To obtain identifications and accurate coordinates, two types of follow-up observations were obtained. For those objects were no finding chart was available at the time, CCD images were obtained with interference filters centered ON and OFF the H emission line, the images being centered on the published coordinates (see Sect. 3.1). This allowed to recover one (or more) emission-line objects in the field. For other PNe, the identification was taken from the literature. For all the identified PNe candidates, entire Schmidt plates were scanned, covering the whole LMC, to determine accurate coordinates and a rough broad-band magnitude.
Images of about 50 PN candidates have been taken, from 1990 to 1991, with several telescopes in ESO La Silla (NTT, 2.2 m and the Dutch 90 cm) with respective fields of view of , and . A few objects turned out to be resolved in the H images (SMP011, SMP017, SMP024, SMP030, SMP060, SMP086, J26). All candidates in these fields could be recovered, with the exception of Sa108, Sa119, Sa125 and Sa127 which are therefore rejected as PNe candidates (see Sect. 2.2). Many broad band images were also recorded, just before spectroscopy, to allow a proper positioning of the objects on the slit. These images also provide high quality finding charts.
We have scanned 5 SERC J (Blue) and 3 ESO (Red) Schmidt plates (to cover a maximum area of the LMC) with the MAMA (Machine à Mesurer Automatique) at Paris Observatory. MAMA is a linear scanner of 1024 diodes mounted on a stable marble table in a cooled room and provides excellent stability and accuracy (see Berger et al. 1991). The dynamics is 12 bits, therefore with very deep exposures all the stars of or brighter are already saturated. For stars brighter than = 12 the saturation is too strong, and then no precise centering can be done. This dynamics is however perfectly adequate for copy Survey Plates such as the ones we used, where the intrinsic dynamics is anyway limited. The J plates, with a field about 6 by 6, are mapped in a mosaic of 33 by 33, pixel images, while the mosaic is 28 by 28 in the case of the red plates. The total J image is pixels and occupies about 2.2 Gbytes of disk space. The pixel size is 10 m or 067. The main characteristics of these plates are shown in Table 1 (click here). The coordinates are the (, ) of the plate center in J2000 equinox derived from our astrometry. The date of observation is also given. The last column presents respectively the initial and the final number of PPM (Röser & Bastian 1988) stars taken into account for the astrometric reduction (see below).
Table 1: Parameters of the used Schmidt plates
The published coordinates for the various PNe candidates are given in several
equinoxes (1950, 1975 and 2000).
Most of the astronomical reductions have probably been done (although not
explicitly said, except by Jacoby 1980) with earlier catalogues in the FK4
system, whose accuracy of positions and proper motions are no longer
satisfactory.
Therefore, in order to make the comparison with our new positions, we
converted all the coordinates to the J2000 equinox with the matrix suggested
by Lieske (1979).
On each SERC or ESO plate, all the PPM references (in equinox J2000 and in
the FK5 system) stars have been searched (the mean number is about 200), and
their positions on the plates determined in ().
For all the (bright) PPM stars we used a spike centering method.
The Schmidt plate is then mapped with a bidimensional 3-degree
polynomial in order to transform the () to (, ) or to
make the reverse transformation.
The program is iterative, in the sense that it rejects the reference stars
which deviate by more that 3 from the calculated position and
recalculates the new coefficients of the fit.
On the average, the program iterates only 3-4 times and rejects only about
10 stars randomly positioned on the plate (and not at the outer part as would
be the case for a bad fitting on the borders).
Figure 1: Differences between our new positions and
previously published ones (new-old)
Figure 1 (click here) shows the differences between our new positions and the
ones given in previous papers.
In the case of Sanduleak et al. (1978) objects the original coordinates
have been taken only if better one's were not available.
The difference can be as large as and therefore the
identification is made impossible in such dense fields from coordinates
alone.
We used the same symbols for Figs. 1 (click here), 2 (click here) and
3 (click here): squares (Sanduleak et al. 1978, 1984; and
Meatheringham et al. 1991a,b); crosses (Jacoby 1980);
circles (Morgan & Good 1992; Morgan 1994).
Figure 2 (click here) is an enlargement of the central part of Fig. 1 (click here)
and shows only the points with better original coordinates (Jacoby 1980;
Morgan & Good 1992; Morgan 1994; and Meatheringham et
al. 1991a,b). The accuracy is still not satisfactory.
Moreover we see a systematic difference in coordinates, both in and
, of respectively 1 and 2.
The bulk of data is not centered at the origin indicated by the cross.
Figure 2: Enlargement of the previous figure.
Systematic difference in positions (new-old) of about is
clearly visible.
The cross indicates the center at position 0,0
The origin of this discrepancy is probably multiple.
It could be due to the use of film copies for some earlier measurements, to
systematic differences between astrometric catalogues (PPM versus SAO,
Perth70, ...), to differences between reference systems (FK5 versus FK4,
amounting to about in and in
(Schwan 1988)), to unaccounted proper motions of standard stars, etc ...
The measurements presented here are however based on the best available
material (glass copies, digitized images, PPM stars).
