alter the assumed stellar profile. This greatly improved 
the quality of the fit. After some more adjustments to the 
stellar profile, we obtained our best single-component fit to 
the e Ind data, shown in Figure 3. The parameters of the fit 
are given in Table 2, including the line velocities, Dopplcr 
parameters {h\ logarithmic column densities (log A^), and 
optical depths (t) of the H i and D i absorption lines. The 
quality of all the fits in this paper can be assessed either by 
looking at the residuals displayed in the figures or by noting 
the Xv values listed in the tables (see, e.g., Bevington & 
Robinson 1992). The errors given in Table 2, which were 
estimated using Monte Carlo techniques, represent only the 
random errors in the fitting procedure and do not include 
the systematic errors that certainly dominate the uncer- 
tainty in the analysis. 



No. 2, 1996 



e IND AND / AND 



1161 




121,").0 1215,2 1215 4 1215 6 1215 

Wavt'lenglh 



1216 



1216,2 



Fig. 3. — Our best single-component fit to the H i and D i absorption lines of t Ind, with the residuals displayed below the fit. The data are shown in 
histogram form. The thin .solid line is the assumed stellar Lyot profile, and the thick solid line is the fit. The parameters for the fit are given in Table 2. 



For I And, we adopt as our initial estimate for the intrin- 
sic stellar profile the Lya profile used in Paper I for the G8 
star of the Capella binary system. This is reasonable 
because the spectral type of k And (G8 IV-III) is very 
similar to that of the primary star of the Capella system (G8 
III). We narrow the profile slightly to force the wings of the 
model profile to agree better with the shape of the wings of 
the observed Lya profile, and then we fit the data as we did 
for e Ind. Figure 4 shows our best single-component fit 
to the X And data. The parameters of the fit are given in 
Table 2. 

The residuals of the e Ind fit in Figure 3 reveal systematic 
discrepancies between the fit and the data. Attempts to 
correct these discrepancies by altering the assumed stellar 
Lya profile resulted in unreasonable fine structure in the 
stellar profile. For example, to correct for the flux excess 
near 1215.47 A one must assume that the stellar profile flux 
near 1215.47 A is at least 3 times larger than the flux at 
1215.45 A. A flux increase this steep is not reasonable. The X 
And fit in Figure 4 is a much better fit than the c Ind fit, but 
both suffer from another problem, namely that the param- 
eters of the fits in Table 2 suggest that D i and H i have 
different velocities and temperatures. The same problem 
was found in Paper III for the single-component fits to the a 
Cen A and B spectra. The H i hnes of both e Ind and X And 
appear to be blueshifted relative to the D i Hnes by about 3 
km s ^ Temperatures for H i and D i can be computed 
from the b values in Table 2 by using the equation 
h'^ = 0.0165T/A -I- ^^, where A is the atomic weight (1 for 



H I, 2 for D i) and t, is the nonthermal velocity. We cannot 
measure ^ as we did in Papers I III without observations of 
the Mg II and Fe ii absorption lines, so throughout this 
paper we will simply use the value measured in Paper III, 
1^ = 1.2 km s" ', which is also consistent with the resuhs of 
Papers I and II. The exact value used is not too important 
since Papers I-III have conclusively shown that LISM H i 
and D i lines are dominated by thermal rather than non- 
thermal broadening. For e Ind, the H i temperature, 8620 
K, is significantly hotter than the D i temperature, 6980 K. 
A similar discrepancy is seen for X And, where the H i and 
D I temperatures are 13,040 and 10,000 K, respectively. 

For X And, we computed another single-component fit in 
which we forced the velocities and temperatures of D i and 
H I to be consistent. The resulting fit was very poor, so we 
used the residuals of that fit to dramatically alter the 
assumed stellar Lya profile, and then we fit the data again. 
The result is shown in Figure 5. In trying to force consis- 
tency between D i and H i, we were forced to make the 
stellar profile (Fig. 5, solid line) very asymmetric with a very 
strong red peak. This is not a very reasonable profile, but in 
any case, the fit resulting from the use of this profile still is 
not very good. The flux excess at 1215.3 A seen in Figure 5h 
indicates that the D i line velocity is more redshifted than 
the fit suggests, and the flux deficit near 1215.45 A indicates 
that the H i line velocity is actually more blueshifted than 
the fit suggests. Trying once more to use the residuals of the 
fit to alter the stellar profile and improve the fit, we derive 
the profile shown as a dotted line in Figure 5a. This profile 







TABLE 2 
Parameters for the Single-Component Fits 






Star 


Ion 


Line 


Velocity 
(kms"') 


b 
(kms"') 


log N 


T 


xi 


6 Ind 

^ And 


H I 
D[ 

Hi 
Di 


1215.670 
1215.339 

1215.670 
1215.339 


-12.6 + 0.1 
-9.9 ±0.1 

+ 3.3 + 0.1 
+ 6.6 + 0.1 


11.98 + 0.09 
7.68 ± 0.19 

14.71 +0.16 
9.16 + 0.18 


18.292 + 0.003 
13.127 ±0.011 

18.645 + 0.002 
13.658 + 0.013 


123,000 
1.31 

227,000 
3.74 


1.361 
1.361 

1.095 
1.095 



1162 



WOOD, ALEXANDER, & LINSKY 



Vol. 470 




1214.5 1215.0 1215.5 1216.0 

Wavelength 



1216.5 



1217.0 



Fig. 4.— Out best single-component fit to the H i and D i absorption lines of X And, with the residuals displayed below the fit. The data are shown in 
histogram form. The thin sohd line is the assumed stellar Lya profile, and the thick solid line is the fit. The parameters for the fit are given in Table 2. 



is even less plausible, however, because of the rapid flux 
changes between 1214.3 A and 1215.5 A. We conclude that 
the only way to force consistency between D i and H i is to 
add a second H i absorption component. 

