A&A 489, 1271-1289 (2008)
DOI: 10.1051/0004-6361:200809869
T. Tsuji
Institute of Astronomy, School of Sciences, The University of Tokyo, Mitaka, Tokyo 181-0015, Japan
Received 29 March 2008 / Accepted 27 May 2008
Abstract
Aims. We determine carbon, oxygen, and their isotopic abundances based on CO and OH spectra in 23 red giant stars, and identify possible origin of difficulty in abundance analysis of cool luminous stars.
Methods. We apply the line-by-line analysis based on the classical micro-turbulent model and 1D model photospheres.
Results. We found empirically that there is a critical value of log
(W is the equivalent width and
the wavenumber) above which the observed lines do not follow the classical line formation theory based on the micro-turbulent model and that the classical abundance analysis can be applied only to the weak lines of log
.
The carbon, oxygen, and their isotopic abundances in 23 K-M giant stars obtained from such weak lines are approximately consistent with evolutionary models, although the 12C/13C puzzle (observed 12C/13C ratios are too small compared with theoretical predictions) remains unsolved. It is already known that the strong lines of log
are contaminated by the contribution from outer molecular layers. The less strong but saturated lines of
(intermediate-strength lines) cannot be understood at all with the abundance and turbulent velocity based on the weak lines. By studying the behavior of these lines and considering other observations such as the detection of H2O lines, not only in the late M giants but also in the early M and K giants, we show that the intermediate-strength lines are essentially identical to the strong lines in that they also include contamination from the extra molecular layers. Thus, the behavior of the intermediate-strength lines, including those with LEP as high as 2 eV, appears to be nothing but a manifestation of the warm molecular envelope or MOLsphere.
Conclusions. The infrared spectra of K-M giant stars are a hybrid of at least two components originating in the photosphere and MOLsphere. Only the weak lines, mostly high excitation, are relatively free from the effect of MOLsphere and can be used to extract photospheric abundances. The strong lines and the intermediate-strength lines, which dominate the observed infrared spectra, are affected badly by contamination from the MOLsphere. For this reason, they provide little information about the photosphere, but instead can be important proves of the warm molecular envelope for which little is known yet. In the interpretation and analysis of the infrared spectra of K-M giant stars, it is essential to consider their hybrid nature.
Key words: stars: abundances - stars: atmospheres - stars: mass-loss - stars: late-type - line: formation - infrared: stars
The basic principle of stellar abundance determination was established almost half a century ago (e.g. Unsöld 1955), but high quality spectroscopic data have become available to complete accurate abundance analyses only from a far later date. Abundance determinations have been done primarily using classical methods based on grids of model photospheres (e.g. Gustafsson et al. 1975; Tsuji 1978; Kurucz 1979; Plez et al. 1992). Local thermodynamical equilibrium (LTE) was assumed to hold in many applications (e.g. Unsöld 1955) and favored for its simplicity in abundance determinations. The next major step was to treat line formation without the restriction of the LTE assumption (e.g. Mihalas 1978). Classical photosphere models study only the structure in the radial direction, that is are one-dimensional; the limitations of these models have been challenged by authors who considered the hydrodynamics of convection in 3 dimensions, that is three-dimensional models (e.g. Asplund 2005). In the regime of the coolest stars such as red giant and supergiant stars, however, the situation may be a bit different and there may be another problem.
In cool giant stars representing evolution from the red giant branch (RGB) to the asymptotic giant branch (AGB), CNO and their isotopic abundances experience drastic changes and accurate abundance determination should be vital not only to testing the stellar evolutionary models but also to understanding the chemistry of interstellar matter, which is enriched by the mass-loss from red giant stars. Despite their importance, few abundance analyses of high resolution spectra have been carried out for cool giant stars, in contrast to studies of other stellar types. It should be remembered that high quality infrared spectra needed for this purpose have been available, thanks to the Fourier transform spectroscopy (FTS) pioneered by P. & J. Connes (e.g. Connes 1970) and developed further in the 1970's (e.g. Maillard 1974; Hall et al. 1979; Ridgway & Brault 1984), even before highly efficient detectors such as CCD innovated optical spectroscopy. Our obstacles were therefore not in the quality of the observational data but difficulties that are inherent in the spectral analysis of very cool stars. Some efforts were made to determine the CNO and their isotopic abundances for the oxygen-rich case (e.g. Harris & Lambert 1984; Smith & Lambert 1985, 1986, 1990; Harris et al. 1985; Tsuji 1986, 1991; Aoki & Tsuji 1997) as well as the carbon-rich case (e.g. Lambert et al. 1984; Harris et al. 1987; Ohnaka & Tsuji 1996, 1999; Abia & Isern 1997; Ohnaka et al. 2000; Wahlin et al. 2006; Abia et al. 2008). Although many interesting results were shown by these works, we were not convinced entirely that the stellar abundance analysis could be completed for cool giant stars.
In our previous analysis of the CO first overtone bands in M giant stars, we found that most lines could be interpreted consistently with the classical model in terms of the micro-turbulent velocity and carbon abundance (Tsuji 1986, hereafter Paper I). The strong lines of low excitation could not be included in this scheme, but it was demonstrated that these strong lines were contaminated by the contributions of outer molecular layers (Tsuji 1988, hereafter Paper II). We then analyzed the CO second overtone bands of the same sample of M giant stars and found that the resulting carbon abundances appear to be higher by about a factor of 2 compared with those based on the CO first overtone bands (Tsuji 1991, hereafter Paper III). The origin of these discrepant results remains unsolved. The difficulty in abundance analysis of very cool stars was also noticed by Smith & Lambert (1990), who found that the strong OH 1-0 and 2-0 bands correspond to unreasonably high oxygen abundances compared with those based on the weaker bands and that strong CO lines cannot be used for abundance analysis.
The carbon and oxygen abundances in other stars are by no means well
established. Even for the Sun, the classical values widely adopted
(e.g. Anders & Grevesse 1989) were challenged by a new approach based
on 3D models; a reduction in carbon and oxygen abundances by
more than 50% from the classical values was proposed (Allende Pietro
et al. 2002). The classical values, however, were defended by a
detailed analysis of CO lines with semi-empirical model solar
photospheres by Ayres et al. (2006).
These results imply that the accurate modeling of photospheric
structure is a critical issue in abundance determinations.
It is unknown whether the difficulties in abundance
analyses in M giant stars described above are due to limitations in
the classical 1D model photospheres for cool giant stars or
to other problems. For example, spectra of red giant stars have
unexpected features such as H2O lines, not only for the early
M giants and K5 giant Tau (Tsuji 2001) but also the earlier
K giant star
Boo (Ryde et al. 2002). These features cannot
be reproduced by our present models of the photospheres of red giant
stars, and imply that we have a lack of understanding of the
photosphere and/or the atmosphere extending beyond the photosphere.
For this reason, we analyze the nature of the difficulty in
interpreting the spectra of cool giant stars in some detail
in this paper.
Our sample consists of one K giant and 22 M giant stars listed in
Table 1 and shown in Fig. 1. We use high resolution infrared
spectra observed with the FTS of KPNO (Hall et al. 1979).
Spectra of the H and K band regions are taken
from the KPNO archive, and we also observed some spectra of the H and K band regions and most spectra of the L band region.
Most of these spectra were also used in our previous analyses of CO
(Papers I-III), SiO (Tsuji et al. 1994), and NH & CN (Aoki &
Tsuji 1997). Some details of the spectra including the resolution,
S/N ratio, and observing dates are summarized in Table 2. We measured
equivalent widths (EWs) of isolated lines of 12C16O,
13C16O, and 12C17O in the H and K band
regions, and those of 16OH in the H and L band regions.
The logarithms of
(W is the measured equivalent width and
is the wavenumber of a line) are given in Tables 3a-3w.
Obj(BS/HD) | Sp. type |
![]() |
![]() |
![]() |
K | mag | ![]() |
||
![]() |
K5+III | 3874 | -1.7 ![]() |
1.5 ![]() |
![]() |
M0.5III | 3790 | -2.2 ![]() |
1.6 ![]() |
![]() |
M1III | 3812 | -2.2 ![]() |
1.7 ![]() |
![]() |
M1.5IIIa | 3909 | -3.2 ![]() |
3.6 ![]() |
![]() |
M2.5III | 3596 | -3.4 ![]() |
2.2 ![]() |
![]() |
M2.5III | 3852 | -3.4 ![]() |
3.7 ![]() |
![]() |
M2.5II-III | 3603 | -3.3 ![]() |
2.2 ![]() |
![]() |
M3+IIIa | 3712 | -2.9 ![]() |
2.0 ![]() |
![]() |
M3III | 3643 | -3.3 ![]() |
2.3 ![]() |
![]() |
M3III | 3643 | -2.4 ![]() |
1.4 ![]() |
10 Dra (5226) | M3.5III | 3730 | -2.9 ![]() |
2.1 ![]() |
![]() |
M4II | 3523 | -4.1 ![]() |
3.2 ![]() |
BS6861 (6861) | M4 | 3600 | -5.2 ![]() |
6.3 ![]() |
![]() |
M4II | 3420 | -5.5 ![]() |
5.5 ![]() |
RR UMi (5589) | M4.5III | 3397 | -3.4 ![]() |
1.6 ![]() |
![]() |
M5Ib-II | 3293 | -5.8 ![]() |
5.0 ![]() |
OP Her (6702) | M5II | 3325 | -4.4 ![]() |
2.3 ![]() |
XY Lyr (7009) | M5II | 3300 | -5.1 ![]() |
3.7 ![]() |
R Lyr (7157) | M5III | 3313 | -4.3 ![]() |
2.1 ![]() |
RZ Ari (867) | M6III | 3341 | -3.5 ![]() |
1.5 ![]() |
30g Her (6146) | M6-III | 3298 | -4.2 ![]() |
2.0 ![]() |
SW Vir (114961) | M7-III: | 2886 | -4.4 ![]() |
1.3 ![]() |
RX Boo (126327) | M7.5-8 | 2850 | -4.6 ![]() |
1.3 ![]() |
a
Based on the infrared flux method. The
uncertainty in
![]() ![]() |
![]() |
Figure 1:
Program stars shown by filled circles on the HR diagram
and recent evolutionary tracks (Claret 2004) shown by grey lines.
Stellar masses (in unit of solar mass ![]() |
We apply classical 1D model photospheres that are almost identical
to those adopted in our previous works (Tsuji 1978; Papers I-III),
apart from the fact that we
now assume that the photosphere is spherically symmetric (SS) rather than
plane-parallel (PP) and some molecular opacity data were updated (Tsuji 2002).
