Up: A UKST survey of

Subsections

# 3 Analysis

## 3.1 Spectral types

The reduced spectra were input into the STARLINK package, DIPSO [12, (Howarth et al. 1993)] for further analysis. Initially the aim was to provide approximate spectral types for all the spectra which had a sufficient signal. Because of our interest in identifying early-type objects and given the relatively low spectral resolution and signal to noise ratio of our spectra, the following scheme was used to guide the classification.

O-type: Presence of HeII lines at 4200 and 4541 Å (note that no stars of this type were identified).

Early-B-type: Relatively weak and narrow Balmer lines coupled with well developed neutral helium spectrum. The CaII line (at 3933 Å), MgII close doublet (at 4481 Å) and (if observed) the NaI doublet at 5890 Å should all be weak or absent (the interstellar components in the CaII and NaI lines should normally appear weak at this spectral resolution).

Mid-B-type: Stronger and broader Balmer lines coupled with a weak but still distinct neutral helium spectrum. The CaII, MgII and NaI features, if present, are weak (with, for example, the HeI line at 4471 Å being stronger than MgII).

Late-B/A-type: Strong and broad Balmer lines coupled with either a very weak or absent neutral helium spectrum. Normally moderate CaII, MgII and weak NaI features (with, for example, the MgII doublet being at least as strong as the HeI line at 4471 Å).

Late-type: Strong CaII and NaI lines with the G-band present. No neutral helium lines visible.

Planetary Nebulae (PN): Balmer lines in emission with underlying continuum being weak. Often the neutral helium lines and the [NII] doublet at 6548 and 6583 Å are also observed in emission. We have cross identified any object showing strong emission lines with entries in the SIMBAD database, and some of them are already known PN. Note that at the available spectral resolution, the newly identified targets could also be post-asymptotic-giant-branch stars such as LSIV -12 111 [3, (Conlon et al. 1993).]

Subdwarf (sd): Strong Balmer lines coupled with the absence of MgII or CaII lines expected in late-B/A-type stars.

These criteria were only used as guides as sometimes features were not identifiable (possibly due to a low signal-to-noise ratio) and in a few cases the spectrum characteristics were inconsistent. Additionally our classification as "late-type'', effectively covers all stars later than A-type. As our primary intention was to identify early-type stars, we have not attempted to further refine this classification.

The preliminary classifications are summarised for the three fields in Tables 3 to 5. Also listed are the mean stellar counts and a brief description of the spectrum. The counts refer to the wavelength ranges 3900 to 4600 Å in the blue and 5600 to 6800 Å in the red; the instrumental response is such that the extreme blue (3860-4000 Å) and red (6700-6900 Å) regions of the spectra had counts typically a factor of two less than these mean values. The gain of the CCD used in FLAIR is approximately 1 e-/adu, hence the tabulated values are close to the number of detected photons; the signal-to-noise ratio of the stellar continuum can thus be estimated directly from them. Although there is a correlation between the mean counts and the photographic V (or I) magnitude, there are some significant discrepancies. These may in part be due to errors in the latter but we believe are mainly due to a combination of three factors. The first is the positioning of the fibres on the plate itself - although the fibres have a 7 diameter on the sky, the semi-manual positioning method incurs placement inaccuracies. Secondly fibre to fibre through-put variation contributes significantly, and thirdly as field acquisition and rotation are achieved with five fiducial fibres fed to a direct viewing TV system, the positioning of the whole plate on the sky may cause differential discrepancies across the field.

The spectral descriptions are necessarily brief with the designation of lines as "strong'', "weak'' etc. being in the context of the whole dataset. In the case of field 456, the comments include both the red and blue spectra. The absence of lines may in some cases be due to the poor signal-to-noise ratio in the spectra; however an attempt was made to only search for lines that should have been visible. In Fig. 3, the range of quality of the spectra is illustrated for the field F456 (for which there are both blue and red spectra); star 5 represents the lower range of usable spectra and star 58 the upper range. In the comments, the following abbreviations have been used, low g: gravity appears to be less than that of a main sequence star and the target may be either a giant/supergiant or an evolved object; He+: target may have an enhanced helium spectrum. In the comments, a question mark indicates that the line identification is marginal and possibly suspect, whilst for a spectral type, it implies that either the quality of the spectra or inconsistencies between different criteria puts the classification in significant doubt.