For c Ind, we mention yet another problem with the fit in 
Figure 3. The upper part of the assumed stellar profile is 



X And 




'^mt 



HH l kyf ^'-** 



' ^*^y^'^f* l' U 



1215.0 1215.5 1216.0 1216.5 

Wavelength 

Fig. 5. — Our best single-component fit to the H i and D i absorption 
lines of A And in which H i and D I are forced to have the same velocity and 
temperature. The vertical dashed lines mark the boundaries of the saturat- 
ed core of the H i absorption. Between these lines, the shape of the assumed 
stellar profile has no effect on the quaUty of the fit. The assumed profile, 
shown as a solid line in (a), has been dramatically altered from that 
assumed in Fig. 4 in order to maximize the quality of the fit, but (b) reveals 
that there are still significant discrepancies between the fit and the data. 
When the residuals of the fit are used to alter the stellar profile in an 
attempt to improve the fit, the profile shown as a dotted line in (a) results. 
This is not a reasonable profile, suggesting that H [ and D i cannot be 
forced to be consistent in a single-component fit. 



redshifted by 5 10 km s ' from the stellar radial velocity of 
-38.9 ± 0.3 km s" ' (Buscombe & Kennedy 1968). The far 
wings of € Indi's Lya Une, away from the effects of the 
interstellar absorption, do not show this redshift, and there 
is no evidence that the Lya lines of inactive main-sequence 
stars like e Ind should have significant redshifts. For 
example, the Lyman-a lines of a Cen A, a Cen B, and 
Procyon appear to be well centered on the stellar radial 
velocity (see Papers II, III; Wood et al. 1996). All of the 
single-component fits to the e Ind data that we have tried 
(assuming different stellar profiles, forcing D i and H i to be 
consistent, etc.) suggested log JVhi « 18.3. We will now 
show that this value cannot be correct because this value 
is inextricably hnked with unreasonably redshifted stellar 
profiles. 

Primarily because of the large radial velocity of € Ind 
{v = -38.9 + 0.3 km s" ' ; Buscombe & Kennedy 1968), the 
interstellar H i absorption is redshifted by about 30 km s ~ ^ 
from the center of the stellar emission line, and therefore, 
there is more absorption in the red wing of the line than in 
the blue wing. Thus, the wings of the observed Lya profile 
are blueshifted from the expected velocity of — 38.9 km s '. 
By assuming different values for JVhi, we can reconstruct 
the wings of the intrinsic stellar Lya profile. When the 
correct value for log N„ , is assumed, the result should be a 
profile centered at —38.9 km s" '. This process is illustrated 
in Figures 6 and 7. The numbers in Figure 6 are assumed 
values for log Nf^ „ and the dotted lines associated with 
these numbers are the wings of the intrinsic stellar Lya line 
computed on the basis of these values by multiplying the 
observed flux by exp ( + t J, where z^ is the optical depth of 
the damping wing of the interstellar hydrogen absorption 
line. We assume that the interstellar H i absorption is cen- 
tered on the D i velocity listed in Table 2. These computa- 
tions rely on completely smooth estimates of the observed 
fine wings (Fig. 6, thick solid lines) obtained using poly- 
nomial fits to the data. 

In order to determine which value of log Nhi leads to an 
intrinsic stellar profile whose wings are best centered on the 
stellar radial velocity, we compute bisectors for the com- 



No. 2, 1996 



e IND AND X AND 



1163 



2 



r f) 



0.0 




100 -50 

Velocilv (km s ') 



InO 



Fig. 6. — Reconstructions of the wings of the Lya profile of e Ind, assuming different values for the interstellar H i column density, log N„ ,. The data are 
displayed in histogram form. The thick solid lines are polynomial fits to the wings of the observed Lya profile. The wings of the intrinsic stellar Lya profile 
{dotted tines) are computed for five different values for log N,,, (the numbers in the figure), using the polynomial fits as smoothed representations of the data. 
The correct value of log Nh i should produce a stellar profile centered on the stellar radial velocity of — 38.9 km s~ ', which is marked by the dashed line. 



puted profiles. Figure 7 shows the bisectors of the intrinsic 
stellar Lya line computed for many different values of 
log Nh I (the numbers in Fig. 7). The radial velocity of the 
star is indicated by the vertical dashed line. For log A^h j = 
17.6-17.8, the bisector is blueshifted relative to the expected 
velocity by 2-3 km s ', whereas for log Nhi = 18.2-18.4, 
the bisector is redshifted relative to the expected velocity by 
2-6 km s '. The value of log N„[ that produces a stellar 
Lya profile that is best centered on the radial velocity of the 
star is logNHi= 18.03. The bisector for this profile is 
shown as a solid line in Figure 7. 



Ideally, this bisector should be a vertical line exactly at 
the expected velocity. The obvious deviations from this 
ideal are probably due to the uncertainties involved in esti- 
mating the shape of the observed profile using polynomial 
fits and the uncertainties involved in the computation of the 
bisectors. Since this analysis is dependent on an accurate 
stellar radial velocity and an accurate wavelength cali- 
bration, uncertainties in both of these things are potential 
sources of error. After consideration of all the sources of 
uncertainty mentioned above, we report a value and error 
for the hydrogen column density of log Nhi = 18.0 + 0.1. 



2.0 



< 1.5 



1.0 



3 0.5 



01 



- 


' 1 ' 






' 


1 y -r- 




' ' 1 1 1 


' ' 18.3. ■;■: 


- 








\ 






18.2 .'/■■/■; 
. '.■.■'.■'.■'.''.■.■..■ 




- 








1 




IB.l ■-'/ 






: 






17 9 

• 


t 


18 / 
■ ;•■ ■■/■ ■■ 








- 


17.7 
• 
17.6 


17 a 

.... *.. ... 




■!■■■: 

1 


/ 












S^\Mi^ 


y 












- 


^%|| 


L 


mm 


y.-- 


^^^ 









42 



-40 



-38 36 

Velocily (km s ') 



-34 



-32 



30 



Fig. 7. — Dotted lines are bisectors of the intrinsic Lya profile of e Ind computed (as illustrated in Fig. 6) for many different values for log N^, (the 
numbers in the figure). The bisector that best agrees with the stellar radial velocity of —38.9 km s^ ' (dashed tine) is shown as a solid line. This bisector was 
computed assuming log N„ , = 18.03. 



1164 



WOOD, ALEXANDER, & LINSKY 



Vol. 470 




1215.0 1215,2 1215-4 1215.6 1215.8 

Wavelength 



1216.0 



1216.2 



Fig. 8. — Single-component fit to the H I absorption line of e Ind, in which the D i Une has been removed using the D i parameters listed in Table 2. The 
assumed stellar Lya profile was created using the log Nh , = 18.03 value measured from Fig. 7. In this fit, we have in fact forced log N„ , = 18.0. The resulting 
fit is very poor, demonstrating that a single-component fit cannot fit the data if log Nh , is in fact 18.0 ± 0.1. 



Note that log Nh, = 18.3, the value suggested by the single- 
component fits, is well outside these error bars. 

Now that we have a measurement of log Nh i, we can 
create a stellar profile designed with this value in mind. We 
start with Lya fine wings computed like those in Figure 6 
assuming log Nhi = 18.0. We then model the center of the 
line profile with the a Cen B profile, which was also used in 
deriving the profile in Figure 3. Using this profile, in Figure 
8 we perform a single-component fit to the H i line, 
requiring log Nhi to be 18.0. For this fit, we removed the 
D I line from the data using the D i parameters in Table 2 in 
order to focus attention on the H i absorption, the idea 
being to show that log Nh i = 18.0 cannot lead to an accept- 
able single-component fit regardless of whether or not the 
parameters for D i and H i are consistent. This is illustrated 
by the very poor quahty of the fit in Figure 8. Absorption 
lines with log Nhi = 18.0 have sides that are simply too 
steep to match the e Ind data. Thus, a single-component fit 
cannot work for the e Ind line of sight. 