We include 34 elements in the solar composition (Table 1 of Tsuji 2002)
except for C, N, and O for which we assume two cases a and b listed in Table 4.
Case a represents the mean C and N abundances of early M giants and case b of
late M giants based on our previous analyses (Paper III;
Aoki & Tsuji 1997). The masses, radii, effective temperatures, and log g of our model
photospheres are listed in Table 5
.
We consider masses of one and two solar masses to study the
effects of mass. However, we use the models with
in the
following analyses, since this appears to be closer
to the masses of most of the red giant stars we study (Table 1).
Also, the changes in the photospheric structure are rather minor if
the masses are increased beyond 2
,
since the differences
between the SS and PP models tend to be reduced for larger masses.
We assume that the micro-turbulent velocity of the models is 3 km s-1
at all times.
Some characteristics of the resulting models are given in Table 5,
that is the surface temperature T0 at
(
is the
optical depth defined by the continuous opacity at
m) and extension parameter defined by
.
Some of our models are compared with those by Plez (1992)
in Fig. 2. These models are based on similar assumptions apart from that
Plez et al. (1992) applied the opacity sampling method while we used
the band model method in taking the blanketing effect of molecular lines
into account. The agreements are in general good apart from at
the very surface of the coolest model.
In evaluating the flux, we calculate spectra at a sampling interval
of 0.05 cm-1 using a detailed line list including CO (Guelachivili
et al. 1983; Chackerian & Tipping 1983), OH (Jacquinet-Husson et al. 1999), CN
(Cerny et al. 1978; Bausclicher et al. 1988), H2O (Partridge & Schwenke 1997), and SiO (Lavas et al. 1981; Tipping &
Chackerian 1981). Based on the resulting spectra, the flux
in the L band of the wide-band photometry, FL, and log RL,
where
with the bolometric flux
(
:
Stefan-Boltzmann
constant), are obtained (Table 5).
Case a | Case b | |
log
![]() |
8.35 | 8.08 |
log
![]() |
8.11 | 8.45 |
log
![]() |
8.69 | 8.69 |
a Tsuji (1991); b Aoki & Tsuji (1997); c Allende Prieto et al. (2002).
Case | R |
![]() |
log g | T0a | db | log FLc | log RLd |
![]() |
K | K | |||||
M = | ![]() |
||||||
a | 50 | 4000 | 1.04 | 2196 | 1.06 | 12.206 | -2.542 |
a | 50 | 3900 | 1.04 | 2185 | 1.08 | 12.193 | -2.572 |
a | 75 | 3800 | 0.69 | 2117 | 1.13 | 12.185 | -2.609 |
a | 75 | 3700 | 0.69 | 2094 | 1.13 | 12.171 | -2.641 |
a | 100 | 3600 | 0.44 | 2023 | 1.18 | 12.160 | -2.679 |
b | 100 | 3600 | 0.44 | 2074 | 1.18 | 12.159 | -2.677 |
b | 100 | 3500 | 0.44 | 2037 | 1.18 | 12.144 | -2.711 |
b | 150 | 3400 | 0.09 | 1857 | 1.30 | 12.137 | -2.755 |
b | 150 | 3300 | 0.09 | 1628 | 1.29 | 12.118 | -2.787 |
b | 200 | 3200 | -0.16 | 1396 | 1.40 | 12.102 | -2.825 |
b | 200 | 3100 | -0.16 | 1291 | 1.38 | 12.064 | -2.842 |
b | 250 | 3000 | -0.36 | 1079 | 1.51 | 12.031 | -2.866 |
b | 250 | 2900 | -0.36 | 1021 | 1.47 | 11.974 | -2.868 |
b | 300 | 2800 | -0.52 | 895 | 1.58 | 11.931 | -2.886 |
M = |
![]() |
||||||
a | 50 | 4000 | 1.34 | 2251 | 1.04 | 12.200 | -2.536 |
a | 50 | 3900 | 1.34 | 2234 | 1.04 | 12.187 | -2.566 |
a | 75 | 3800 | 0.99 | 2172 | 1.06 | 12.178 | -2.602 |
a | 75 | 3700 | 0.99 | 2152 | 1.06 | 12.163 | -2.634 |
a | 100 | 3600 | 0.74 | 2089 | 1.08 | 12.152 | -2.670 |
b | 100 | 3600 | 0.74 | 2142 | 1.08 | 12.151 | -2.668 |
b | 100 | 3500 | 0.74 | 2107 | 1.08 | 12.135 | -2.702 |
b | 150 | 3400 | 0.39 | 2003 | 1.13 | 12.124 | -2.742 |
b | 150 | 3300 | 0.39 | 1884 | 1.12 | 12.104 | -2.773 |
b | 200 | 3200 | 0.14 | 1652 | 1.17 | 12.083 | -2.806 |
b | 200 | 3100 | 0.14 | 1525 | 1.16 | 12.041 | -2.819 |
b | 250 | 3000 | -0.06 | 1342 | 1.20 | 11.998 | -2.833 |
b | 250 | 2900 | -0.06 | 1234 | 1.19 | 11.936 | -2.830 |
b | 300 | 2800 | -0.22 | 1102 | 1.22 | 11.882 | -2.837 |
a Temperature at
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
![]() |
Figure 2:
Some of our model photospheres (solid lines) are compared with those by
Plez et al. (1992) (dots).
![]() ![]() ![]() |
The stellar mass has a significant effect on the surface
temperature T0 and the extension parameter d especially
at low values of
,
and we examine its effects on abundance
determinations (Sects. 4.3 and 5.3). The calculated
values of log RL differ only slightly, however, from our
previous values (Tsuji 1981), which would be expected because the L band
region is relatively free from the heavy line-blanketing effect.
For
K, we compare the results for cases a
and b, but the differences are rather minor (Table 5).
For our sample of 23 red giant stars,
the effective temperatures are determined by applying the
infrared flux method (Blackwell et al. 1980) using the log RL
values from Table 5 and observed values of log RL (Tsuji
1981). The resulting values of
agree with our previous
results (Tsuji 1981) in general to within 50 K. Our
scale is also consistent with the empirical ones based on the
measured angular diameters by Ridgway et al. (1980) and more
recent results (e.g. Dyck et al. 1998; Perrin et al. 1998;
van Belle et al. 1999; Mondal & Chandrasekhar 2005).
We found that our individual values of effective temperature
agree with the empirical values of these authors, in most cases, to
within about 100 K; we therefore assume an error bar of 100 K for our
estimations of
.
With the Hipparcos parallaxes
(ESA 1997)
and the bolometric luminosities obtained by the integration of the
SEDs (Tsuji 1981), the absolute bolometric magnitudes are determined.
The uncertainty in
is dominated by errors in the parallaxes.
The errors in the photometric data are estimated to be about
0.1 mag, including absolute calibration errors of a few percent (Bersaneli
et al. 1991).
Our sample is shown on the HR diagram (Fig. 1) and an evolutionary mass
is estimated by comparison with the evolutionary tracks by
Claret (2004). Unfortunately, nearly half of our sample cannot be
accounted for by the models of Claret, and we have to extrapolate the
evolutionary tracks to be able to derive the stellar masses given in
Table 1. The error bars of the masses are derived from the errors
in
and
.
For objects with
lower than
3000 K, the extrapolations needed are so significant that the
uncertainties should be much larger, and we hope that theoretical
evolutionary tracks can be extended to describe these objects.
Given a model photosphere characterized by a set of mass, radius,
and luminosity (or reduced to
and log g
in PP models), a given EW can be interpreted in terms of abundance and micro-turbulent
velocity within the framework of the classical theory of line
formation. Abundances of 34 elements are assumed in
our model photosphere, and we in fact determine corrections
to the abundances assumed. In our analysis, we apply the line-by-line
analysis based on the classical micro-turbulent model.
In this method, we plot the abundance correction
log Aobtained for an assumed value of
against the
log
of the line. We expect that
larger (smaller) abundance corrections are needed if lower (higher) values
of the micro-turbulent velocity are assumed especially for more
saturated lines, while abundance corrections are independent of the
assumed
value for very weak lines free from saturation effect.
The micro-turbulent velocity
is determined so that
abundance corrections for individual lines show the least dependence
on EWs (as for detail, see Paper I, Sect. 5.1).
The probable errors (PEs)
in the resulting abundance (correction) and micro-turbulent velocity
are determined from the scatter in the data points plotted on
the log A-log
plane, as described
in Paper I (see its Fig. 2 and related explanation). Accordingly,
the probable errors are robust assessments of the internal errors in our
analysis and any result that has a large PE should be assumed
as to have a large uncertainty (e.g. some values of the micro-turbulent
velocity with large PEs in Table 6).
Apart from the internal error, we notice that there should be
a systematic error due to a selection of the lines.
The analysis, which is based on the classical micro-turbulent model
outlined above, should be applied only to lines weaker than a
critical value, since otherwise a large systematic error will
be introduced, as will be shown in Sects. 4 and 5.
We analyzed the lines of the CO first and second overtone bands separately in Papers I and III, respectively, and we summarize briefly the difficulties in these analyses. We then reanalyze the CO lines using the same observed and spectroscopic data as before, but consider the first and second overtone bands together.
The method outlined in Sect. 3 was applied to the lines of the
12C16O first overtone bands (see Figs. 1a-g of Paper I).
The analysis in Paper I was done using PP model
photospheres, but we confirm that the results are similar for
SS model photospheres of the same
.
As an example,
we repeat the line-by-line analysis of
Her in Fig. 3a; for
assumed values of
,
3.5, and 4.0 km s-1,
the CO lines of
(those shown by filled
circles) show behavior expected from the classical micro-turbulent
model outlined in Sect. 3, and it is possible to find a
value for which the abundance corrections do not
depend on
values. The resulting values of
km s-1 and log
provide a null mean logarithmic abundance correction as confirmed
in Fig. 3b
.
Although it appeared to be possible that a self-consistent abundance
analysis could be done for the lines of log
,
it was found that the lines of log
could not be described by the above scheme,
but behave quite differently (also see Figs. 1 and 2 of Paper II).
From a detailed analysis of the line profiles, strengths, and velocity
shifts of these strong lines, we suggested that these lines should be
contaminated by the contributions of extra molecular layers
above the photosphere as detailed in Paper II and hence
these lines cannot be used for abundance analysis.
We then analyzed the lines of the CO second overtone bands by the
same method as for the CO first overtone bands (Paper III).
