A3 'outlier7-vsn-2' gives a general overview of outlier lines with the aid of
Kurucz's line- and gf-values. It contains tests and results of investigations
into wavelengths and gf-values together with some wavelength areas including all
(known) lines with an equivalent width > 5 milli Angstrom (.005 x 0.1nm).

What is the reason for the so-called 'outliers'? In the following we investigate
a number of large differences between model- and observed-values of the
equivalent widths (0.1pm) of some lines.

From an inspection of the files "ltemodel", "observed", "signifi" we find a
number of instances where model- and observed-values diverge appreciably.
(In actual fact we should test all lines for this phenomenon.) We take 10% as a
lower limit of contribution.
In some instances gf-values used may be too low. or too high with respect to the
accuracies estimated for them. In the same
way the estimated accuracies of the observed values of the equivalent widths may
be too optimistic. We find however that another identification for high Teff
observations may be needed.

Below is a selection of lines that have problems in modelling equivalent widths
in comparison with observed values, as given in the files:
"ltemodel", "observed", "signifi". We must therefore compare outliers as
indicated in the various files, and we see that in some cases extra modelled
lines need to be taken into account, in other lines it seems that the gf-values
used are either somewhat too low, or too high, and that we should also check
when the given accuracies of the measurements and/or of log(gf) are too
optimistic.

This is possibly the case in the following lines ( lamda in 0.1nm):
Numbers indicate observation number as e.g. 5) for obs5

5051.63 A trial of 5052.17  600 for high temperature is equiblend at 6500 K.
We adopt this identity for 6,8-12
5125.23 A trial of 5125.12 2600 for high temperature gives good results.
We adopt this identity for 8-12
5127.36 A trial of 5127.87 2601 for high temperature gives good results.
We adopt this identity for 13,15-26,28
5243.78 A trial of 5243.46 2401 for high temperature gives better results, but
identity not certain. We use for 13,17
5432.55 .97 operate at lower temperatures, and 5432.95 2601 at higher
temperatures
5444.59 A trial of 5445.05 2600 was already incorporated. We now delete the
identity 5444.59 and keep 5445.04 for 5
5476.90   check gf, is conform Kurucz     (We take no action)
5493.50 A trial of 5493.88 1400 gives better results, wavelength more consistent
cf. original publication. Adopt for 5
5501.46 A trial of 5502.07 2401 for high temperature gives good results.
We adopt this identity for 6,27
5593.74 A trial of 5593.30 1301 for high temperature gives a slight improvement,
but is ident correct? Keep both for 6,14
6246.32 A trial of 6245.64 2101 for high temperature gives good results.
We adopt this identity for 6,7,14

A too low tolerance character B, C etc may lead to high chi-square for a line.
Some values of gf seem low resp. high
5172.68 -- gf tolerance of 15 % is possibly too optimistic?
(In all these cases we take no action)
5183.60 -- gf tolerance of 10 % is possibly too optimistic?
5325.55 -- gf tolerance of 25 % is possibly too optimistic?
5737.06 -- gf tolerance of 15 % is possibly too optimistic?

Blends may play a role, lines are approximately 1 x  0.1nm wide:
(We take no action for these lines)
5349.74 blend with 5349.76 2301 or  identification? try 5350.36 2301
5353.37 blend with 5353.25
5357.19 blend with 5357.35
5371.49 blend with 5371.33 .44 .60
5497.52 blend with 5497.41
6413.32 blend with 6413.11

We follow the development along the temperature scale, and see that in a number
of instances extra lines have to be calculated and the values of EW have to be
combined with the EW of the identified (main contributing)line to an estimated
accuracy of 10% in the model contributing to the equivalent width. The table
shows equivalent widths (0.1pm) for lines with lamda (0.1nm), element number and
indicated model-atmospheres (function of temperature). In the  columns under the
used models we give calculated model EW's (0.1pm) followed by their sigma's
({Delta}EW) due to the accuracy of gf.

Note: The models used are Teff=9250, 8250, 7250, 6250, 5250, and 4250 K
                          logg= 2.0,  1.0,  0.5,  0.5,  1.0, and  0.0
                            vt= 6.0,  6.0,  6.0,  6.0,  6.0, and  6.0 km/s