4. TWO-COMPONENT FITS TO THE DATA 
4.1. Analysis of the € Ind Data 

We have concluded that a second H i absorption com- 
ponent must be present for both e Ind and / And to fit the 



data adequately and to explain the apparent discrepancies 
between H i and D i. Based on the results of Paper III, the 
most likely sources for the additional absorption are solar 
and/or stellar hydrogen walls. 

For e Ind, initial two-component fits were performed, and 
the stellar profile was then altered slightly to improve the 
quaUty of the fits. In all of our two-component fits, we force 
the D I and H i lines for each component to have the same 
velocities and temperatures for both components. We also 
force the two components to have the same D/H ratio, 
although the D i column density of the second component is 
too small to produce any noticeable absorption. 

The final result is shown in Figure 9, and the fit param- 
eters are given in Table 3. The dotted line in Figure 9 rep- 
resents the absorption only from the ISM. The second 
absorption component {dashed line) is blueshifted relative 
the ISM component, explaining why the H i line was blue- 
shifted relative to D i in the single-component fits. The 
two-component fit to the e Ind data is clearly a better match 
to the data than the single-component fit. The single- 
component model did not fit the data well, mostly because 
the H I absorption line appears to have an asymmetry in 
which the red side of the line is somewhat steeper than the 
blue side (see Fig. 3). This asymmetry is nicely explained by 
the two-component fit (see Fig. 9). 



TABLE 3 
Parameters for the Two-Component Fits 



Star 


Velocity 
(kms') 


''hi 
(kms"') 


logN„, 


D/H 

(10-') 


T 


xl 


find 

AAnd 


-9.4 + 0.1 
-27.8 ± 1.5 

+ 6.5 + 0.1 
-2.0 ± 1.1 


12.15 + 0.10 
41.39 ± 1.20 

13.87 + 0.18 
31.89+ 1.66 


17.949 ±0.019 
14.244 ± 0.031 

18.455 + 0.002 
14.722 + 0.120 


1.74 + 0.12 
(1.74 ±0.12) 

1.72 + 0.06 
(1.72 ± 0.06) 


55,300 
3.21 

155,000 
12.5 


1.117 
1.117 

1. 149 
1. 149 



No. 2, 1996 



c IND AND X AND 



1165 




1215.0 1215.2 1215,4 1215.6 1215.6 1216.0 12162 

Wavelength 

Fio. 9. — Our best two-component fit to the H [ and D i absorption lines of e Ind, with the residuals displayed below the fit. The data are shown in 
histogram form. The thin sohd line is the assumed stellar Lya profile. The dotted line is the component representing absorption from the ISM, the dashed fine 
is the component representing absorption from a hydrogen wall surrounding e Ind (see § 4.1), and the thick solid line is the combination of the two 
absorption components. The parameters for the fit are listed in Table 3. 



Because hydrogen-wall material around a star should be 
decelarated by its interaction with the stellar wind, the 
velocity of this material should generally lie between the 
stellar velocity and that of the ISM. Thus, for the e Ind line 
of sight, the solar hydrogen wall should have a velocity 
between and —9.2 km s '. However, any material at a 
velocity in this range should cause the H i absorption line to 
be redshifted relative to the D i line rather than blueshifted 
as observed. We measure the second absorption component 
of the H I absorption line to be at —27.8 km s~' (see Table 
3), inconsistent with the solar hydrogen wall. Furthermore, 
the temperature inferred for this component from the 
Doppler parameter (/>„! = 41.39 km s'') is 104,000 K, 
which is much hotter than the temperature measured for the 
solar hydrogen wall in Paper III. Since the velocity and 
temperature of the second component are inconsistent with 
the solar hydrogen wall, we instead attribute this com- 
ponent to an analogous hydrogen wall surrounding the 
star. The velocity of the second component is certainly con- 
sistent with this interpretation as is the temperature for 
reasons that we will present in § 6. 

Finally, we discuss the effects that multiple ISM com- 
ponents might have on our analysis. Unfortunately, we do 
not have observations of narrower ISM absorption lines of 
heavier atomic species, such as Mg ii and Fe ii, that would 
show whether multiple components exist. We believe, 
however, that even if there are multiple velocity com- 
ponents, a very hot component like the second component 
in Figure 9 must still be present to fit the observed absorp- 
tion profile. No combination of typical T k 7000 K LISM 
absorption components can explain the curved appearance 
of the blue side of the H i absorption line, and no alteration 
of the assumed stellar profile can explain this either without 
introducing unreasonable fine structure into the profile (see 
§ 3). However, a very hot T % 100,000 K absorption com- 
ponent with a low column density like that shown in Figure 
9 fits the data quite nicely. 



4.2. Analysis of the k And Data 

The "bisector method" for measuring Nhi could not 
have been used for either a Cen of CapeUa in Papers l-III, 
because for these lines of sight the velocity separations 
between the stellar emission and the interstellar absorption 
are small. For the same reason, we cannot use the bisector 
method in the analysis of the A And data. Therefore, for our 
inital two-component fit to the Lya profile of X And, we 
simply assume the same stellar profile that was used for the 
single-component fit in Figure 4. As we did for the two- 
component fits in §4.1, we force the velocities and tem- 
peratures of the H I and D i lines to be the same, and we 
require the D/H ratio to be the same for both components. 
Figure 10 shows the resulting fit. The dotted line shows the 
absorption only from the ISM component, and the dashed 
line represents the absorption from the second component. 
We would like to associate this second component with 
either the solar hydrogen wall or a stellar one. To see if this 
is possible, wc first determine if the velocity of the second 
component, —31 km s ', is consistent with such an inter- 
pretation. 

