It is well established that the Balmer lines are good temperature indicators for
because of their small gravity and metallicity
dependence (Smalley & Dworetsky 1993; Furhmann et al.
1994). In this work, effective
temperatures for both non-variable and RR Lyrae stars have been
obtained from the comparison, using a least-squares fitting technique,
between the observed
profile and a grid of synthetic profiles
using ATLAS8 (Kurucz 1979). The gravity and metallicity were
fixed to
and [M/H] = -1.0 for the RR Lyrae stars and to
and [M/H] = 0.0 for the sample of non-variable
stars. Effective temperatures of the observed stars are given in
Tables 3 (click here), 4 (click here) and 5 (click here).
It has been pointed out that for stars cooler than 8500 K,
convection is starting to become efficient enough to modify the temperature
gradient and hence the Balmer profiles with the exception of
(van't Veer-Menneret & Megessier 1996). However, we have not
found any evidence for this. Comparing our effective
temperatures derived from
and those obtained from the literature
for the non-variable stars shows a mean difference and a standard deviation
of
. Moreover, previous comparisons between effective
temperatures derived from
from the Infrared Flux method and
spectrophotometric methods (Solano & Fernley 1997) did not
show systematic differences either.
The crowding of the observed spectral region made the line selection difficult and only a small number of Fe I lines were selected for the abundance calculation (Table 6 (click here)): weak lines (with associated large errors due to the moderate S/N ratio of our observations) and strong lines (sensitive to saturation effects) were not considered. The relatively low signal-to-noise together with the crowded spectral region also cause an uncertainty in the measured equivalent width. An estimation of this uncertainty was obtained by comparing the equivalent widths of lines of non-variable stars with two or more observations. This gave a typical error of 7%. For the RR Lyrae stars, which generally have lower S/N, the error in EW is higher and we estimate 10%.
Synthetic equivalent widths were calculated for a grid of
temperatures and metallicities using the spectrum synthesis code
XLINOP and the ATLAS8 models (Kurucz 1979). Log gf
values were taken from Thévenin (1989)
(Table 6 (click here)). The abundance values were calculated by
computing the sum of squares of the differences between the
observed equivalent widths and the synthetic equivalent widths for
five values of metallicity. The sum of squares were then plotted
against the metallicities and a parabola was fitted, the adopted
metallicity being the minimum of the parabola. In order to check
the internal consistency of the whole process we have calculated the
abundances of Procyon (HR 2943) and some other standard stars. A
microturbulence of 2.0 and a surface gravity of
were used for all the stars. The results are given in
Table 3 (click here). Comparing these values and those obtained from
the literature shows good agreement (mean difference and a standard
deviation of
).
The microturbulence velocity has been set to
(Lambert et al. 1996) for most of the RR Lyrae stars since the
scattered in the abundances derived from the lines was too large or the
number of lines used too small to distinguish any systematic differences in
the abundances derived from weak and strong lines. Also, following
Fernley & Barnes (1996), a value of
was adopted for all the RR Lyrae stars.
The derived abundances from the individual RR Lyrae spectra are listed in
Tables 4 (click here), 5 (click here). For stars with two or more abundance
determinations it can be seen that there is good agreement, the differences
are typically . The final abundance for each RR
Lyrae star is listed in the Appendix. Also listed are values from the
literature. The two sets of values are compared in Fig. 1 (click here)
where it can be seen that, with the exception of XZ Aps, there is reasonable
agreement (rms of the difference 0.18 dex). Since the literature values have
a typical error of
, this rms difference implies a
similar error in our work although there is some suggestion that our
metallicities are systematically higher (mean difference 0.07 dex). The
sensitivity of the derived abundances on the atmospheric parameters has
also been studied. Errors of
and
have been assumed for effective
temperatures and microturbulence velocities which produce errors of
and
respectively. The influence of the errors in
on the
derived abundances is negligible.
From Tables 4 (click here), 5 (click here) we can see that stars with
abundances calculated at phases out of the minimum light interval show good
agreement with the values calculated at minimum light even if at
these phases the stellar atmosphere may suffer from
fast acceleration which produces large ranges in the physical parameters in a very short
time-scale. Clementini et al. (1995) also found similar
results. The explanation of this is related to the fact that the
steepest variations in a -phase diagram only occurs in the
interval
(Fernley et al. 1989). In
the rest of the phase interval the variation of
is such that,
if the ephemeris are accurate enough, the temperature can be determined
with a small error. This is also valid for the variations in gravity (e.g.
Liu & Janes 1990) with the additional advantage of the
little influence of
on the abundance determination.
Identification
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Log gf Lower level energy
( )
415.1945
-1.570 28604 . 605
Fe I 415.2056
-0.990 32873 . 617
Fe I 415.2169
-3.180 7728 . 056 415.3899
-0.590 27394 . 688
Fe I 415.4098
-1.500 27394 . 688
Fe I 415.4500
-1.040 22838 . 318
Fe I 415.4805
-0.640 27166 . 818 417.6577
-0.680 27166 . 818 418.7047
-0.660 19757 . 033
Fe I 418.7588
-1.320 27666 . 346
Fe I 418.7617
-2.410 29356 . 740
Fe I 418.7795
-0.650 19562 . 439 419.5329
-0.720 26874 . 548
Fe I 419.5618
-1.710 24335 . 760
Fe I 419.6208
-0.920 27394 . 688
Fe I 419.6533
-2.190 23783 . 613 421.3653
-1.550 22946 . 809 421.9355
-0.720 28819 . 945
Fe I 421.9419
-1.650 24118 . 814 422.2221
-0.980 19757 . 033 425.0130
-0.350 19912 . 494
Fe I 425.0787
-1.040 12560 . 930
Fe I 425.0893
-2.040 24772 . 018 440.4080
-2.530 31686 . 346
Fe I 440.4750
-0.250 12560 . 930 . Only used in metal-poor RR Lyraes (saturation effects
may occur in metal-rich RR Lyraes).
Figure 1: A comparison between the [Fe/H] values of the present work and
previous work