Below we discuss the abundance pattern of the stars studied (consult Table 4 (click here) for a summary).
Lithium - as derived from two lines at 
and
the
abundance is normal, compared with normal A-type stars, in HD108651 and
HD138213 and slightly underabundant in HD116657 (see Fig.1 (click here)).
While in the spectrum of HD108651 the lithium
lines dominate the absorption feature observed, in the two hotter stars
Li lines weaken and unidentified opacity occurs on the long-wave side
of this feature at about
. This opacity could not be explained
by means of adjusting the 
value of Fe I 6708.885Å
as increasing its would have lead to unacceptable
strengthening of the line in the spectrum of the cooler star.
Recently, Hack et al. (1997) also pointed to a similar lack of
opacity in the atomic data at 
in 
CrB. Thus,
the Li abundance estimated for the two hotter stars is rather an upper limit.
Figure 1: A close look at the Li 6708 region in HD116657.
Synthetic spectrum computations were made with
the following abundances 
, 2.80, 3.10
(from top to bottom)
Carbon - the abundance of carbon can be derived from two lines: C I 6683.970 which includes a slight blend of Fe I, and C I 6688.794 which stands alone. Three other C I lines are only complements of blends. One line, 6711.323 Å, though not-negligible in the computed spectra, is not present in those observed. A mild deficit of carbon in all three stars favors Am characteristics.
Nitrogen - seven lines could be identified in blends, but only one of them, 6722.610 Å, makes a significant contribution to the wing of Si I 6721.848. This enabled us to estimate nitrogen abundance. While normal in HD116657, nitrogen was slightly overabundant in the two other stars.
Aluminium - the lines of Al I significantly contribute to the absorption at 6696 Å and 6699 Å. Aluminium is moderately and slightly underabundant in HD108651 and HD116657 respectively, and is normal in HD138213.
Silicon - Si I 6721.848 causes a substantial absorption at the feature observed, which enables us to derive its abundance with accuracy. The other lines, 6683.161, 6696.044, 6706.980, 6719.609 and 6720.908 Å, occurring alone or in well-defined wings, confirm the derived value. Silicon is slightly to moderately overabundant in all three stars.
Calcium - Ca I 6717.681 dominates one of the strongest features in the observed spectra. Ca I 6679.352 complements the blend in the long-wave wing of the strongest Fe I line. Due to an unidentified opacity source within Ca I 6718 the abundance given in Table 4 (click here) is rather an upper limit as shown in Sect. 4.2. Nevertheless, the abundances derived confirm at least a slight calcium deficit in HD108651 and HD116657, while in HD138213 the abundance is normal at most. The calcium deficit underlines the Am characteristics of those two stars.
Titanium - is normal as estimated from Ti I 6680.133 which contributes significantly to the blend on the wing of the strongest Fe I line.
Iron - the procedure for deriving iron abundance was described in the previous section. Its abundance correlates with the other approximately 35 Fe I and Fe I lines. Iron might be slightly overabundant in HD108651, while it is normal in the two other stars.
Nickel - only in HD108651 are two lines identified with Ni I 6680.123 and Ni I 6700.890. The former Ni line contributes to Ti I 6680.133 and suggests a moderate overabundance. There is a minute depression in the noise at this place in Ti I line. The third Ni I line at 6711.575 Å in the observed spectrum is unsignificant. Thus, the value estimated represents rather an upper limit of the abundance.
Cerium, Samarium, Gadolinium - three lines of cerium, 6 of samarium and 3 of gadolinium are suitable minor complements of blends with other lines. The values derived are rather upper limits of the abundances.
The following analysis is based on data from Boesgaard (1987), Burkhart & Coupry (1989, 1991, 1997) and this paper as far as the Ca/Fe (the ratio of measured equivalent widths of Ca I\ 6718 and Fe I 6678 blends, i.e. comprising also e.g. Fe I 6677 line) and Li is concerned, and from Batten et al. (1989) as to the orbital elements. Consequently, one should be aware of the disadvantages of such data compilations from various sources, e.g. the sample is not homogeneous.
Note that the only observed Ca I 6718 line blends with a strong
Ti I 6718 line, as recently mentioned by Burkhart & Coupry
(1989). Unfortunately, there remains a lack of opacity in the atomic
data so that one cannot derive reliable Ca abundance from this line.
This follows from Fig. 2 (click here), where we plot Ca/Fe against 
.
The dashed line is the simulated synthetic Ca/Fe ratio calculated for
solar abundances with 
. One can see that it is
apparently less than the one observed for normal stars, pointing to the
mentioned missing opacity.
Nevertheless, one can see that the normal and Am stars are
well separated in such a plot and that the Ca/Fe ratio in normal stars
is not very sensitive to their and is mainly within
0.8-1.0. Such scatter is well in accordance with the
expected uncertainty in the equivalent width measurement. The latter
is usually 
so that the observed relative error in the Ca/Fe ratio
is 
.
We can see that three out of the four long period
Am binaries clearly occupy the bottom of Fig. 2 (click here) (see also Table
6 (click here)). This is a strong indication of their pronounced Am anomalies.