Our observations of / And were made at orbital phase 
<i) = 0.91, when the G8 IV HI star has a velocity of +3.3 
km s~' (Batten et al. 1978). However, if the solar analogy is 
any guide, it should take at least a year for stellar wind 
material to reach the hydrogen wall of X And, if it has one, 
and since the orbital period of /. And is only 20.5 days, we 
doubt that the hydrogen wall is affected by orbital motions. 
Thus, the ccntcr-of-mass velocity of the A And system, +6.8 
km s"', is the relevant velocity for estimating the velocity 
expected for a stellar hydrogen wall. The velocities of the 
solar and/or stellar hydrogen walls should not be too differ- 
ent from the velocity of the interstellar absorption ( + 6 km 
s"'), because both the Sun's velocity (0 km s"') and the 
center-of-mass velocity of the X And binary system ( + 6.8 
km s'') are not that different from this value. While we 



1166 



WOOD, ALEXANDER, & LINSKY 



Vol. 470 



\:d 






/"■' rA 




; X And 




\ i\ 


1.0 


^ 


/ 


\ /■ \ 






/ 


\ 


0.8 




/ 


\ i: 


0.6 


- 




■^ v; 






/' 






12M.5 1215.0 1215,5 1216.0 

Wavelength 



1216.5 



1217.0 



Fig. 10. — A two-component fit to the H l and D i absorption lines of X And, assuming the same stellar Lya profile (thin solid line) used in the 
single-component fit in Fig. 4. The data are shown in histogram form. The dotted line is the component representing absorption from the ISM, the dashed 
line is the second component for which we have no interpretation (see § 4.2), and the thick solid line is the combination of the two components. 



expect the second component to be blueshifted relative to 
the interstellar absorption component at -1-6 km s"', in 
order to explain why H i is blueshifted relative to D i (see 
Table 2), the blueshift of the second component in Figure 10 
( — 31 km s~ ') is clearly too large to be explained by a solar 
or stellar hydrogen wall. 

We tried another two-component fit in which we forced 
the velocity of the second component to be km s"S but 



the quality of that fit was not particularly good. We then 
experimented with diff'erent intrinsic stellar profiles to see 
whether we could find one that would lead to a two- 
component fit with a second component that we could 
attribute to a hydrogen wall. We found that broader stellar 
profiles designed to produce fits with smaller values for Nh i 
yielded promising results. Figure 1 1 shows the best of these 
fits. The parameters of this fit are given in Table 3. The 



_3 




12145 12150 1215.5 1216.0 

Wavelength 



1216.5 



1217.0 



Fig. 11. — Our best two-component fit to the H i and D [ absorption lines of X And, with the residuals displayed below the fit. The data are shown in 
histogram form. The thin solid line is the assumed stellar Lya profile. The dotted Une is the component representing absorption from the ISM, the dashed line 
is the component representing absorption from a hydrogen wall surrounding X And (see § 4.2), and the thick solid line is the combination of the two 
components. The parameters for the fit are given in Table 3. We consider this fit to be superior to the two-component fit in Fig. 10 because for this fit we can 
interpret the second absorption component as being due to a stellar hydrogen wall, which is not the case for the fit in Fig. 10. 



No. 2, 1996 



e IND AND A AND 



1167 



velocity of the second component is —2 km s~^ which is 
close enough to the solar rest frame to be consistent with 
the solar hydrogen wall. However, the Doppler parameter 
of this component (/) = 31.89 km s"^) suggests a tem- 
perature of 62,000 K, which is much hotter than the 29,000 
K temperature measured for the solar hydrogen wall in 
Paper III. In fact, at a temperature of 29,000 K, the solar 
hydrogen wall does not produce an absorption component 
broad enough to be detected, given the very broad H i 
absorption feature we observe in our A And data. Thus, we 
believe that the second component in Figure 1 1 must be due 
to a stellar hydrogen wall. The high temperature of the 
stellar wall component and the 8 km s"' blueshift of this 
absorption with respect to the interstellar absorption and 
A And rest frames have reasonable explanations that will 
be discussed in § 6. 

For e Ind, we argued that the hot second component is 
necessary to fit the data regardless of whether or not there 
are multiple ISM components along the line of sight. 
However, we cannot be sure that this is the case for X And. 
In contrast to e Ind, the H i line of A And is nicely fitted by a 
single ISM component, the only problem being that the 
parameters of H i and D i are inconsistent. It is not imme- 
diately apparent why the second absorption component for 
A And cannot be an ISM component with a much higher 
column density and much lower temperature than the 
second component in Figure 1 1. We tried to force our two- 
component fits to produce such a component. The resulting 
fits were not as good as that shown in Figure 11, but they 
were not too bad, and the discrepancies between the fits and 
the data were far more subtle than the discrepancies seen for 
the e Ind fits in Figures 3 and 8, for example. Nevertheless, 
the fit shown in Figure 9 is our best two-component fit for A 
And, and the parameters for the hot component do conform 
nicely to what we expect for a hydrogen wall around A And 
(see § 6), but these results must be considered tentative. If 
future observations reveal multiple ISM components for 
this hne of sight, these data should be reanalyzed with this 
in mind. 

4.3. Estimating Systematic Errors 

The errors cited in Table 3 are only the random errors of 
the fitting process and do not include estimates of system- 
atic errors, such as those that result from uncertainties in 
the adopted stellar Lya profiles. In an attempt to assess 
these systematic errors, we experimented with a large 
number of possible stellar profiles to determine which pro- 
files produce two-component fits that match the data well 
but have significantly different parameters than the fits 
shown in Figures 9 and 11. Deciding whether or not a fit 
matches the data well enough to be considered acceptable is 
somewhat subjective, but this is a necessary step in estimat- 



ing the uncertainties in the assumed stellar profiles, which 
are the dominant sources of systematic error in the analysis. 

For e Ind, the interstellar H i column density derived 
from the bisector analysis (log N^i = 18.0 + 0.1) provides 
an important guideline that can be used to separate the 
acceptable fits from the unacceptable. For A And, we 
required that the second component of the two-component 
fit be at a velocity consistent with a stellar hydrogen wall. 
As Figure 10 suggests, not all assumed stellar profiles 
resulted in fits that met this criterion. Although absorption 
from a hydrogen wall around a Cen was not really detected 
in Paper III, the possibility that this absorption is present 
and affects the data was considered in estimating errors for 
the fit parameters. Similarly, we experimented with three- 
component fits to the Lya line of c Ind to see how the 
addition of a component representing the solar hydrogen 
wall could change the derived parameters of the other two 
components. The results from Paper III were used to con- 
strain the parameters of the solar hydrogen-wall com- 
ponent. Since the temperature of the solar hydrogen wall 
measured in Paper III implies an absorption feature too 
narrow to have affected the very broad H i absorption line 
seen in the A And data, we did not experiment with three- 
component fits to the Lya fine of A And. 

In Table 4, we show the results of these experiments. This 
table summarizes the results of the two-component fit 
analysis, and the uncertainties in the fit parameters include 
our estimates of the systematic errors. 