In Field 456, four of the blue objects showed PN type features with HI strongly in emission often accompanied with HeI and SII emission lines. These were cross-identified with objects in the SIMBAD catalogue and all of them are previously known planetary nebulae (the PK designations are from [20, Perek & Kohoutek 1967).] The two objects in Field393 which also show PN features do not have known counterparts, and hence are new identifications. We have also searched for known objects within a 20 radius around the positions of all stars which have a V (or I, in the case of Field 456) magnitude brighter than 13. Four of the brighter stars in Field454 were found to be known objects, while the rest of the stars in the sample have no other identifications, within the SIMBAD database. Star number 26 in Field456 is LS4784 for which we already have high-resolution spectra [33, (Smartt et al. 1999)] and which appears to be a normal, distant, B-type star. Hence we included this star to test whether our methods would retrieve such an object. No attempt was made to cross-identify the rest of the sample with magnitudes below this cut-off. As these objects are generally unremarkable apart from being blue, they are unlikely to appear in any other catalogue.

 Figure 3: Typical spectra are shown for two stars in the Field 456. Star 5 represents the lower range of usable data and has a signal to noise ratio of approximately 10, whilst star 58 is amongst the best spectra obtained with a signal to noise ratio of approximately 60

For some of the stellar spectra, more detailed descriptions are appropriate and these are provided below:

Field 393, star 36: The Balmer lines are narrow and relatively weak and are compatible with an early-B-type classification. However, although the signal-to-noise ratio is relatively low, the absence of a neutral helium spectrum implies a later spectral type. The CaII at 3933 Å appears to be present but its strength cannot be estimated. Hence this star is either a late-B/A-type with a low gravity or, less likely given the absence of the G-band, a late-type object.

Field 393, star 45: The strong Balmer lines coupled with a marginal identification of neutral helium lines implies a late-B-type classification. However the SiII doublet near 4130 Å is very strong with SiII lines at 3954 Å and 4200 Å also present. Numerous other metal lines (including MgII and CaII) are present in the spectrum and we therefore classify it as Bp or Ap-type.

## 3.2 Early-type candidates

For the early-type candidates (designated at early-B, mid-B or late-B/A in Tables 3 to 5), an attempt has been made to obtain a more reliable classification and to estimate atmospheric parameters. Equivalent widths were estimated for the Balmer, H, H, H, lines by arbitrarily assigning the continuum level at 16 Å from the line centre. Line strengths were also estimated for the neutral helium lines at 4026, 4143, 4387 and 4471 Å (with the continuum defined at 8 Å) and metal lines, including MgII at 4481 Å and CaII at 3933 Å (with the continuum at 4 Å). The equivalent widths are listed in Table 6, with ":'' implying a marginal measurement, "p'' that a line is present but its strength cannot be reliably estimated and "p?'' that a line may be present. Where no estimate is provided, the line could not be identified at the signal-to-noise and resolution of the spectrum.

Assuming a normal helium to hydrogen ratio, the relative strength of the neutral helium and hydrogen lines can provide an estimate of the stellar effective temperature. We have used a grid of models generated with the ATLAS9 code of [15, Kurucz (1991)] and an LTE model atmosphere code to predict line strengths for these feature; further details of the methods can be found in, for example, [31, Smartt et al. (1996b).] The range of effective temperatures considered was from to  K with logarithmic gravities from 4.5 dex down to near the Eddington limit and a helium to hydrogen fraction of 0.092 by number was assumed. When calculating equivalent widths, the continuum was defined so as to be consistent with the observational measurements.

 Figure 4: Blue spectra for the early B-type candidates that we have identified. For comparison, we also show a FLAIR spectrum of LS4784 which has been previously confirmed to be a B2IV star from high resolution data [33, (Smartt et al. 1999)]

Over this range of atmospheric parameters, the theoretical predictions for the equivalent width of the Balmer, H and H, lines normally agree to better than 5% (we did not consider H as it is blended and is in a spectral region, where the counts were normally low). Given the uncertainties in the observational data, we have therefore used just the mean in the equivalent width of the two lines (designated ) in our comparisons. We have also assumed that this would also be appropriate to the equivalent width of either line (when for example the measurement of the equivalent width of the other line was not possible).

The ratio of the strength of the neutral helium lines to  as a function of effective temperature is illustrated in Fig. 5, for the HeI line at 4026 Å. Two logarithmic gravities, 3.0 and 4.0 dex, were considered; ratios are only available up to an effective temperature of  K for the former, due to the effect of radiation pressure. This ratio does not depend significantly on the adopted gravity. For example in the lower temperature regime a change of the gravity by 1.0 dex would affect the temperature estimate by normally less than 1000 K, while at higher temperatures, the ratio becomes temperature insensitive but again the limit does not significantly depend on gravity.