The () positions of all objects have been measured with the 2
completely different methods.
The first method is a Gaussian centering fit inside the MIDAS Image
Processing Package of ESO.
In very crowded fields, like the Bar, and/or with tightly blended stars, the
fit does not converge properly and this measurement has not been retained.
The second is a home-made program (Alard 1995) which separates the blended
stars, measures the central position and does the photometry of all the stars
found in the field.
This method failed only in a few peculiar cases were no position and
photometry could be derived (7 objects in total from the BAR: SMP047, SMP048,
J12, J22, J26, Mo15 and Mo23).
These cases are indicated by a colon in Table 4 (Col.
3 after the flag number).
Figure 3: Half difference between the two centering methods.
No systematic effect can be found
Figure 3 (click here) shows that for all the objects, which could be measured
by the two methods, the two determinations agree quite well.
They gave similar results with a difference always less than 030
and a of about 009.
Both methods have advantages and disadvantages.
The first one, with manual identification, is indispensable when an object
has poor coordinates but can be identified on a finding-chart.
Its limits are reached in crowded fields not because of identification
problems, but because the centering algorithm in MIDAS is not able to
properly distinguish between sky background and adjacent objects.
The second, automatic method is optimized for these difficult cases but,
because no a-priori identification is made, the PN candidate cannot be
selected out from the outputs when its initial coordinates are too poorly
determined.
Both methods are thus complementary and have been used in sequence.
We first identified the candidate, to center it in an extracted sub-frame of
pixel (in order to produce the finding-charts).
We then use first and second methods in this sub-frame only to determine the
position and the magnitude of the PNe.
In all the cases where the first method works properly, the two ()
position determinations agree perfectly.
Therefore, in order to improve the accuracy and to reduce dispersion, the two
determinations have been simply averaged.
The final PNe coordinates are directly determined from the proper ()
to (, ) transformation, and averaged if several positions
were measured on different Schmidt plates.
In the case of multiple determinations, if one is coming from a plate corner,
check of the reliability is done automatically and the point is
rejected if its distance to the mean value is greater than 030
(see the final position Table 4
in Appendix).
The definition of a corner is the area within 4 cm, in both directions, of
the scanned border zone.
Elsewhere, no distortion problem has been noted.
To look at the internal consistency, we use PNe in common on all plate
pairs.
The mean position error of the PPM stars at the LMC declination is
011 and the estimated residual systematic deviation between PPM
south and the FK5 system has a typical size of 005.
The external consistency has been checked over all overlapping PPM stars on
each pair of plates, and no systematic error could be found nor border
fitting problems.
The residual errors are of the order of 02.
Figure 4 (click here) shows the results of this test for the plate pair
SERC 056J/SERC 085J.
Table 2 (click here) lists the number of PNe in common, as well as the corresponding sigma in coordinate differences. This test was done on the SERC plates only (overlap on ESO
Figure 4: PPM position differences on Plates SERC 056J and SERC 085J
plates is too small) but ensures the accuracy and consistency of the whole astrometry. Figure 5 (click here) represents the histogram of coordinates differences for all the overlapping PNe on the 8 plates. Very few objects fall outside 05, and all the bad determinations came from a PNe pair with at least one lying in a plate corner. The average error is about 03 and the corresponding mean standard deviation is about 017. But these objects are taken from the borders of the plates, and therefore the expected errors in the inner part of the plates should be much smaller.
Figure 5: Histogram of the position differences for all the PNe present on
at least 2 Schmidt plates
The automatic centering program (Alard 1995) also furnishes magnitudes, by
profile reconstruction of point-like sources and integration inside the
profile.
This method has the advantage to give good results even when the central part
of the star is saturated.
We have therefore directly the magnitude (in B band) for all the PNe where
crowding was not too important.
Calibration was obtained from B and V photometry of faint stars
in the LMC (Linde et al. 1988) for very faint sources and from the Atlas of
star clusters with color magnitude diagrams in the Magellanic Clouds
(Alcaino 1975) from which we extracted 12 photometric sequences in LMC.
In total we used about 250 stars, covering a large magnitude domain from
= 9.1 to = 22.7.
B and V photometry has been used to produce
magnitudes for the standard stars, following the recipe given Blair &
Gilmore (1982): .
Figure 6 (click here) shows the calibration curve with all the standard
stars (except a few rejected outlying points due to confusion, saturation, etc...) and a fitted 6-degree polynomial, taken from two different plates
(SERC 056J and SERC 085J).
No systematic effect can be seen between these two plates.
The linear domain extends from to 22 with a precision of about
0.20 magnitudes which is the expected internal error from the published
measurements.
The stars brighter than are saturated, due to the limited dynamical
range of MAMA.
This phenomenon is clearly seen in Fig. 6 (click here), where the slope
at brighter magnitudes is smaller, and therefore the precision is diminished.
But a magnitude determination is still possible down to 10-11th magnitude.
From 114 stars which have been measured more than once, we derive an internal
error smaller than 0.25 magnitude.
Only for a few objects is the uncertainty larger, possibly due to peculiar
local problems.
Figure 6: Schmidt plates calibration