5. THE PROPERTIES OF THE INTERSTELLAR MEDIUM 

The Va and Fag flow vectors of Lallement & Berlin (1992) 
predict LISM velocities for the e Ind Hne of sight of —8.5 
and —8.9 km s~ ', respectively. These two velocities are too 
close for us to determine whether e Ind is in the G cloud or 
in the AG cloud. Our measured velocity of —9.2 + 1.0 km 
s~ ' (see Table 4) is consistent with both velocities. Because e 
Ind is located in the general direction of the Galactic center, 
it would seem more likely that e Ind is in the G cloud. 
However, the LISM temperature that we have measured 
(8500 + 500 K) is much hotter than the temperature 
(5400 + 500 K) measured for the a Cen line of sight, which 
is definitely in the G cloud (Paper III). The e Ind tem- 
perature is also somewhat hotter than the temperatures of 
7000 ± 900 K and 6900 ± 380 K measured for the Capella 
and Procyon lines of sight, respectively, which are in the AG 
cloud (Paper II). Observations of the interstellar Mg 11 and 
Fe II lines are needed to verify that the Doppler parameter 
we have measured is indeed an accurate measure of the 
LISM temperature and has not been artificially broadened 
by the presence of more than one velocity component. 

The D/H ratio that we have measured for the e Ind Hne of 
sight, (1.6 + 0.4) X 10"', is in excellent agreement with the 



Star 



TABLE 4 
Summary of Quantitative Results (Systematic Errors Included) 



Velocity 
(km s ') 



(km s 



') 



T 

(10^ K) 



logN„, 



D/H 

(10') 



Source of Absorption 



€ Ind . . 

/I And. 



-9.2 + 1.0 


11.9 + 0.3 


8.5 + 0.5 


18.0 + 0.1 


1.6 + 0.4 


ISM 


-27 + 6 


41+4 


100 + 20 


14.2 + 0.2 


(1.6 + 0.4) 


Stellar H i wall 


+ 6.5 + 1.0 


13.8 + 0.3 


11.5 + 0.5 


18.45 + 0.15 


1.7 + 0.5 


ISM 


-2 + 3 


32+ 5 


62 + 18 


14.8 + 0.2 


(1.8 ± 0.5) 


Stellar H i wall 



1168 



WOOD, ALEXANDER, & LINSKY 



Vol. 470 



values measured in Papers I-III. This agrecmenl is very 
encouraging, since the H i column density was measured 
using an entirely different technique than in previous 
analyses namely the bisector method. The H i column 
density toward e Ind, log Nhi = 18.0 + 0.1, implies an 
average number density of «hi = 0.094 ± 0.022 cm '. This 
density is consistent with the densities measured toward 
Procyon and a Cen, which are both less than 4 pc from the 
Sun, as is e Ind. Note that the bisector analysis and our best 
two-component fit resulted in an H i column density a 
factor of 2 different from that computed in the single- 
component fit (log Nhi = 18.292). This illustrates that 
absorption components that are present but left out of the 
analysis can result in inaccurate column densities, regard- 
less of the quality of the data. 

The flow velocities predicted for the / And line of sight by 
the Vc and V^a vectors are +9.8 and +7.6 km s ', respec- 
tively. We expect the A And line of sight to be in the AG 
cloud on the basis of its Galactic coordinates, and the 
observed velocity of the ISM absorption, +6.5 + 1.0 km 
s ', is indeed close to the expected velocity. However, the 
temperature of 11,500 + 500 K that we have measured is 
substantially higher than the temperature of 7000 K mea- 
sured for the Capella and Procyon lines of sight, which are 
both predominantly in the AG cloud. This temperature is 
also higher than the 8500 K value that we have estimated 
from the e Ind data. For e Ind, we expressed concern that 
unresolved ISM components might be broadening the D i 
and H i absorption lines. This concern is even greater for / 
And, because the discrepancy with previous measurements 
is larger and because the longer (23 pc) line of sight to A And 
makes it far more likely that multiple components are 
present. If future observations should confirm that multiple 
components are present toward A And, our spectrum of the 
Lya line should be reanalyzed with this in mind, as we 
previously emphasized in § 4.2. It is of course possible that 
the average ISM temperature toward X And really is hotter 
than for the Capella and Procyon lines of sight. 

The H I column density listed in Table 4 for the / And 
line of sight, log N^i = 18.45 + 0.15, indicates an average 
density of «„! = 0.040 + 0.014 cm"^. This density is much 
lower than that found for the shorter (d < 4 pc) lines of sight 
toward a. Cen, Procyon, and e Ind, but it is very similar to 
the density measured toward Capella (d = 12.5 pc; «„, = 
0.0450 + 0.0006 cm"^). In Paper II, this difference was 
interpreted to mean that most of the Capella line of sight is 
not inside the local cloud but is instead within a very hot 
(T « 10* K) and tenuous phase of the ISM containing 
essentially no neutral hydrogen. This is a reasonable 
hypothesis, since the Sun is believed to be located within a 
region called the Local Bubble, in which most of the volume 
of the bubble consists of this very hot interstellar matter. It 
is this hot interstellar material that is thought to be 
responsible for the soft X-ray background (see, e.g.. Cox & 
Reynolds 1987). This interpretation of the Capella data is 
supported by Lemoine et al. (1995), who measured an iden- 
tical H I column density for this absorption component 
along a much longer line of sight only 7 from Capella (see 
§ 1). If the local cloud extended to and beyond Capella, 
Lemoine et al. (1995) should have measured a larger column 
density than that measured toward Capella. 

Does the low average H i density toward A And mean 
that /I And lies inside the very hot, 10* K phase of the ISM? 
We believe that this is unlikely, because hydrogen walls 



such as the one we believe we have detected around A And 
cannot exist unless the surrounding interstellar material 
contains a substantial amount of neutral hydrogen. Thus, 
although it is likely that a portion of the A And line of sight 
lies within the hot phase of the ISM, A And itself must He 
within a warm, neutral cloud like that which surrounds the 
Sun. It is also possible that the low ISM density we have 
measured toward A And is simply indicative of density gra- 
dients in the local cloud (and the cloud surrounding A And, 
if it is a different cloud) rather than the presence of very hot 
and completely ionized ISM material along this line of 
sight. 

We find that D/H = (1.7 ± 0.5) x 10'* for the A And 
data. Both the e Ind and A And data appear to support the 
value of D/H = 1.6 x 10"' established in Papers I III as 
the best estimate for the LISM. However, we consider the e 
Ind result to represent a more significant confirmation of 
this value, because the bisector method used to constrain 
possible values of iV„ , for € Ind is a quantitative method 
that produces a clear result. In contrast, the method used to 
constrain N„ , for A And relies on qualitative assumptions, 
namely the requirement that the second absorption com- 
ponent be interpretable as a hydrogen-wall component. The 
analyses in Papers 1 111 clearly show that the abundance 
ratio D i/Mg ii does vary significantly in the LISM, but 
difficulties in accurately measuring H i column densities 
have prevented us from determining whether it is D i or 
Mg u or both that are varying. Therefore, there is as yet no 
conclusive evidence from the GHRS studies of many lines of 
sight that D/H varies in the LISM, as many analyses of /L £ 
and Copernicus data have suggested (see, e.g., Duprce, 
Baliunas, & Shipman 1977). 