In this case, many weak lines were available and the abundances
determined were relatively insensitive to the value of
.
At the same time, the value of
could not be specified
well in some cases (Fig. 2 of Paper III). The resulting carbon
abundances based on the CO second overtone bands, however, appeared
to be about twice as high of those based on the CO first overtone bands.
The origin of such a systematic difference between the
carbon abundances based on the CO first and second overtone bands
is difficult to understand, since both the analyses of the first and
second overtone bands were internally self-consistent.
Since weak lines are preferred in abundance analysis, it may be possible to adopt the carbon abundances based on the CO second overtone bands and abandon the results based on the first overtone bands. However, the CO first overtone bands are useful for investigating isotopic abundances, since 13C16O and 12C17O lines of the first overtone bands are good indicators of 13C and 17O abundances, respectively (Sect. 6). We should then investigate in some detail the problems with the first overtone bands if we are to use them for investigating isotopic abundances. We assumed that the isotopic ratios can be determined even if the elemental abundances themselves cannot be determined accurately in our preliminary analysis of the isotopic ratios (Tsuji 2007), but we believe that it is necessary to reconsider such an assumption.
Although the OH analysis detailed in Sect. 5 suggests that
lines stronger than a critical value of
behave differently from the weaker lines, we are
unable to observe this effect in the CO lines of Fig. 3, since only
a few lines of the CO first overtone bands with
could be measured especially in
the later M giant stars (e.g. only 3 lines could be measured in
Her shown in Fig. 3). To include more weak lines, we tried to analyze
together the first and second overtone bands of CO.
The results for
,
3.0, and 3.6 km s-1
are shown in Fig. 4a, where the lines of the first and second overtone
bands of CO are shown by circles and triangles, respectively.
All these lines are analyzed together, regardless of whether
they are of the first or the second overtone bands.
Unlike Fig. 3a based on the CO lines of the first overtone bands alone,
it now appears that there is a clear distinction between the lines of
(shown by the open symbols) and those of
(shown by the filled symbols).
Only the lines of
show the behavior
expected from the classical micro-turbulent model of line formation
theory noted in Sect. 3.
It is evident in Fig. 4a that it is impossible to find a turbulent
velocity that provides a consistent abundance correction to all
lines measured covering
. Thus,
the critical value of
found for the
CO lines is purely empirical as for OH to be discussed in Sect. 5.1.
Unfortunately, we thought previously (Paper I) that
such a critical value is
rather
than -4.75. In this paper, we confirm that this new critical value of
should apply to all red giant stars that we are to study
.
From the weak lines which behave as expected for
different values of
in Fig. 4a, we can determine
value using the method described in Sect. 3. The solution
is found to be
km s-1 and
log
,
for which the mean abundance
correction is null as confirmed in Fig. 4b.
On the other hand, the intermediate-strength lines
remain unexplained. We previously concluded that
lines with
,
most of which originate
in the low excitation levels, should include excess absorption
originating in the outer molecular layers. It is, however, unclear
if this interpretation can be extended to the
intermediate-strength lines and we return to this subject in
Sect. 7.2 after we examine more cases.
The results of the same analysis on other late M giants show
similar patterns as shown for Lyr and 30g Her in
Figs. 5a,b.
It appears that the abundance corrections for the second overtones
(filled triangles) are
often larger than those for
the first overtones (filled circles), but we assume that
such a difference may be within the margins of error of the measurements,
including the difference in the assumed continuum levels.
The case of RX Boo shown in Fig. 5c is the most difficult case
with only 3 weak lines, but nevertheless we could find a solution
in this case. The PEs of both
and
log
,
estimated by the method noted in Sect. 3 are,
however, large, which reflects the poor quality of the data
available for this latest M giant in our sample.
We then examine earlier M giants and apply the same analysis
to the lines of the CO first and second overtone bands of
Peg, as shown in Fig. 6a. We applied the same method of the
line-by-line analysis to the first and second overtone bands separately
in Fig. 1c of Paper I and Fig. 2 of Paper III, respectively.
The lines of the second overtone bands are all weak
and our results were
km s-1 and
log
(Paper III). Although the
lines of the first overtone bands cover the range of
,
these lines could also be
analyzed consistently, and the results were found to be
km s-1 and log
(Paper I).
We now analyze precisely the same lines of the first and second
overtone bands together, and our results for
= 1.6, 2.0
and 2.4 km s-1 are shown in Fig. 6a.
It appears again that no consistent solution can be
obtained for all these lines covering
.
Instead the lines are divided into two groups separated at
,
and a consistent abundance
correction is derived for only the weak lines. The resulting values
of
km s-1 and
log
are confirmed to provide
null abundance corrections as shown in Fig. 6b.
The resulting carbon abundance agrees well with that of Paper III but
differs substantially from that of Paper I. It is now clear that the
analysis of Paper I included the lines of
,
which were inappropriate for abundance analyses.
The results of the same analysis on other early M giants (including
a K giant) show similar patterns. Some examples are shown for
Tau,
Vir, and
Cet in Figs. 7a-c. It is
noted that the anomalous behavior of the lines stronger than
the critical value already appears in the K5 giant
Tau,
and in all the early M giants that we have examined. The nature of the
intermediate-strength lines, however, is quite difficult to understand.
We have previously suggested that the lines with
may be contaminated by the contribution of the warm molecular
layers above the photosphere, but this interpretation has
been accepted only for low excitation lines observed in late M giants
(Paper II). It is noted, however, that the anomalous behaviors of
the intermediate-strength lines shown on the
log A - log
diagram are similar between
the late (Figs. 4 and 5) and early (Figs. 6 and 7) M giants.
Finally, we noticed that the case of
Per is somewhat different.
We again apply the line-by-line analysis with
,
3.5, and 4.0 km s-1 as shown in Fig. 8a. The resulting
km s-1 and log
based on the weak lines are confirmed in Fig. 8b.
It is to be remembered that exactly the same lines of the first
overtone bands alone (omitting lines with
)
resulted
km s-1 and
log
(see Fig. 1e of Paper I), although
it now appears that almost all lines used in Paper I were
intermediate-strength lines, which may be inappropriate for abundance
analysis. Although the weak lines behave as expected, the
intermediate-strength lines do not at all in that the abundance
corrections are now mostly negative, while they were mostly positive
in the cases examined so far (see Figs. 4-7).
We found more or less similar patterns in other objects such as
Eri,
R Lyr, and SW Vir shown in Figs. 9a-c.
We note that the presence of the warm molecular layers
is already known in late M giants including
Per, R Lyr,
and SW Vir (Paper II) and the strong low excitation lines
in Figs. 8 and 9 show positive abundance corrections consistent
with the excess absorption in the outer molecular layers, even though
weaker intermediate-strength lines show negative abundance corrections.
Thus, the presence of the extra molecular layers in these objects
can still be considered, but we return to this subject after
we also analyze OH lines (Sect. 7.2).
We conclude that the weak lines of CO behave quite well as expected from the classical micro-turbulent model described in Sect. 3 and hence photospheric abundances can be determined with reasonable accuracy from these lines, although the behaviors of the intermediate-strength lines cannot yet be understood. Thus, our initial purpose to determine carbon abundance could be achieved, and we summarize the resulting carbon abundances based on the lines of the CO first and second overtone bands for 23 objects in the 6th column of Table 6. We found that the results do not agree at all with those of Paper I but agree rather well with those of Paper III except for a few cases. We conclude that the carbon abundances of Paper I cannot be correct because they were based on the lines that we found to be inappropriate for abundance analysis on a purely empirical basis.
Although our new
values agree rather well
with those of Paper III, we notice for a few objects that there are
differences as large as 0.3 dex, which have a direct effect on the
determination of the nitrogen abundances based on CN lines (Aoki & Tsuji
1997). Roughly speaking, if
and
are
determined by CO and N2 formations, respectively,
We have only considered internal errors (PEs) in our results
shown in Table 6, but the PEs are useful indicators of the reliability
of the results, as noted in Sect. 3. For example, inspection of Table 6
reveals that the PEs for
are larger than
1 km s-1 for BS6861, SW Vir, and RX Boo, and this may be because
the numbers of weak lines available in these cases are only a few.
A large systematic error due to an improper selection of lines in
Paper I is corrected with the recognition of the possibly true
critical value of log
in this paper. Other possible
uncertainties are due to errors in the input parameters such as
,
M, R, and to our modeling of the
photospheres. We previously examined the effect of uncertainties
in
and
within the framework of
PP models. The results (Fig. 14 of Paper III) indicate that the
uncertainties are
for
K and
for
.
The use of SS models, however, introduces
another problem; inspection of Table 5 reveals that
the photospheric extensions and hence surface temperatures show
considerable differences between models of one and two solar masses
especially at low
.
In the model photospheres of
K, for example, the photospheric extension
increases from 0.20 R to 0.51 R and the surface temperature decreases
from 1342 K to 1079 K if mass is reduced from
to
(Table 5). We assume 2
in our
analysis at all times, but mass is the most uncertain parameter.
For this reason, we examined the effect of mass for the typical example
of
Her. We derived the results shown in Table 6
(
and
km s-1)
with a model of
(b/2.0/150/3300). We repeat
the same analysis with a model of
(b/1.0/150/3300).
The result is
and
km s-1. The difference in
derived carbon abundance is 0.11 dex and turbulent velocities differ
by less than 10%. Thus the conclusion obtained with PP models remains almost unchanged if we use SS models.
The reason for this may be due to the use of only
the weak lines which are formed rather deep in the photosphere.
Also CO is quite stable at low temperatures because of its high
dissociation energy; its abundance may be hardly affected by some
changes in the physical conditions. This is in
marked contrast to the case of the Sun in which CO is
just forming at temperatures considerably higher than in M giants
and its abundance is highly sensitive to temperature.
We compare our carbon abundances with those by Smith & Lambert
(1990) for objects analyzed in both analyses presented in Table 8.
Some significant differences are found for
Vir,
Peg, and
Per, which show
differences of 0.38, 0.15, and 0.19 dex, respectively.
The agreements for other objects, however, are generally within
the internal errors of both analyses.
Unfortunately, we are unaware of other high resolution studies
of photospheric abundances of cool giant stars.
We note, however, an interesting intermediate-resolution analysis for a large sample of 70 M giant stars (Lazaro et al. 1991). Their result implied rather low carbon abundances consistent with our previous result of Paper I, which we just discarded. Since their intermediate-resolution analysis necessarily included the intermediate-strength lines as we also did in our previous analysis in Paper I, their result demonstrates that low and high resolution analyses provide similar results if applied to the same types of lines. Certainly, intermediate-resolution analyses should be useful for applications to faint objects in remote stellar systems. However, the problem of how to treat the stronger lines that have not been well modeled by the classical line formation theory must be settled before lower resolution spectra can be used successfully in abundance analyses.