Figure 2: Ca/Fe ratio versus . Data are from
Boesgaard (1987), Burkhart & Coupry
(1989, 1991, 1997) and this paper. Squares stand for normal stars;
open circles for Am stars; full circles for our long period Am binaries and
16 Ori (which is moreover distinguished also by an asterix); dashed line is
calculated Ca/Fe behaviour for solar abundances
| HD108561 | HD116657 | HD138213 | |
| Ca I | 23.7 | 13.0 | 25.6 |
| Fe I | 95 | 73* | 56 |
| Ca/Fe | 0.25 | 0.18 | 0.46 |
We proceed to explore any possible dependence of Ca/Fe on
orbital parameters.
There might be a decrease of Ca/Fe with the orbital period, but it
is not convincing (Fig. 3 (click here)a). Apparently, HD138213
does not fit into such a pattern having 
.
However, notice that it is a special case of long period binaries
(with 
) having a circular
orbit. Such stars may not obey the favoured behaviour of increasing
metallicity with 
(Budaj 1997).
Nevertheless, looking at Fig. 3 (click here)b, we see a remarkable decline of
Ca/Fe with eccentricity.
To speak in more objective terms, we have calculated Pearson's linear
(
) as well as Spearman's rank order (
) correlation
coefficients together with their two-sided significances or p-values
(
, 
, according to Press et al. 1986). The
latter simply represent the probability of the appearance of a better
correlation coefficient than that found here under the assumption that the
quantities do not correlate at all. Generally, a p-value less than about
0.05 is accepted as strong support for the presence of a correlation.
We see in Table 7 (click here) that there is sufficient evidence that
Ca/Fe decreases with eccentricity.
Ca/Fe  | Li | |||
 | 
| | | |
| | -0.51 | -0.39 | -0.64 | -0.40 |
|
| 0.036 | 0.12 | 0.019 | 0.18 |
|
| -0.63 | -0.57 | -0.95 | -0.46 |
|
| 0.007 | 0.16 |  | 0.12 |
Figure 3: Ca/Fe ratio and Li versus eccentricity and .
Open circles stand for Am binaries from Boesgaard (1987)
and Burkhart & Coupry (1989, 1991, 1997) (16 Ori is
moreover distinguished by a special symbol); full circles are used for our
long period Am binaries; arrows denote upper limits. Typical error in Ca/Fe
is about 20% and in Li abundance about 0.2 dex (displayed in the top
right-hand corner of 3c,d)
Similar analysis was also performed in the case of Li
(see Figs. 3 (click here)c and 3 (click here)d). Typical errors of most Li abundances
are mainly due to the uncertainties in effective temperatures and
amount to about 0.2 dex (Burkhart & Coupry 1989).
We also observe here a tendency of
Li abundance to decrease with orbital period, but it is again not very
convincing. However, there is sufficient evidence that Li declines with
eccentricity (check Table 7 (click here) for corresponding significances).
This strongly supports our above findings concerning the Ca/Fe
behaviour because the anomalies in both quantities
behave in the same manner, namely, they increase with rising
eccentricity
and exhibit a less pronounced but similar tendency with orbital period.
Unfortunately, it is hard to distinguish from such sparse data whether
the latter behaviour is not affected by the former, or the opposite,
because in general there is a predominance of eccentric orbits at large
periods. Nevertheless,
this cannot be a mere coincidence. Rather, it is exactly the opposite of what
has recently been found in the peculiarity of Ap binaries by Budaj
et al. (1996) (see also Gerbaldi et al. 1985;
Budaj 1995), namely, that Ap peculiarity diminishes at larger
eccentricities. This means that apart from (1) orbital periods and (2)
frequency of occurrence among SB2's (Abt & Snowden 1973;
Abt & Levy 1985) these two basic subgroups of CP stars are
furthermore distinguished by their (3) eccentricities. This supports the
idea that there is a basic set of parameters connected with binarity which
determines CP characteristics. Such eccentricity effects could be accounted
for by a different degree of pseudo-synchronization, which is apparently
much higher for eccentric than for circular orbits at comparable orbital
periods. Following Budaj (1997) and Budaj et al.
(1996) this could not only lower the efficiency of tidal mixing in
corresponding Am binaries, but have a much more serious impact on
stellar magnetism, for example... Thus, based on the observed decline of Li
with eccentricity (or its marginal decline with ), a potential
Li deficit in HD116657, and the above-mentioned relationship between
binarity and CP phenomena we propose that the 16 Ori phenomenon might be due
to the suppressed tidal mixing. The latter, being more intensive in low
eccentric or short period orbits, suppresses diffusion and disperses a
potential Li cloud as well as any other abundance anomalies (including
Ca/Fe) below the superficial convection zone. This is not the case with 16
Ori or HD116657, which, having the longest (within the range
), and large eccentricities, could build relatively
pronounced abundance anomalies in their stable atmospheres.
Figure 4: Spectrum synthesis of the region 6675 - 6725 ÅÅ of the
program stars. Model atmosphere parameters are from Table 3. The derived
abundances are listed in Table 4. The synthetic "non-rotated"
spectra are for HD 108651 and HD 116657 and they are cut below 0.85 and
0.98, respectively
and eccentricity together with their
corresponding significances