6. THE HYDROGIiN WALLS SURROUNDING e INDI AND / 
ANDROMEDAE 

To understand the properties of the hydrogen walls that 
we believe we have detected around e Ind and A And, it is 
necessary to evaluate the velocity and direction of the inter- 
stellar How in the rest frames of these stars. We assume that 
e Ind is in the G cloud and that A And is in the AG cloud. As 
discussed in § 4.1, it is possible that € Ind is actually in the 
AG cloud, but since the K^ and F^g vectors are very similar 
(see § 1), it does not matter too much which vector we use. 
The use of the I-^g vector to represent the ISM flow around 
A And is more questionable. At a distance of 23 pc, A And 
likely lies well outside the AG cloud. However, since the 
velocity of the interstellar absorption agrees very well with 
the velocity expected from the F^g vector, the flow vector 
for the cloud in which A And resides is probably similar to 
the Kag vector. 

The distances, radial velocities (v,), and proper motions 
ifi^ and /i^; Hirshfeld, Sinnott, & Ochsenbein 1991) of e Ind 
and A And are listed in Table 5. When combined with the 
flow vectors assumed above and the known positions of the 
stars, this information allows us to compute the interstellar 
flow vectors in the rest frames of the stars. The velocities ((„) 
and directions (/ and b) of these flow vectors are also listed 
in Table 5. These flow vectors may be compared with the 
Kag vector (Uq = 25.7 km s"', / = 186.1, h= - 16.4) that 
represents the interstellar flow in the solar rest frame. The 
interstellar wind velocity in the stellar rest frame Vq, is much 
higher for c Ind and A And than it is for the Sun. We believe 
that this is the reason that the hydrogen walls of € Ind and A 
And have higher temperatures than the solar hydrogen 



No. 2, 1996 



€ IND AND ;. AND 



1169 



TABLE 5 
Interstellar Wind Properties in the Rest Frames of € Ind and i And 



Star 


Cloud 


Distance 

(PC) 


(kms^') 


/'a 
(arcsec yr ') 


(arcsec yr ') 


(kms"') 


/ 
(deg) 


h 
(deg) 



(deg) 


€ Ind 

A And 


G 
AG 


3.54 
23 


-38.9 
+ 6.8 


+ 3.96 
+ 0.17 


-2.54 
-0.42 


64.0 
47.7 


36.2 
169.4 


-13.4 
+ 61.6 


60.3 
88.9 



wall. Figure 12 shows the hydrogen-wall temperatures as a 
function of interstellar wind velocity. Baranov & Malama 
(1995) and Pauls et al. (1995) predict a temperature of about 
20,000 K for the solar hydrogen wall. V. B. Baranov (private 
communication) informs us that if the Sun were traveling 
through the ISM at a velocity of 64.0 km s " ' like e Ind, his 
models would predict a hydrogen-wall temperature of 
about 80,000 K rather than 20,000 K. Thus, the dependence 
of temperature on wind velocity seen in Figure 12 is qualit- 
atively consistent with these models. 

One might think that hydrogen should be almost com- 
pletely ionized at these high temperatures. However, it is 
not collisions that heat neutral hydrogen but charge 
exchange processes. Interstellar protons are compressed 
and heated by their collisions with solar wind protons. 
Through charge exchange with these protons, the neutral 
hydrogen is also compressed and heated, creating the hot 
hydrogen wall. Since the collisional ionization mean free 
path for hydrogen atoms at very low ISM densities is 
extremely long (%0.1 pc|. hydrogen atoms heated in the 
hydrogen wall are not collisionally ionized there despite the 
high temperatures. 

On the basis of the models of the heliosphere, we expect 
the solar hydrogen wall to have different properties for dif- 
ferent lines of sight. In Paper 111, the solar hydrogen wall 
was observed toward a Cen at an angle of <? = 52 ' from the 
upwind direction. The last column of Table 5 lists the angles 
(relative to the upwind direction as seen from the star) 
observed through the hydrogen walls of € Ind and X And. 
Fortunately, these angles (60 and 89 ) are not drastically 
different from the solar angle observed in Paper III, allow- 
ing us to compare results as we do in Figure 12. 

The column density of the k And hydrogen wall, logN^, , = 
14.8 + 0.2, is very similar to the column density measured 
for the solar hydrogen wall in Paper III (log N^, = 14.74 



1^ 



<>t Ind 



<> A And 



-L 4 



^^ 



<|> Sun 



^0 40 60 80 

Inler.slellar Wind Velocity (km s" ) 

Fig. 12. — Hydrogen-wall temperature as a function of the interstellar 
wind velocity in the stellar rest frame. 



± 0.24). The hydrogen wall column density for e Ind, 
however, is much lower log N„|= 14.2 + 0.2. Perhaps 
the high interstellar wind velocity for e Ind has resulted in a 
thinner hydrogen wall. 

We expect the hydrogen walls to generally have velocities 
between the stellar radial velocity and the velocity of the 
ISM absorption, because the hydrogen-wall material is 
expected to be decelerated relative to the unperturbed inter- 
stellar wind. Baranov & Malama (1993) predict that i.. ;% 
0.5i'o in the upwind (i.e., fl = ) direction and r, -t 0.75t!„ at 
6 = 90 \ where v^ is the velocity component in the direction 
of the unperturbed flow. Away from the upwind direction, 
the hydrogen-wall material can also have a velocity com- 
ponent, i)^, perpendicular to the original flow direction as 
the interstellar flow is deflected around the heliopause (see 
Fig. 1 ). For e = 90 , Baranov & Malama (1993) predict that 

t'x « 0.21=0 • 

To estimate the expected hydrogen wall velocity for e Ind, 
we crudely interpolate the results of Baranov & Malama 
quoted above to = 60 and assume v^ = 0.65(;u and (\. = 
O.lSt'o- The expected velocity is then simply v = i\ + t'. 
cos d — 1-'^ sin = —21 km s '. This velocity agrees per- 
fectly with the measured velocity in Table 4 of — 27 + 6 km 
s~'. Since 6 is essentially 90 for ?. And, the t;_, velocity 
component has no projection along the line of sight, and 
therefore it is the i\ component that is responsible for the 
observed blueshift of the hydrogen wall with respect to the 
interstellar absorption and the stellar radial velocity noted 
in § 4.2. For A And, the predicted velocity is —3 km s~', 
which agrees well with the measured velocity listed in Table 
4(-2 + 3kms-'). 