Another interesting attempt of medium resolution spectroscopy
was reported by Decin et al. (2003), who analyzed infrared spectra
observed with ISO SWS.
They carefully compared the observed and synthetic spectra and
estimated basic stellar parameters, including the CNO and their
isotopic abundances. For synthetic spectra of medium resolution,
it may again be difficult to avoid the use of the intermediate-strength
and strong lines inappropriate for abundance analysis.
Nevertheless, the resulting carbon abundances agree rather
well with the other results except for Cet, for which our
result shows unusually high C abundance.
Hydroxyl radical OH shows rich spectra due to ro-vibrational transitions in the infrared region. The fundamental and first overtone bands are observed well in the L and H band regions, respectively, in cool oxygen-rich stars. Spectroscopic data such as the line positions and intensities of OH are taken from the GEISA databank (Jacquinet-Husson et al. 1999), which covers well the high excitation transitions of OH observed in stellar spectra.
![]() |
Figure 12:
Lower excitation potentials (LEPs) of the lines
plotted against
![]() ![]() ![]() |
First, we applied the same method as used for CO to the lines of the
OH fundamentals. We included the carbon abundance determined in
the previous section as an input abundance for each star.
Since OH formation depends on
rather than on
,
determination of the carbon abundance is
a prerequisite for OH analysis to determine the oxygen abundance.
An example is shown for
Her in Fig. 10a, in which the results
of the analysis with
,
3.0, and 4.0 km s-1are shown. We again encountered the same difficulty as for the case
of the CO first overtone bands in that the stronger lines do not follow
the pattern expected from the weaker lines. Furthermore, it appears very
clearly that these weaker and stronger lines are divided at log
and not at log
suggested from our
previous analysis of the CO first overtone bands alone (Paper I).
In fact, we first noticed that log
should be
an important value for dividing the lines into
two groups by this analysis of OH fundamental bands.
With this result in mind, we also confirmed that this value is similarly
important for the CO lines, but clearly seen only if the first and second
overtone bands are analyzed together (Sect. 4.2).
As for CO, the weaker lines show the behavior expected from
the classical micro-turbulent model noted in Sect. 3.
However, the stronger lines, which should suffer heavier saturation
effect, do not show the behavior expected for the saturated lines.
Thus, we apply the same grouping of lines as defined for CO (see
footnote 5) to OH. From the weak lines, we found that the abundance
corrections from different lines show a consistent value for
km s-1 as shown in Fig. 10b.
We found again that it is impossible to smooth out the
reflection at log
by any choice of
the micro-turbulent velocity, and thus the critical value of log
can be deemed as well defined empirically.
This result is consistent with Smith & Lambert (1990), who found that
OH lines of 1-0 and 2-1 bands (mostly stronger than the lines of
3-2, 4-3, 5-4 bands) provide unreasonably large values of the oxygen
abundance and cannot be used for abundance determinations.
It is unclear if the same problem exists for the OH lines of K giant
Tau, since almost all lines measured are the weak lines.
We recall that the anomalous behavior of the intermediate-strength
lines of CO was already noticed in the K giant
Tau
(Sect. 4.2). We are able to use almost all OH lines
detected for
Tau in the abundance analysis because they
are the weak lines and the result is shown in Fig. 11a.
However, all the other late M giants show similar patterns as
for
Her. As examples, the results for R Lyr and SW Vir
are shown in Figs. 11b and 11c, respectively (30g Her, RZ Ari,
and RX Boo also show similar patterns).
We assumed that the photospheric abundances of oxygen could be
obtained exclusively from the weak lines, and summarize the resulting
oxygen abundances and turbulent velocities obtained from the
OH fundamentals in Table 6 for 7 objects for which the L band spectra are
available (those objects with non-zero
in the 11th column).
Again, a problem is why the intermediate strength lines
cannot be used for abundance analysis.
We suggest that some contributions from the outer molecular
layers will disturb the strongest lines in the case of CO. To
examine if this may also be the case for OH, we plotted the lower
excitation potentials (LEPs) of all the lines in Fig. 10a against
log
in Fig. 12b. For comparison, we made a similar plot in
Fig. 12a for CO lines of
Her analyzed in Fig. 3.
Unlike the case of the strong CO lines originating in the levels
of very low LEP less than 0.5 eV, the intermediate strength lines
of OH do not necessarily originate in the very low LEP, but rather
from the levels with LEP
0.5-1.5 eV. However, the
intermediate strength lines of CO do not originate in levels of
very low excitation as well, and the intermediate strength lines
of CO and OH may have the same origin as discussed in Sect. 7.2.
Unfortunately, we have only 7 spectra of the L band region, and we analyze the OH first overtone bands observed in the H band region to extend our sample. Unlike the most molecules for which the overtone bands are weaker than the fundamental bands, the case of OH is rather special in that the first overtone bands are as strong as the fundamental bands or even stronger. As a result, it is more difficult to measure weak lines in the first overtone than in the fundamental bands especially in cooler objects, and we cannot analyze the first overtones of OH in some cool objects. Fortunately, most of these objects are those for which we have the L band spectra and oxygen abundances have already been determined from the OH lines of the fundamental bands (Sect. 5.1).
We apply the same method as used for the fundamental bands and an
example is shown in Fig. 13a for Per. For this case of the
overtone bands, compared to the case of the fundamental bands
discussed in Sect. 5.1, it is shown more clearly that the OH lines
are divided into two groups that behave quite differently.
Although abundances can be determined rather well from the weak
lines, it is hopeless to determine abundances from
the intermediate-strength lines. It is also interesting to note that
the abundance corrections for the intermediate-strength lines show a
similar behavior to those for the intermediate-strength lines of CO
in
Per (Figs. 8a, b).
In the K giant
Tau, most lines measured are the weak lines
as in the fundamental bands, and we use most of these lines for our
abundance determination (Fig. 14a). In the M3 giant
Lib,
the abundance corrections for the intermediate-strength lines show large
positive deviations compared with those for the weak lines, for which
the abundance corrections converge to null values for
log
and
km s-1 (Fig. 14b).
The M5 giant R Lyr shows similar patterns as for
Per (Fig. 14c).
The resulting micro-turbulent velocity is unreasonably large, but
the large PE implies that such a result cannot be realistic.
This is because small errors in the measured equivalent widths
seriously affect the determination of
value
when only a few weak lines are available, and this is a difficulty
of our line-by-line analysis. Nevertheless, the abundance
can be determined rather well because of the use of the weak lines,
which depend little on the value of
.
The difficulty of measuring weak lines in the first overtones
of OH is more significant for some late M giants including
Lyr, 30g Her,
Her, SW Vir, RX Boo, and we were
unable to analyze the H band spectra of these stars.
We confirm both in the CO and OH lines that the behaviors of the
intermediate-strength lines are quite complicated and
show different patterns for different objects. For example, the
intermediate-strength lines, both of OH and CO, do not necessarily
show positive abundance corrections compared with those for the weak lines
in Per and R Lyr.
It is interesting to note that CO and OH lines show similar patterns
on
log A-log
diagram if these OH and CO lines
are measured from the same spectra as are the cases of
Per and
R Lyr
.
We summarize the resulting oxygen abundances based on the weak lines
of the OH first overtone bands for 15 objects in
Table 6 (those with non-zero
in the 9th column).
The oxygen abundances could be determined
independently from the fundamental and first overtone bands
for three objects,
Tau, R Lyr, and RZ Ari. The resulting
oxygen abundances, compared directly in the 12th column of Table 6,
agree well within the internal errors of both analyses.
The micro-turbulent velocities
also agree well except for R Lyr, for which the reason for the
difficulty was discussed already (see Fig. 14c).
The micro-turbulent velocities of
Tau based on the fundamental
and overtone bands agree quite well (
km s-1), but disagree with the result obtained from the
CO lines (
km s-1).
This result again prompts the question of what the micro-turbulent
velocity represents.
We notice in Table 6 that the values of micro-turbulent velocity
are sometimes unexpectedly large both for CO and OH, and that
most cases of
km s-1 are
found when the numbers of weak lines used are below 10.
The
value by our analysis based on the
method outlined in Sect. 3 is
susceptible to small errors in the measured EWs especially if only
a small number of lines can be used. The poor quality of the
data is reflected in the large PE of the derived
in Table 6, and the resulting
value is therefore
unreliable. For example,
values
as large as 6 km s-1 for
Aqr and R Lyr are almost
unusable because they have very large PEs.
Unlike CO, OH may be more susceptible to a change in the
physical conditions. To compare with the case of CO (Sect. 4.3),
we again examined the case of the cool giant Her.
We repeated the same analysis of the OH lines of
Her
done for a model of
(b/2.0/150/3300) giving
the results shown in Table 6 (
and
km s-1). Now, for a model of
(b/1.0/150/3300), the results are
and
km s-1. The differences are
again unexpectedly small. The reason for this may again be due to the
use of only the weak lines, which originate deep in the photosphere.
Finally, we compare our results with those by Smith & Lambert (1990) and by Decin et al. (2003) in Table 8. For most cases of oxygen abundances, the results agree reasonably well within the error bars of these analyses. It is interesting that the agreements are reasonably good despite possible differences in the model photospheres used.