 Figure 5: The ratio, R(4026), of the equivalent width of the HeI line at 4026 Å to the mean equivalent width of the H and H lines () is plotted as a function of effective temperature. Ratios are shown for two logarithmic gravities, viz. solid line - log g = 3.0; dotted line - log g = 4.0

The behaviour of the ratios for the other observed helium lines is qualitatively similar and hence they can be used independently to estimate (or, at least, set a lower limit on) the effective temperature. Then the mean hydrogen line strength, (H), can be used to estimate the gravity. This procedure was employed for the targets listed in Table 6. However it became apparent that whilst the HeI lines at 4026, 4143, 4387 Å gave consistent results (within the observational uncertainties), that at 4471 Å led to consistently lower temperatures estimates (by typically 2000 K). This discrepancy probably reflects a systematic over estimate in the theoretical equivalent widths for this line; hence we gave this ratio lower weight when finalising our atmospheric parameters estimates, which are summarised in Table 6.

We note that, even for the best observed stars, there will be considerable uncertainties in the atmospheric parameters. Realistic errors estimates would be at least  K in effective temperature and in the logarithmic gravity. The uncertainties for the lower quality spectra could be larger. However, this is still sufficient for us to classify a star and, in particular, decide whether it is suitable for further high resolution spectroscopy. Figure 4 shows the blue spectra of the stars that we have identified as early-types in Table 6. Also shown for comparison purposes is the FLAIR spectrum of LS4784, a typical early-type star (see Sect. 3.1).

Below we discuss the atmospheric parameters for each group of preliminary spectral-types. To simplify the discussion, stars are represented by two numbers (in square brackets) representing the field and the star:

Early-B-type candidates: For our ten candidates, eight appear to be in the high effective temperature limit of the line ratios and we are therefore only able to give lower limits for both the temperature and gravity. For four targets ([393,03],[393,13],[393,32] and [454,10]), these values should be secure; for the other four targets, uncertainties in the line strength implies that it is possible (but unlikely) that these stars might have lower temperatures than  K. Assuming that these stars are main (or near) main sequence stars with logarithmic gravities less than  4.2 dex [5, (Claret 1995),] then we can constrain the upper values of their effective temperatures. These would range from  K for [393,55] to  K for [393,03] and are consistent with the lack of an observed HeII spectrum. For example, at an effective temperature of  K and a logarithmic gravity of 4.0 dex, our model atmosphere calculations of the HeII line at 4541 Å predicts an equivalent width 0.2 Å, which increases rapidly with temperature.

Two stars have ratios that imply effective temperatures of approximately  K but in these cases the observational uncertainties are consistent with them being in the high temperature limit. However if these are main sequence targets, then their effective temperatures cannot be greater than approximately  K as this would lead to too large an estimate for the gravity. Encouragingly, for all ten targets the estimated atmospheric parameters are consistent with our original spectral classification, although our preliminary identification of [456,39] as possibly helium rich is not supported by the equivalent width estimates. In fact the helium to hydrogen line ratios for this star are all close to the high temperature limits. We further checked that none of these objects have been identified in previous catalogues and found no possible known counterparts in SIMBAD.

Mid-B-type candidates: Atmospheric parameters could be estimated for all the stars apart from [393,24]. Its spectrum was of low quality and although the helium lines appear to be present, they could not be reliably measured. We have taken the conservative approach that their equivalent widths were less than 1 Å to estimate limits. Although the HeI line strengths for the other stars are often poorly determined, they vary rapidly with the effective temperature. Hence the estimates of the atmospheric parameters should normally be reliable.

Our effective temperature estimates range from to  K and indicate that our selection criteria have led to a homogeneous sample. However although characterised as mid-B spectral type, the adopted criteria have also included some of the later-B spectral types. This reflects the difficulty of discriminating by eye between a range of weak neutral helium line strengths. Most of the stars appear to be on (or near) the main sequence but a significant minority (eight stars) have lower gravities. These are likely to be a mixture of relatively young giants and old evolved stars (such as the post-asymptotic-giant-branch - post-AGB - objects discussed by [17, McCausland et al. 1992).] With the current data it is not possible to distinguish between these possibilities. The lack of emission in the Balmer H line for the targets in the Field 456 would be consistent with the former, although we note that such emission is not present in the spectra of all post-AGB stars.

Late-B/A-type candidates: For all these stars the neutral helium lines were either absent or too weak to measure. We have therefore assumed that they have equivalent widths of less than 0.5 Å. This should be appropriate for the spectra with low counts but may be too conservative for stars, such as [393,28]; however it was not felt that the data warranted assigning different equivalent width limits to each spectrum as these would, in any case, also have been wavelength dependent. The limits on the atmospheric parameters estimates are given in Table 6 and the stars appear to fall into two groups. Five targets have maximum gravity that are consistent with them being main sequence targets, whilst the other two have lower gravities. As for the previous group, these may either be relatively young stars evolving from the hydrogen main sequence or older post-AGB objects.

Up: A UKST survey of

Copyright The European Southern Observatory (ESO)