Because the parameters of the hot components seen 
toward e Ind and A And agree very well with the expected 
properties of hydrogen walls around these stars, we believe 
the hydrogen wall interpretation is the best interpretation 
for the hot components. Nevertheless, there is no way we 
can prove that the material responsible for the absorption 
in these components is in fact surrounding the stars. Other 
interpretations of the hot components cannot be completely 
ruled out. For example, the local warm neutral cloud is 
embedded within a very hot (^lO* K), tenuous ISM. If 
there is an evaporative interface between the local cloud 
and this hot ISM (see, e.g., Slavin 1989) along the hues of 
sight to c Ind and A And, the hot components might poss- 
ibly be associated with this interface. For the line of sight 
toward Sirius, Berlin et al. (1995) claim to have detected a 
hot H I absorption component that is redshifted relative to 
the interstellar absorption. They propose that this absorp- 
tion component is due to an evaporation flow from the local 
cloud into the hot ISM at an evaporative interface along 
this line of sight. We point out, however, that the hot com- 
ponents seen toward e Ind and A And are definitely blue- 
shifted relative to the interstellar absorption, making it 
much harder to associate these components with an evapo- 
rative interface. 



1170 



WOOD, ALEXANDER, & LINSKY 



Vol. 470 



7. THE STELLAR WINDS OF € INDI AND X ANDROMEDAE 

In § 4.2, we noted that the presence of neutral hydrogen 
in the ISM is necessary for the formation of a hydrogen 
wall, meaning e Ind and k And must be inside warm, neutral 
clouds. Another necessary ingredient for the formation of 
hydrogen walls is a stellar wind, the implication being that e 
Ind and X And must have previously undetected winds. To 
our knowledge, no one has ever detected a solar-like wind 
around a star before. Other types of stellar winds have been 
detected and studied extensively : for example, the hot winds 
of OB stars that are driven by radiation pressure, the cool 
winds of red giant and supergiant stars, and the winds of 
pre-main-sequence stars such as T Tauri and Herbig Ae 
stars. However, all these winds are fundamentally different 
from the solar wind in many ways. The wind acceleration 
mechanisms are almost certainly different, and the mass- 
loss rates of these winds are orders of magnitude higher 
than for the solar wind. Also, the stars generating these 
winds are very different from the Sun. 

On the basis of what is known about the origins of the 
solar wind, there is no reason to believe that stars like e Ind 
and X And do not have similar winds. In the original solar 
wind model of Parker (1958), the solar wind is driven by the 
thermal expansion of material heated to coronal tem- 
peratures (T w 2 X 10* K). Therefore, all stars that have 
hot coronae are expected to have solar-like winds. Coronal 
X-ray emission has been detected from both e Ind and I 
And. The X-ray luminosity of c Ind (Lx = 1.6 x 10^'' ergs 
s"'; Wood et al. 1994) is very similar to that of the Sun, 
implying that e Ind and the Sun have similar coronal 
properties. The X-ray luminosity of k And, on the other 
hand (Lx = 3.7 x 10^° ergs s"'; Dempsey et al. 1993a), is 
over 1000 times larger than that of the Sun. The tem- 
perature of k And's corona is also significantly higher 
(Dempsey et al. 1993b), implying a much more active 
corona than the Sun, and perhaps different wind properties. 

Detecting a solar-like wind emanating from a star is 
much more difficult than detecting the corona that presum- 
ably provides the acceleration mechanism. Hot, ionized 
winds should be sources of free-free emission, so nonde- 
tections of radio emission from stars have allowed upper 
limits to be placed on the mass-loss rates of these stars. For 
example, Brown et al. (1990) list upper limits for mass-loss 
rates of many A- and F-type stars, the lowest upper limit 
being M<7 x lO"'' Mq yr"' for Procyon (F5 IV-V). 
Lim, White, & Slee (1996) place an upper limit of M < 7 
X 10"'^ Mq yr"' on the mass-loss rate of Proxima Cen 
(M5.5 Ve). However, these upper fimits are more than 2 
orders of magnitude higher than the solar mass-loss rate of 
M = 2 X 10"'* Mq yr'MFeldman et al. 1977), illustrating 
just how difficult it is to detect such winds. The creation of 
"hydrogen walls" when these winds collide with the ISM 
provides indirect evidence for such winds. With the detec- 
tion of hydrogen walls around e Ind and k And, perhaps we 
can now add the solar wind to the long list of solar pheno- 
mena that have been detected on other stars. 

8. SUMMARY 

We have observed the Lya lines of e Ind and k And using 
the GHRS instrument aboard the HST. Our analysis of the 
interstellar absorption features present in these lines reveals 
that the velocities and temperatures of the D i lines appear 
to be different from those of the H i lines. A similar discrep- 



ancy was seen in Paper III for the Lya lines of a Cen A and 
B. In Paper III, the discrepancy was resolved by fitting the 
H I fine with two components, one representing absorption 
from the ISM and the other representing absorption from 
the hydrogen wall surrounding the Sun. This hydrogen wall 
has been predicted by models of the hefiospheric interface 
region where the solar wind interacts with the ISM. 

We have also fitted the H i fines of e Ind and k And with 
two components in order to resolve the discrepancy 
between D i and H i. The results of these two-component 
fits are summarized in Table 4. We found, however, that the 
parameters of the second components for e Ind and k And 
were inconsistent with the solar hydrogen wall. The tem- 
peratures are too high, and for e Ind the velocity is clearly 
wrong. However, the parameters of these second com- 
ponents are consistent with those expected for hydrogen 
waUs surrounding the stars. The hydrogen walls of e Ind 
and k And are expected to be hotter than the solar hydro- 
gen wall because of the higher interstellar wind velocities in 
the stellar rest frames. In recent models of the hefiospheric 
interface region, hydrogen walls are a natural consequence 
of the interaction between an ionized wind and a partially 
ionized interstellar medium. Thus, the detection of hydro- 
gen wafis around e Ind and k And represents a detection of 
winds emanating from these stars. We believe this to be the 
first detection, albeit indirect, of solar-like winds around 
stars other than the Sun. 

As for the measured parameters of the LISM listed in 
Table 4, the H i column density measured for the e Ind line 
of sight implies an average density, Oh i = 0.094 ± 0.022 
cm"', that is consistent with previous measurements made 
for other short {d<A pc) fines of sight. The H i column 
density was not measured by the two-component fit alone. 
Instead, we used a promising new technique for estimating 
TVh , that is feasible when there is a large velocity separation 
between the stellar Lya emission line and the interstefiar 
absorption. The advantage of this technique is that it does 
not require one to estimate the shape of the intrinsic stellar 
Lya profile. This technique could not be used for k And, but 
our requirement that the second component of the two- 
component fits be interpretable as either a solar or steUar 
hydrogen wafi component resulted in substantial con- 
straints on the shape of the assumed Lya profile and on N^^ , 
(see Table 4). 