Star | NHa |
![]() |
NKb |
![]() |
log ![]() |
![]() |
NHc |
![]() |
NLd |
![]() |
log ![]() |
![]() |
log
![]() |
km s-1 | km s-1 | ||||||||||||
![]() |
13 | 13 | 25 | 9 | 8.38 ![]() |
2.20![]() |
... | ... | 28 | 25 | 8.79 ![]() |
3.85 ![]() |
8.05 |
23 | 21 | ... | ... | 8.76 ![]() |
3.81 ![]() |
||||||||
![]() |
5 | 5 | 21 | 6 | 8.27 ![]() |
2.46![]() |
42 | 27 | ... | ... | 8.83 ![]() |
2.84 ![]() |
8.58 |
![]() |
6 | 6 | 22 | 10 | 8.13 ![]() |
2.14![]() |
55 | 20 | ... | ... | 8.84 ![]() |
2.67 ![]() |
8.11 |
![]() |
6 | 6 | 11 | 3 | 8.64 ![]() |
1.95![]() |
38 | 18 | ... | ... | 8.98 ![]() |
4.25 ![]() |
7.73 |
![]() |
12 | 12 | 14 | 2 | 8.23 ![]() |
2.09![]() |
51 | 19 | ... | ... | 8.82 ![]() |
2.03 ![]() |
8.15 |
![]() |
5 | 5 | 13 | 4 | 8.56 ![]() |
1.97 ![]() |
30 | 7 | ... | ... | 8.92 ![]() |
6.24 ![]() |
7.88 |
![]() |
12 | 12 | 29 | 7 | 8.27 ![]() |
2.04 ![]() |
53 | 16 | ... | ... | 8.77 ![]() |
3.12 ![]() |
8.11 |
![]() |
6 | 5 | 20 | 1 | 8.41 ![]() |
3.57 ![]() |
25 | 9 | ... | ... | 8.87 ![]() |
3.38 ![]() |
7.94 |
![]() |
19 | 19 | 34 | 9 | 8.32 ![]() |
2.00 ![]() |
36 | 16 | ... | ... | 8.81 ![]() |
2.58 ![]() |
... |
![]() |
16 | 16 | 19 | 4 | 8.50 ![]() |
2.20 ![]() |
36 | 11 | ... | ... | 8.84 ![]() |
2.47 ![]() |
7.90 |
10 Dra | 5 | 5 | 15 | 3 | 8.43 ![]() |
2.69 ![]() |
30 | 15 | ... | ... | 8.92 ![]() |
2.44 ![]() |
8.11 |
![]() |
13 | 13 | 22 | 2 | 8.27 ![]() |
3.59 ![]() |
43 | 7 | ... | ... | 8.87 ![]() |
3.11 ![]() |
8.15 |
BS6861 | 5 | 4 | 12 | 2 | 8.38 ![]() |
4.91 ![]() |
... | ... | ... | ... | ... | ... | ... |
![]() |
5 | 5 | 13 | 1 | 8.22 ![]() |
2.86 ![]() |
19 | 3 e | ... | ... | ... | ... | 8.41 |
RR UMi | 15 | 15 | 15 | 4 | 8.26 ![]() |
2.48 ![]() |
34 | 5 | ... | ... | 8.76 ![]() |
2.66 ![]() |
8.07 |
![]() |
11 | 10 | 17 | 3 | 8.40 ![]() |
2.82 ![]() |
22 | 0 | 42 | 16 | 8.85 ![]() |
2.92 ![]() |
8.14 |
OP Her | ... | ... | 21 | 4 | 8.27 ![]() |
2.53 ![]() |
... | ... | ... | ... | ... | ... | ... |
XY Lyr | ... | ... | 14 | 4 | 8.15 ![]() |
2.96 ![]() |
... | ... | ... | ... | ... | ... | ... |
R Lyr | 14 | 14 | 17 | 2 | 8.26 ![]() |
2.95 ![]() |
... | ... | 52 | 31 | 8.61 ![]() |
2.99 ![]() |
8.06 |
25 | 4 | ... | ... | 8.62 ![]() |
6.53 ![]() |
||||||||
RZ Ari | 12 | 12 | 26 | 7 | 8.10 ![]() |
2.54 ![]() |
... | ... | 28 | 15 | 8.57 ![]() |
3.35 ![]() |
8.44 |
52 | 21 | ... | ... | 8.67 ![]() |
3.49 ![]() |
||||||||
30g Her | 6 | 5 | 24 | 3 | 8.35 ![]() |
2.61 ![]() |
23 | 0 | 38 | 26 | 8.75 ![]() |
2.49 ![]() |
7.92 |
SW Vir | 11 | 6 | 15 | 1 | 8.26 ![]() |
4.21 ![]() |
... | ... | 29 | 16 | 8.54 ![]() |
1.83 ![]() |
8.46 |
RX Boo | 6 | 2 | 13 | 1 | 8.32 ![]() |
3.41 ![]() |
... | ... | 13 | 12 | 8.55 ![]() |
2.04 ![]() |
8.26 |
a The number of CO lines measured in the H band
region, of which the number of weak lines (log
![]() |
We applied the line-by-line analysis to the
isolated lines of 12C16O, 13C16O, and
12C17O. We used only the weak lines but included
12C16O lines of both the first and second overtone bands.
The
values determined from the 12C16O
lines of the first and second overtone bands in Sect. 4.2 were used to
determine abundances of 13C16O and 12C17O.
Given a
value, abundance can be
determined directly for each line. Some examples of the results
are shown in Figs. 15a-d for 10 Dra, RZ Ari,
Oph, and
Gem.
Unfortunately, the number of lines of the isotopic species measured is
rather small and the weak lines could not be measured at all for some
cases especially for 12C17O.
For example, the lines measured for 10 Dra are 12C17O
(2, 0) R 25, R 27, and R 32 in the spectrum shown in Fig. 16a. This
is a favorable case for which three 12C17O lines could
be measured, and these lines are used for the analysis of Fig. 15a.
In the case of RZ Ari shown in Fig. 16b, the 12C17O lines
are weaker and we could measure only two lines (R 25 and R 32) for
the line-by-line analysis of Fig 15b. The 12C17O lines
are still weaker in Aqr and RR UMi shown in Figs. 16c and d,
respectively, and no line could be measured in these cases.
The difficulty of measuring weak lines for 13C16O is more or
less the same, although at least one line could be measured in most
objects except for RX Boo.
The 12C/13C and 16O/17O ratios based on the resulting 12C16O, 13C16O, and 12C17O abundances are given in the 4 and 5th columns, respectively, of Table 7 with their probable errors (PEs). The resulting 12C/13C ratios are mostly larger compared with our previous result (Tsuji 2007), and this is because 12C16O abundances are increased due to the use of lines of both the first and second overtones.
We applied the spectral synthesis method as a check of
our results given in Sect. 6.1. With the carbon abundance
and micro-turbulent velocity from Table 6 and with the isotopic
ratios from Table 7,
we generated synthetic spectra between 4279 and 4294 cm-1,
where 12C17O (2-0)
R 25 - 33 lines are
found. Before we compare the synthetic spectra
with observed ones, we corrected for the effects of the
instrument broadening with the Norton-Beer's apodization function 2
(Norton & Beer 1976), which was used in the apodization of the observed
spectra (see Paper I),
and also of the macro-turbulent broadening based on the measured
intrinsic line widths (see Table 4 of Paper III).
As an example, we compared the observed spectrum of 10 Dra (dots) with
the synthetic one (heavy solid line) in Fig. 16a. In this region, the weak
line of 12C16O (2-0) R 92 can clearly be seen at 4292 cm-1,
which serves as a check of the carbon abundance
.
The lines of 12C17O (2-0) R 25, 27, and 32 used to
determine 12C17O abundance in Sect. 6.1 can be
reproduced well by the synthetic spectrum.
Some other lines of 12C17O such as (2-0) R 26 and
29 should include extra components, since the synthetic
spectrum computed without 12C17O at all (light solid line)
shows some absorption features. Also some 12C17O lines are
blended significantly with other lines, but all lines of
12C17O are not in disagreement with the synthetic
spectrum computed for 16O/17O ratio of 151 (Table 7).
As another example, the case of RZ Ari is shown in Fig. 16b. Although
the 12C17O lines are weaker than in the case of 10 Dra, some
unblended features such as (2-0) R 25 and 32 and all other lines of
12C17O are approximately similar to those in the synthetic
spectrum computed for 16O/17O ratio of 607 (Table 7).
Obj. |
![]() |
![]() |
12C/13C | 16O/17O |
![]() |
2 | 0 | 10.6 ![]() |
>1000 |
![]() |
3 | 2 | 11.1 ![]() |
387 ![]() |
![]() |
5 | 0 | 8.7 ![]() |
>2000 |
![]() |
2 | 2 | 11.1 ![]() |
586 ![]() |
![]() |
2 | 0 | 7.5 ![]() |
>1500 |
![]() |
2 | 0 | 7.9 ![]() |
>1000 |
![]() |
3 | 0 | 7.7 ![]() |
>2500 |
![]() |
1 | 1 | 12.4 ![]() |
687 ![]() |
![]() |
6 | 1 | 10.5 ![]() |
798 ![]() |
![]() |
4 | 0 | 12.3 ![]() |
>2500 |
10 Dra | 3 | 3 | 14.8 ![]() |
151 ![]() |
![]() |
2 | 0 | 9.7 ![]() |
>1000 |
BS6861 | 1 | 0 | 48.5 ![]() |
>1000 |
![]() |
2 | 2 | 16.2 ![]() |
465 ![]() |
RR UMi | 1 | 0 | 10.0 ![]() |
>2000 |
![]() |
2 | 4 | 11.1 ![]() |
102 ![]() |
OP Her | 2 | 1 | 11.3 ![]() |
329 ![]() |
XY Lyr | 1 | 3 | 15.0 ![]() |
223 ![]() |
R Lyr | 1 | 3 | 6.4 ![]() |
368 ![]() |
RZ Ari | 3 | 2 | 7.9 ![]() |
607 ![]() |
30g Her | 2 | 3 | 12.5 ![]() |
211 ![]() |
SW Vir | 2 | 2 | 22.0 ![]() |
432 ![]() |
RX Boo | 0 | 3 | ... | 233 ![]() |
a Number of weak lines of 13C16O; b number of weak lines of 12C17O.
The major purpose of computing the synthetic spectra, however,
is to estimate the 16O/17O ratio for the cases in which
the 12C17O lines are too weak for EWs to be measured
accurately. For example, all 12C17O features are too weak for
their EWs to be measured in the case of Aqr shown in Fig. 16c
and these features, including relatively undisturbed ones such as
(2-0) R 32 and 33, do not contradict with the synthetic spectrum
computed for 16O/17O ratio of 1000. As another example,
the observed spectrum of RR UMi shown in Fig. 16d should be in
approximate agreement
with the synthetic spectrum computed for 16O/17O ratio
of 2000. This conclusion depends largely on subtle depressions in
(2-0) R 25 and 32. On the other hand, definite absorption
seen at the position of R 26 and 29 also appear in the synthetic
spectrum computed without 12C17O at all (light solid line)
and are not due to 12C17O. However, it is possible
that the features assumed to be due to 12C17O may still
include unknown impurities, and the 16O/17O ratio estimated
on such subtle features should at best be regarded as a lower estimate
to the 16O/17O ratio. The 16O/17O ratios estimated
this way are given in the 5th column of Table 7 without error bars.