The column density we have measured toward k And 
impfies a LISM density, «„! = 0-040 + 0.014 cm"^ that is 
lower than densities measured along much shorter fines of 
sight, implying either a density gradient in the LISM or 
very hot, ionized ISM material along much of this fine of 
sight. The LISM temperature listed for the k And line of 
sight in Table 4, T = 11,500 ± 500 K, is significantly higher 
than previous measurements for the LISM. Considering the 
length of this line of sight, this measurement could be the 
result of multiple ISM components rather than high tem- 
peratures. Observations of narrower interstellar absorption 
lines, like those of Mg ii and Fe ii used in Papers I II, are 
needed to detect multiple components. If future obser- 
vations prove that multiple ISM cmponents exist for the k 
And line of sight, the Lya line of k And should be rea- 
nalyzed with this in mind, and the existence of the hydrogen 
wall absorption component for this line of sight should be 
reassessed. The D/H measurements listed for € Ind and k 
And in Table 4 agree very well with the value of D/H = 
1.6 X 10 * established in Papers I-III as the best esti- 



No. 2, 1996 



€ IND AND A AND 



1171 



mate for D/H in the LISM. Thus, the conclusions in Papers 
I-II regarding this value and its implications for cosmology 
remain valid. 

We would like to thank V. B. Baranov for providing us 
with useful information about the models of the helio- 
spheric interface region, and F. Bruhweiler for helpful sug- 



gestions. This work is supported by NASA Interagency 
Transfer S-56460-D to the National Institute of Standards 
and Technology. W. R. A. also acknowledges the support of 
NASA grant GO-0100.01-92A from the Space Telescope 
Science Institute, which is operated by the Association of 
Universities for Research in Astronomy Inc., under NASA 
contract NAS 5-26555. 



REFERENCES 



Baliunas, S. L., & Dupree, A. K. 1979, ApJ, 227, 870 

Baranov, V. B., & Malama, Y. G. 1993, J. Geophys. Res., 98, 15,157 

. 1995, J. Geophys. Res., 100, 14,755 

Batten, A. H., Fletcher, J. M., & Mann, P. J. 1978, Publ. Dom. Astrophys. 

Obs. Victoria, 15, No. 5 
Berlin, P., Vidal-Madjar, A., Lallement, R., Ferlet, R., & Lemoine, M. 1995, 

A&A, 302, 889 
Bevington, P. R., & Robinson, D. K. 1992, Data Reduction and Error 

Analysis for the Physical Sciences (New York: McGraw-Hill) 
Brown, A., Veale, A., Judge, P., Bookbinder, J. A., & Hubeny, I. 1990, ApJ, 

361,220 
Buscombe, W., & Kennedy, P. M. 1968, MNRAS, 139, 341 
Cox, D. P., & Reynolds, R. J. 1987, ARA&A, 25, 203 
Dempsey, R. C, Linsky, J. L., Fleming, T. A., & Schmitt, J. H. M. M. 

1993a, ApJS, 86, 599 
Dempsey, R. C, Linsky, J. L., Schmitt, J. H. M. M., & Fleming T. A. 1993b, 

ApJ, 41 3. 333 
Dupree, A. K., Baliunas, S. L., & Shipman, H. L. 1977, ApJ, 218, 361 
Feldman, W. C, Asbridge, J. R., Bame, S. J., & Gosling, J. T. 1977, in The 

Solar Output and Its Variations, ed. O. R. White (Boulder: Colorado 

Assoc. Univ. Press), 351 
Ferlet, R., Lecavelier, A., Vidal-Madjar, A., Bertin, P., Deleuil, M., Gry, C, 

Langrange- Henry. A.-M., & Lallement, R. 1995, A&A, 297, L5 
Gilliland, R. L. 1994, GHRS Instrument Science Report 063 (Baltimore: 

Space Telecope Science Instutute) 
Gliese, W., & Jahreiss, H. 1991, Third Catalogue of Nearby Stars 

(VerofTentlichungen des Astronomischen Rechen-Instituts, University of 

Heidelberg) 
Gry, C, Lemonon, L., Vidal-Madjar, A., Lemoine, M., & Ferlet, R. 1995, 

A&A, 302, 497 
Herbst, W. 1973, A&A, 26, 137 
Hirshfeld, A., Sinnott, R. W., & Ochsenbein, F. 1991, Sky Catalogue 

2000.0, Vol. 1 (Cambridge: Cambridge Univ. Press) 
Jenkins, L. F. 1952, General Catalog of Trigonometric Parallaxes (New 

Haven: Yale Obs.) 



Lallement, R., & Bertin, P. 1992, A&A, 266, 479 

Lallement, R., Bertin, P., Ferlet, R., Vidal-Madjar, A., & Bertaux, J. L. 

1994, A&A, 286, 898 
Lallement, R., Ferlet, R., Lagrange, A. M., Lemoine, M., & Vidal-Madjar, 

A. 1995, A&A, 304,461 

Lemoine, M., Vidal-Madjar, A., Ferlet, R., Bertin, P., Gry, C, & Lallement, 
R. 1995, in The Light Element Abundances, ed. P. Crane (Berlin: 
Springer), 233 

Lim, J. White, S. M., & Slee, O. B. 1996, ApJ, 460, 976 

Linsky, J. L., et al. 1993, ApJ, 402, 694 (Paper I) 

Linsky, J. L., Diplas, A., Wood, B. E., Brown, A., Ayres, T. R., & Savage, 

B. D. 1995, ApJ, 451 , 335 (Paper II) 

Linsky, J. L., & Wood, B. E. 1996, ApJ, 463, 254 (Paper HI) 

McClintock, W., Henry, R. C, Linsky, J. L., & Moos, H. W. 1978, ApJ, 225, 

465 
Morton, D. C. 1991, ApJS, 77, 1 19 
Murthy, J., Henry, R. C, Moos, H. W., Vidal-Madjar, A., Linsky, J. L., & 

Gry,C. 1990, ApJ, 356, 223 
Parker, E. N. 1958, ApJ, 128, 664 
Pauls, H. L., Zank, G. P., & Williams, L. L. 1995, J. Geophys. Res., 100, 

21,595 
Slavin,J. 1989,ApJ,346,718 
Strassmeier, K. G., Hall, D. S., Zeilik, M., Nelson, E., Eker, Z., & Fekel, 

F.C. 1988,A&AS,72,291 
Walker, T. P., Steigman, G., Schramm, D. N., Olive, K. A., & Kang. H.-S. 

1991, ApJ, 376, 51 
Witte, M., Rosenbeuer, H., Banaszkewicz, M., & Fahr, H. 1993, Adv. Space 

Res., 13(6), 121 
Wood, B. E., Brown, A., Linsky, J. L., Kellett, B. J., Bromage, G. E., 

Hodgkin, S. T., & Pye, J. P. 1994, ApJS, 93, 287 
Wood, B.E., Harper, G. M., Linsky, J. L., & Dempsey, R. C. 1996, ApJ, 458, 

761