The most significant uncertainties in the analysis of such faint lines as 12C17O in cool stars are the location of the continuum level and the effect of the blending by many weak lines. As a guide to examining these problems, we computed synthetic spectra similar to those shown in Fig. 16 for the regions with rich lines of 12C17O (4264-4295 cm-1) and 13C16O (4235-4266 cm-1) for all spectra analyzed. This aided the rejection of blended lines in our line list. Furthermore, we examined the correlations between line intensities, line-widths, line-depths, and radial velocities, and rejected those lines that had large deviations. The continuum level was estimated by connecting the highest peaks with a smooth curve and the consistency of the resulting continuum can also be checked by comparison with the synthetic spectra. However, we need to analyze spectra of yet higher resolution to check the accuracy of the continuum used. By comparing results of different authors (see below), the errors in the final isotopic ratios due to these uncertainties are estimated to be as large as 100% (or a factor of two).
The carbon isotopic ratios were studied relatively well; we compare our 12C/13C ratios with those by Smith & Lambert (1990) and other authors in Table 8. In addition to those cited in Table 8, Wallerstein & Morell (1994) reported the 12C/13C ratios for 8 early M giants to be in the range 7-12, although no object common to our sample is included. Our results for 12C/13C ratios appear to agree well with the results of other authors in general.
The determination of the oxygen isotopic ratios is more difficult
but a detailed identification of 12C17O features was
completed for high resolution FTS spectra by Maillard (1974), who estimated
16O/17O ratio in
Her for the first time.
We compare our 16O/17O ratios with those by Harris &
Lambert (1984) and by other authors in Table 8. Wallerstein
& Morell (1994) reported lower limits to 16O/17O ratio
in 8 early M giant stars to be 85-295. Inspection of Table 8 reveals
that 16O/17O ratios by different authors differ by factors
of 2-4. This may largely be due to difficulty to locating the continuum
levels and measuring weak features; more accurate results could be
achieved by using spectroscopic data of higher quality.
The abundance analysis of very cool stars is associated with many difficulties. In particular, absorption lines tend to be strong due to the low continuous opacity at low temperatures in the photospheres of cool giant stars. The consequences are severe blending and uncertainties in the continuum level, which are obvious difficulties by themselves and, in addition, increase the difficulty in measuring other weak lines that are essential in abundance determination. If a sufficient number of weak lines, which are linearly proportional to the effective column densities of a species under consideration, can be measured, abundance determination become straightforward. This ideal case is seldom realized especially for cool stars. However, it was understood to be possible to carry out an abundance analysis by using the micro-turbulent model (or curve-of-growth method) even if saturated lines had to be used. With reasonable success, this model was in fact applied to abundance analyses in general.
The real difficulty in the abundance analysis of cool stars is that the
classical method based on the micro-turbulent model can
no longer be applied for lines of equivalent width higher than a
certain critical value, which we found to be log
for the case of M giant stars (Sects. 4 and 5). Since stronger lines are
formed in the upper photospheric layers where modeling may be
increasingly uncertain, this difficulty can be related to the problem
of the model photospheres. Smith & Lambert (1990) noted that the
CO lines stronger than about log
should not
be used for abundance analysis, since these lines are formed in the
layers above the continuous optical depth of
,
where the photospheric structure cannot be described accurately by the
usual model photospheres. We arrive at a similar conclusion that the
lines stronger than log
cannot be used for
abundance analysis, although by a different argument based on
purely empirical evidence shown most clearly for the OH lines
of the fundamental bands (Fig. 10).
An unanswered question remains why do the intermediate-strength lines show abundance corrections as large as 1.0 dex or more when the weak lines show null corrections? It is unlikely that the uncertain in the thermal structure of the upper layers could produce such a large effect on the intermediate-strength lines, since differences in the model photospheres have only modest effects on abundances as we have seen before (Sects. 4.3 and 5.3). Also, no model of the upper photosphere is known to explain the difficulty of the intermediate-strength lines or the stronger lines. However, the dynamical structure represented by stellar turbulence may have a more dramatic effect on the intermediate-strength lines heavily saturated. Taking advantage of the fact that the line positions and profiles accurately can be measured with FTS, we showed that the line asymmetries and differential line shifts of CO lines are quite large and that a simple model of the depth-independent and isotropic Gaussian micro-turbulence cannot be justified (Paper III).
Even if we know that a simple model of micro-turbulence is not
sufficiently realistic, it is unknown how the behavior of the
intermediate-strength lines (Figs. 4-11 and 13, 14) can be
reproduced by a more accurate description of stellar turbulence.
It is interesting if a more consistent treatment of the dynamical
structure of stellar photosphere by a 3D model
can resolve the present difficulty. A detailed
numerical simulation of surface convection in red giant stars is
reported by Collet et al. (2007). Although their simulations
were not extended to the regime of M giants, their result
for the coolest model of
K and [Fe/H] = 0.0
indicated that the difference between 3D and 1D oxygen abundances
derived from OH lines (not ro-vibrational lines but of A-X system
at 3150 Å) is about 0.1 dex. This appears to be
due to the difference in thermal structures of the 3D and 1D models, but possible effects of the dynamical structure of the
3D models on the saturated lines are still unknown.
Their results indicated that the effect of the 3D structure should
be more important in stars of lower mentallicities, but
it is unclear if such a large difference in abundances derived
from the weaker and stronger lines in our population I sample can be
relaxed by considering the dynamical effect of convection
in the 3D models. Thus, we conclude that the difficulty
we encountered may not be resolved by considering variation in
models of the photospheric structure.
We know that the strong lines of CO in late M giant stars are
affected significantly by the contributions of the extra molecular
layers (Paper II). Since the intermediate-strength lines and
the strong lines behave similarly on
log A-log
diagram (e.g. Figs. 4, 5),
a natural question to ask is; can the intermediate-strength
lines be understood in a similar way to the strong lines?
For the late M giants, the intermediate-strength lines do not
necessarily originate in the levels of low LEP (see Fig. 12a),
but increasingly reduced contributions from higher excitation
lines can be possible for modest kinetic temperatures in the warm
molecular layers, and the trend of intermediate-strength lines
together with the strong lines in these giants (Figs. 4, 5) can be
explained consistently. Given that the presence of the MOLsphere
was demonstrated in the late M giants (Paper II), it should at the
same time produce excess absorption in the intermediate-strength
lines as it does for the strong lines. It is then natural to assume
the same origin for the intermediate-strength lines and the strong
lines in the late M giants.
Interestingly, the intermediate-strength lines of CO in the early
M giants (including a K giant) provide, in general, larger abundance
corrections compared with those from the weak lines (e.g. Peg,
Tau,
Vir,
Cet in Figs. 6, 7).
We have not previously considered the effects of outer molecular
layers in these early M giant stars
.
But why not? The behavior of the intermediate-strength lines on the
log A-log
diagram, in particular, are so
similar for the late M (Figs. 4, 5) and early M
(Figs. 6, 7) giants
that it is natural to assume the same origin for the
intermediate-strength lines in the early and late M giants. Furthermore,
we identified H2O bands on the ISO spectra of the K-early M giants
included in our present sample (e.g.
Tau,
Peg, and
Cet) and one possibility is that these H2O bands
are originating in the outer molecular layers as for the late M giants
(Tsuji 2001). However, this is not a unique interpretation. For example,
Ryde et al. (2002) detected H2O lines in the high resolution
spectrum of the K giant
Boo and suggested that its origin can be
due to the anomalous structure of the photosphere. We note, however,
that the anomalous behavior of the intermediate-strength lines and the
presence of H2O bands in the K-early M giant stars can consistently
be understood by assuming the presence of the outer molecular layers.
In the case of CO lines in late M giants such as
Per, R Lyr,
and SW Vir, and also in the M3 giant
Eri, the
intermediate-strength lines and the strong lines show quite
different behaviors as shown in Figs. 8, 9; the strong lines show
positive abundance corrections, while the intermediate-strength lines
show mostly negative corrections for the carbon abundances derived
from the weak lines. We first thought that this difference should be a
reason for considering alternative explanations for the
intermediate-strength lines and the strong lines.
However, we reach the opposite conclusion that this provides
evidence that the intermediate-strength and strong lines have
identical origins. In fact, a possible contribution from the
outer molecular layers can be either absorption or emission
depending on the extension, temperature, and optical thickness
of the outer molecular layers for a particular line.
For the intermediate-strength lines, we can therefore explain
the negative abundance corrections as reductions in their EWs
due to infilling by emission from the outer molecular layers.
At present, however, the negative abundance corrections for
the intermediate-strength lines cannot be explained if they are
assumed to originate in the photosphere.
Obj. | log
![]() |
log ![]() |
log
![]() |
log ![]() |
12C/13C a | 12C/13C | 16O/17O a | 16O/17O |
![]() |
8.38 ![]() |
8.40 ![]() |
8.78 ![]() |
8.78 ![]() |
11 ![]() |
10 ![]() |
>1000 | 560 ![]() |
8.35 ![]() |
8.93 ![]() |
10 ![]() |
||||||
![]() |
8.13 ![]() |
8.51 ![]() |
8.84 ![]() |
8.91 ![]() |
9 ![]() |
12 ![]() |
||
![]() |
8.64 ![]() |
8.20 ![]() |
8.98 ![]() |
8.93 ![]() |
11 ![]() |
10 ![]() |
||
![]() |
8.27 ![]() |
8.42 ![]() |
8.77 ![]() |
8.81 ![]() |
8 ![]() |
8 ![]() |
>2500 | 1050 ![]() |
8.20 ![]() |
8.93 ![]() |
5 ![]() |
||||||
![]() |
8.32 ![]() |
8.43 ![]() |
8.81 ![]() |
8.82 ![]() |
11 ![]() |
13 ![]() |
798 ![]() |
325 ![]() |
![]() |
8.50 ![]() |
8.61 ![]() |
8.84 ![]() |
8.82 ![]() |
12 ![]() |
16 ![]() |
||
10 Dra | 8.43 ![]() |
8.59 ![]() |
8.92 ![]() |
9.04 ![]() |
15 ![]() |
12 ![]() |
||
![]() |
8.27 ![]() |
8.46 ![]() |
8.87 ![]() |
8.92 ![]() |
10 ![]() |
15 ![]() |
||
![]() |
11 ![]() |
17 ![]() |
102 ![]() |
190 ![]() |
||||
5 ![]() |
450 ![]() |
|||||||
30g Her | 8.35 ![]() |
8.25 ![]() |
8.75 ![]() |
8.73 ![]() |
13 ![]() |
10 ![]() |
211 ![]() |
675 ![]() |
a Present results;
b Smith & Lambert (1990);
c Decin et al. (2003);
d Hinkle et al. (1976); e Maillard (1974); f Harris & Lambert (1984); g Harris et al. (1985). |
Finally, an additional problem is if the intermediate-strength lines of
OH can be understood in the same way as for the CO lines. Unlike CO,
OH may not be so abundant in the outer molecular layers, since oxygen
may be mostly in H2O within cooler environments expected for the
MOLsphere. For this reason, we first believed that the anomalous
strengths of the intermediate-strength lines of OH may not be due to
contamination from the outer molecular envelope, but may instead
originate in the photosphere itself. However, the infrared OH lines
show behavior quite similar to the CO lines (e.g. Figs. 4 and 10),
and it is again natural to assume
the same interpretation for the OH and CO lines. Given that OH has
been detected even in the cooler circumstellar envelopes of red giant
stars by OH masers, it may be possible that OH can be abundant in the
warmer MOLsphere. Furthermore, the assumption of chemical equilibrium
should probably not be applied to the outer layers and we should
instead use the OH lines as probes of non-equilibrium processes in the
outer layers. Then, we conclude that the intermediate-strength lines of
OH observed in the late M giants such as Her, R Lyr, and SW vir
(Figs. 10, 11) as well as in the earlier M giant such as
Lib
(Fig. 14b) also include contributions from the outer molecular layers.
![]() |
Figure 17: Observed abundances of a) carbon (Table 6), b) nitrogen (Aoki & Tsuji 1997; Table 6), c) oxygen (Table 6; if the results derived from the H and L band spectra are available, mean values are used), and d) C + N + O, compared with the predictions of evolutionary models (Claret 2004). The results of FDU and SDU for the case of Z = 0.02 are shown by solid and dash-dotted lines, respectively, in each panel. The observed C + N + O abundances are smaller by about 0.16 dex compared with the predictions based on Z = 0.02 as can be seen in panel d). This initial metallicity correction of -0.16 dex is applied to the predicted C, N, and O abundances in panels a), b), and c), respectively, and the corrected results are shown by dashed lines. In each panel, the solar abundances are shown by open square (Anders & Grevesse 1989; Ayres et al. 2006) and by open diamond (Allende Prieto et al. 2002). |
The possible presence of the outer molecular layers in red giant stars
is supported by direct measurements of the angular extensions.
For example, angular diameters of normal K and M giant stars, including
several objects common to our sample, were measured
in the strong TiO bands at 712 nm and in the nearby
pseudo-continuum at 754 nm (Quirrenbach et al. 1993). The observed
712 nm/754 nm diameter ratios appeared to be larger than
predicted by model photospheres, suggesting that the
photospheric models cannot describe adequately the outer layers
of these red giant stars, and the result may lend support for the
presence of the extra molecular envelopes (Quirrenbach 2001).
Interferometric observations in the infrared bands provided
direct support for the presence of extended warm molecular gas layers not
only in Mira variables but also in late M giants (e.g.
Mennesson et al. 2002; Perrin et al. 2004).
On the other hand, the apparent angular diameter of Her
in the mid infrared was found to be larger by 30% than its
near infrared size, but the high resolution 11
m spectrum does not
show any substantial spectral lines in
Her (Weiner et al. 2003).
This result was explained by a balance of absorption and emission
in a spherically extended envelope, and the mid infrared spectra and
visibilities could be consistent with the warm molecular envelope model
(Ohnaka 2004).
We should remember that the possibility of a warm molecular
envelope in the early M giant stars was noticed previously when we first
identified H2O bands in the spectrum of the M2 giant Peg
observed with ISO SWS (Tsuji et al. 1997). Also, H2O bands were
found in several M giants earlier than M6 by Matsuura et al. (1999)
with IRTS (Infrared Telescope in Space) launched in March 1995 by ISAS.
These results were further extended to several K-M giant stars
(Tsuji 2001) and also confirmed by other authors (e.g. Decin et al. 2003).
The implications of this finding, however, were not fully recognized
even by ourselves, and some time elapsed before this finding was finally
related to the difficulty in the abundance analysis of red giant stars.
Since the temperatures and radial velocities of the MOLsphere differ only
slightly from those of the upper photosphere, it was
difficult to recognize the presence of the MOLsphere using the
spectroscopic method
.
However, we finally recognize the effect of the overlaying warm
molecular layers on the high resolution infrared spectra of dozens of
red giant stars.
Considering several lines of reasonings outlined above, we conclude that the presence of a warm and quasi-static molecular envelope above the photosphere is a common feature in K and M giant stars. It is unexpected and surprising if that both CO and OH lines of intermediate-strength with LEP as high as 2.0 eV (Fig. 12) include the absorption and/or emission originating in the outer molecular layers. We previously noted this possibility mainly for the late M giants (Paper II), but now we find that the presence of the extra molecular envelope is a general phenomenon not only in the late M giants but also in the early M and K giants as well. Also, the fact that the rather high excitation lines are included in these intermediate-strength lines implies that the extra molecular envelope is indeed ``warm'' and should be even warmer than we imagined before from the analysis of the strong lines (Paper II). Yet we know little about this new feature consisting of warm gaseous molecules such as CO, OH, and H2O. We hope to extend our analysis done for the strong lines (Paper II) to the intermediate-strength lines and clarify the nature of the warm molecular envelope in more detail.
We conclude that the observed infrared spectra should actually be a hybrid of multiple components originating in the photosphere and overlaying warm molecular layers. The hybrid nature of the infrared spectra of red giant stars is now confirmed both in the low resolution ISO spectra and high resolution FTS spectra. We must then abandon the traditional idea that the stellar spectra are primary defined by the stellar photospheric structure, at least for the infrared spectra of K-M giant stars.
Even if the photosphere of a cool giant star is veiled by
the warm molecular envelope, the weak lines of high excitation may
remain almost undisturbed.
The weak lines are clearly distinguished from the intermediate-strength
lines on the
log A-log
diagram (Figs. 4-11,
13-14), and we confirmed that the weak lines behaved as
expected from the classical micro-turbulent model (Sects. 4 and 5).
Thus, we are convinced that abundance determination could be possible
using the weak lines of log
.
With this caution,
our results agreed well with those by Smith & Lambert (1990), who have
already used only weak lines in their abundance analyses. We therefore
continue to compare our results with the predictions of evolutionary models.
We plotted the resulting C and O abundances in Figs. 17a and c,
respectively. We plotted N abundances based on our
previous analysis of the same sample (Aoki & Tsuji 1997),
corrected for the effect of the revised C abundances (14th column
of Table 6), in Fig. 17b. For 5 late M giants for which the N abundances based on the ro-vibrational lines of NH are available,
we plotted the mean log
values based on the lines
of CN and NH. Given the C, N, and O abundances, we obtained
the sum of the C, N, and O abundances as plotted in Fig. 17d.
The results are compared with the predictions of the standard
evolutionary models by Claret (2004); the results of the first
dredge-up (FDU) are shown by solid lines and those of the second
dredge-up (SDU) by dash-dotted lines in Figs. 17a-d.
We first examined the sum of the C, N, and O abundances in Fig. 17d.
The model prediction confirmed that the CNO processing
conserved the sum of the C, N, and O abundances corresponding to
Z = 0.02, as assumed by the evolutionary models. The observed
values are generally smaller than the values of the models, which are
log
.
The mean of the observed values in low mass stars
is
log
as shown by
the dashed line in Fig. 17d, and thus
the mean metalliciy of the low mass stars that we have analyzed
is smaller than the value assumed in the evolutionary models by
Claret (2004). This metallicity difference should be
kept in mind when comparing the observed and predicted
C, N, and O abundances. A solar value of the C + N + O abundances based on the C and O abundances by Alende Prieto et al.
(2002) is shown by
the open diamond and a classical value (e.g. Anders Grevesse
1989),
supported by Ayres et al. (2006), is indicated by the open squares
in Fig. 17d; the N abundance is taken from Anders & Grevesse (1989).
The observed carbon abundances are generally smaller than those of the
model predictions, but may be consistent with the predictions after the
metallicity difference between the evolutionary models and actual stars
is corrected as shown by the dashed line in Fig. 17a. The scatters
about the dashed line may be due to those of the metallicity in
our sample, except for a few objects. For example, two stars appear
above the predicted FDU curve; the object showing the largest deviation
is
Cet and the other one is
Aqr. It is
interesting to notice that these two stars also show the largest
under-abundances of N at
in Fig. 17b .
The nitrogen abundances may roughly be consistent with the FDU curve
with and/or without metallicity correction in general.
However, the deviations from the predictions appear to be larger in a few
objects, and it is as if some extra mixing may be needed as in the
case of the 12C/13C ratios discussed in Sect. 7.4.
Most lines of both CN and NH used are weak lines that satisfy
the condition of log
,
and N abundances may be free
from the difficulty encountered in the analysis of the CO first overtone
bands in Paper I (also Sect. 4.1).
Finally, the oxygen abundances in Fig. 17c show a similar pattern as
the C + N + O abundances in Fig. 17d, and the scatters may represent those
of the metallicity in our sample. There is a tendency for the oxygen
abundances to be higher in higher mass objects. The resulting oxygen and
carbon abundances correlate rather well and the mean
ratio for 19 objects for which C and O abundances
are determined is 0.37 (with the standard deviation of
0.10).
![]() |
Figure 18: a) The observed 16O/17O ratios are plotted against stellar mass by filled circles and by the filled triangles if only lower limits can be estimated. Predicted 16O/17O ratios by FDU and SDU are shown by solid and dash-dotted lines, respectively. b) the observed 12C/13C ratios are plotted by filled circles against stellar mass. Dashed curve is the predicted 12C/13C ratios by the ``Compulsory Deep Mixing (CDM: Eggleton et al. 2008)''. |
The resulting 16O/17O ratios (Table 7) are plotted versus stellar mass in Fig. 18a by filled circles or by filled triangles if only lower limits could be estimated. For comparison, predicted results of FDU and SDU by Claret (2004) are shown by solid and dash-dotted lines, respectively. The large variation in 16O/17O of low mass stars is consistent with the predictions of the evolutionary models, confirming the previous analysis by Dearborn (1992) using the observed data known at that time. In the higher mass stars, some objects have very low 16O/17O ratios, but the general trend that the 16O/17O ratios are confined to lower values is consistent with the predicted ones. Considering the possible uncertainties in the stellar masses, the observed and predicted 16O/17O ratios can be regarded to be not in serious contradiction.
The observed 12C/13C ratios (Table 7) are shown
by filled circles and compared with the predicted ones (Claret 2004)
in Fig. 18b. The observed 12C/13C ratios are now around 10
in contrast to our previous 12C/13C ratios, which were mostly
below 10 (Tsuji 2007), but yet too small compared with 12C/13C
20 predicted by the standard evolutionary models. This
contradiction has been known for a long time and we simply confirm this
long-standing puzzle. To resolve this dilemma,
extra mixing of the CN-cycle processed material had to be assumed.
For example, a deep circulation below the bottom of the
convective zone, referred to as ``cool bottom processing'',
was proposed (Boothroyd & Sackmann 1999).
It was shown recently by Eggleton et al. (2008) that a deep
mixing mechanism should operate in all low mass giants due to
a molecular weight inversion resulting from 3He (3He,
2p) 4He reaction in the layers just above the H-burning shell.
The predicted 12C/13C ratios by this ``Compulsory Deep Mixing
(CDM)'' are shown by the dashed line in Fig. 18b. The more modern
prediction appears close to the upper boundary of the observed results,
and discrepancies between observations and predictions are relaxed
if not perfectly.
Given that the presence of the warm molecular envelope or
MOLsphere is a general phenomenon
in normal red giant stars, the next problem is to clarify the
nature of the MOLsphere in detail, and to understand the structure
of the entire atmosphere extending beyond the photosphere.
An initial step towards this goal was achieved by a detailed
MHD simulation of the red giant wind by Suzuki (2007), who showed that
nearly static regions of
K are formed
at several stellar radii above the photosphere.
It is interesting that the mass loss from red giant stars with
wind velocities far smaller than the surface escape velocity can
be understood consistently with his simulation in which the stellar
winds are accelerated from the quasi-static region
at several stellar radii above the photosphere.
The quasi-static region predicted by Suzuki's simulation can be identified with the warm molecular envelope that we confirmed in several observational studies as discussed in Sect. 7.2; we hope that the simulation of Suzuki can be extended to include spectral modeling so that more detailed confrontation between theory and observations can be achieved. We include the measured data on the stronger lines in Tables 3a-3w, in addition to those on the weaker lines used in the abundance analysis, and we hope that these data can be interpreted correctly with more sophisticated models in the future. With this detailed confrontation between theory and observations, we expect to find a consistent solution to the long-standing problem of the origin of mass-loss from red giant stars for the first time.
We initiated our study hoping to clarify the nature of the difficulty
encountered in our previous analysis of CO (Papers I and III ), since
we felt that something is still not understood well in abundance
analyses of cool giant stars. At first, we had to recognize that
line spectra of cool giant stars are highly complicated
and may consist of at least three different groups of lines of
different origins; the weak lines of log
,
the
intermediate-strength lines of
,
and
the strong lines of log
.
Of these three groups, only the
weak lines could be understood well on the basis of the classical
micro-turbulent model of line formation. On the other hand, it was
known that the strong lines could not be understood as being of photospheric
origin alone but should include an extra component originating in the outer
molecular layers (Paper II).
The most difficult case was the intermediate-strength lines. However, we finally concluded that the intermediate-strength lines are essentially the same as the strong lines. This means that not only the low excitation strong lines but also medium strong lines with LEP as high as 2 eV (Fig. 12) are disturbed by the contaminations originating in the outer molecular envelope, which we refer to as MOLsphere for simplicity. This in turn means that most of the dominant molecular lines observed in the infrared spectra of K-M giant stars should be badly disturbed by the contaminations from the extra component beyond the photosphere. At the same time, however, this finding provides further support for the presence of the warm molecular envelope on the basis of the high resolution spectra of dozens of red giant stars.
We found that the largest fault in our previous analysis (Paper I)
was to have assumed that the critical value discriminating the weak
and the intermediate-strength lines was log
rather than -4.75. For this reason,
many lines which are disturbed by the contaminations from the outer
molecular envelope were included in the abundance analysis of Paper I.
It is certainly difficult to attempt abundance analyses
with such lines, and the reason why abundance analysis of cool
giant stars often encountered difficulties appeared to be simple,
at least partly. This result also implies that it should be
difficult in principle to apply low resolution infrared spectra
to abundance analyses of cool luminous stars, and the hybrid nature
of the infrared spectra reinforces the reason why high resolution
spectroscopy is indispensable in abundance analyses.
On the other hand, it is fortunate that the weak lines are relatively free from disturbance by extra component of non-photospheric origin and we are convinced that the abundance analysis of cool giant stars is possible with these lines. With a limited number of the weak lines, we could determine the C, O, and their isotopic abundances with reasonable accuracy. The resulting CNO and their isotopic abundances are roughly consistent with the predictions of the first and second dredges-up during the red giant phase. We noticed, however, that there are some deviations from the evolutionary models especially for the 12C, 13C, and 14N abundances of some stars, and it seems that some additional episodic events may be needed to explain them. Since the CNO isotopes in the universe, especially 13C, 14N, and 17O, are largely supplied by red giant stars, further refinements of both abundance analyses and evolutionary models are required.
In conclusion, high resolution infrared spectra provide a wealth of information about not only photospheric abundances but also the structure of the entire atmosphere extending beyond the photosphere. For this reason, infrared spectra must be of hybrid nature with contributions from multiple components. It is essential to keep the hybrid nature in mind in the interpretation and analysis of the infrared spectra of cool luminous stars.
Acknowledgements
I am most grateful to Drs. K. H. Hinkle and S. T. Ridgway for generous support in observing the FTS spectra at KPNO, in applying the KPNO archival data, and in analyzing the FTS spectra. I also thank an anonymous referee for careful reading of the text and for helpful comments. Data analyses were in part carried out on common use data analysis computer system at the Astronomy Data Center (ADC) of NAOJ. This work was supported by Grant-in-Aid for Scientific Research (C) No. 17540213.
Obj. | Filter | Res. (mK) | S/N | Date of obs. | EW data in |
alf Tau | H | 46.19 | 60 | Aug. 17, 1976 | Table 3a |
alf Tau | K | 46.19 | 76 | Sep. 28, 1976 | Table 3a |
alf Tau | L | 49.90 | 159 | Oct. 21, 1977 | Table 3a |
del Oph | H, K | 49.70 | 75 | Jun. 24, 1977 | table3b |
nu Vir | H | 41.21 | 51 | Apr. 08, 1982 | Table 3c |
nu Vir | K | 41.20 | 115 | Apr. 09, 1982 | Table 3c |
alf Cet | H, K | 46.19 | 69 | Sep. 28, 1976 | Table 3d |
sig Lib | H, K | 46.19 | 89 | Aug. 18, 1976 | Table 3e |
lam Aqr | H, K | 46.19 | 61 | Jul. 06, 1976 | Table 3f |
bet Peg | H, K | 48.76 | 103 | May 12, 1976 | Table 3g |
tau4 Eri | H, K | 46.19 | 59 | Sep. 29, 1976 | Table 3h |
mu Gem | H | 46.19 | - | Nov. 07, 1976 | Table 3i |
mu Gem | K | 16.53 | 80 | Jan. 27, 1988 | Table 3i |
del Vir | H | 46.19 | 54 | Jun. 18, 1976 | Table 3j |
del Vir | H, K | 49.70 | 86 | Jun. 24, 1977 | Table 3j |
10 Dra | H, K | 49.82 | 61 | Jun. 28, 1977 | Table 3k |
rho Per | H | 29.70 | 44 | Apr. 07, 1982 | Table 3l |
rho Per | K | 22.28 | 66 | Apr. 08, 1982 | Table 3l |
BS6861 | H, K | 46.19 | 42 | Sep. 29, 1976 | Table 3m |
del2 Lyr | H, K | 46.19 | 78 | Apr. 11, 1976 | Table 3n |
RR UMi | H, K | 46.19 | 61 | Aug. 18, 1976 | Table 3o |
alf Her | H, K | 49.70 | 106 | Jun. 24, 1977 | Table 3p |
alf Her | L | 16.53 | 45 | Jul. 09, 1987 | Table 3p |
OP Her | K | 16.53 | 142 | Jul. 01, 1988 | Table 3q |
XY Lyr | K | 16.53 | 86 | Jul. 08, 1987 | Table 3r |
R Lyr | H | 29.70 | 70 | Apr. 08, 1982 | Table 3s |
R Lyr | K | 16.53 | 65 | Jul. 07, 1987 | Table 3s |
R Lyr | L | 16.53 | 107 | Jul. 09, 1987 | Table 3s |
RZ Ari | H, K | 46.19 | 88 | Aug. 17, 1976 | Table 3t |
RZ Ari | L | 16.53 | 65 | Oct. 02, 1987 | Table 3t |
30g Her | H, K | 49.10 | 102 | Jun. 30, 1977 | Table 3u |
30g Her | L | 16.53 | - | Jul. 09, 1987 | Table 3u |
SW Vir | H, K | 46.19 | 65 | Jun. 18, 1976 | Table 3v |
SW Vir | L | 16.53 | 88 | Jul. 09, 1987 | Table 3v |
RX Boo | H, K | 48.10 | 62 | Jan. 14, 1976 | Table 3w |
RX Boo | L | 16.53 | 108 | Jul. 09, 1987 | Table 3w |
alf Tau | ||||||
---- | ||||||
Isotope | v![]() ![]() |
Rot. tr. | nu(cm-1) | log gf | L.E.P.(cm-1) | log EW/nu |
C12O16 | 3 1 | P 2 | 4199.480 | -6.282 | 2154.701 | -4.739 |
C12O16 | 2 0 | P 9 | 4222.953 | -6.128 | 172.978 | -4.585 |
C12O16 | 4 2 | R 68 | 4237.805 | -4.241 | 12982.145 | -4.721 |
C12O16 | 3 1 | R 8 | 4238.289 | -5.600 | 2280.402 | -4.611 |
C12O16 | 4 2 | R 36 | 4243.637 | -4.604 | 6763.398 | -4.608 |
C12O16 | 3 1 | R 18 | 4266.090 | -5.248 | 2794.075 | -4.572 |
C12O16 | 3 1 | R 22 | 4275.180 | -5.154 | 3105.648 | -4.553 |
C12O16 | 2 0 | R 92 | 4292.047 | -4.817 | 16004.551 | -5.067 |
C12O16 | 3 1 | R 32 | 4292.750 | -4.969 | 4148.160 | -4.563 |
C12O16 | 3 1 | R 35 | 4296.566 | -4.923 | 4533.898 | -4.595 |