A&A 370, 436-446 (2001)
1 - Observatorio Astronómico Nacional, UNAM, Apdo. Postal 877, CP 22800, Ensenada B.C., México
2 - Instituto de Astrofísica de Andalucía (CSIC), Apdo. 3004, 18080 Granada, Spain
Received 13 December 2000 / Accepted 14 February 2001
We have performed a photometric survey of open clusters in the third Galactic quadrant in order to study the star formation history and spatial structure in the Canis Major-Puppis-Vela region. In this paper we describe a catalogue of CCD photometry of approximately 65000 stars in the fields of 30 open clusters. The data were obtained and reduced using the same telescope, the same reduction procedures, and the same standard photometric system, which makes this catalogue the largest homogeneous source of open cluster photometry so far. In subsequent papers of this series, colour-colour and colour-magnitude diagrams will be presented which, amongst other uses, will allow the determination of an homogeneous set of cluster reddenings, distances, and ages that will constitute the observational basis for our studies of the spatial structure and star formation history in the third Galactic quadrant.
Key words: techniques: photometric - stars: fundamental parameters - Galaxy: open clusters and associations: general - Galaxy: structure - ISM: dust, extinction
In the case of Galactic structure studies the existing compilations of open cluster parameters have allowed some understanding of the structure and development of the Galactic disc (Vogt & Moffat 1972; Janes & Adler 1982; Alfaro et al. 1991; Twarog et al. 1997; Carraro et al. 1998). However, many of the conclusions drawn from open cluster data have been if not controversial, at least affected by considerable uncertainty. Among the reasons for this situation are the lack of homogeneity of derived open cluster fundamental parameters (reddening, distances, ages, metalicities), as well as the small number of studies of certain kinds of clusters (e.g. old clusters and distant clusters).
The Catalogue of Open Cluster Data (COCD) (Lyngå 1981, 1987), which includes information for 1151 clusters (distances for 422) has been the observational basis for many Galactic structure studies. However, the parameters presented in the COCD were derived by different authors using a variety of instrumentation, techniques, calibrations, and analytic criteria, and therefore result in a rather inhomogeneous set. An analysis of the precision expected due to the effects of these inhomogeneities has been performed by Janes & Adler (1982) who have found a typical difference of 0.55 mag in distance moduli determined in different photometric studies. But even this large scatter in distance modulus determinations might be underestimated since several clusters have much larger ranges in their estimated distances. Cases like NGC 2453 which has distance estimates from 1500 to 5900 pc are not rare. These problems have been known to the astronomical community for a long time and have led to several attempts to produce more consistent sets of open cluster parameters. Loktin & Matkin (1994) reanalysed UBV photometry of 330 clusters and obtained a more homogeneous, although less numerous, catalogue of open cluster reddenings, distances and ages. However, Loktin & Matkin (1994) used three different sets of isochrones in their age and distance determinations which is likely to affect the internal precision of the catalogue. More recently, Dambis (1998) has given another contribution towards an homogeneous set of cluster parameters by redetermining the reddenings, distances and ages of 203 open clusters, younger than to avoid chemical composition effects, using a single set of isochrones and an empirical ZAMS.
Although the works of Loktin & Matkin (1994) and Dambis (1998) have generated improved sets of cluster parameters by using the same kind of data ( photometry), and were analysed in an homogeneous fashion, at least two effects still contribute to degrade the precision of their results. The first one is that both works used data from different sources, nominally on the system but in fact were calibrated using various sets of standard stars which can produce considerably different photometry (see Table 8 in this paper, and also Bessel 1995 for a detailed comparison of some versions of the system). The second and perhaps most important factor is the effect of photometric depth. Photoelectric photometry is usually limited to mag, which in many cases is not enough for reliable distance determinations: evolutionary effects will produce redder colours and lead to underestimated distances. Also, for the younger clusters, the nearly vertical shape of the upper main sequence will introduce large uncertainties in distance determinations via ZAMS fitting.
In the light of the above mentioned problems with the current open cluster parameter compilations and of our interest in studying the star formation history and spatial structure of the Canis Major-Puppis-Vela region, we have performed a deep CCD photometric survey of open clusters using the same instruments, reduction methods and standard stars. In this paper we describe the photometric database. Reddenings, distances and ages will be determined in forthcoming papers using the same Zero Age Main Sequence (ZAMS) and evolutionary models.
In the next sections we describe the observations and data reductions, and present the photometric database. Later, in Sect. 6 a comparison with previously published photometry is shown. Finally, Sect. 7 is a discussion of the interstellar extinction law in the direction of our sample.
Data were acquired during five nights in Jan. 1994 and ten nights in Jan. 1998
at the CTIO 0.9 m telescope. Due to technical difficulties (focus), and to
some non photometric nights, only 30 clusters were observed. The
typical seeing during both runs was about
although a few
images presented higher values (
The cluster names, coordinates and observing run are presented in
In both runs, images were taken with a Tek CCD and the standard set of filters available at CTIO. The /pixel plate scale resulted in a field of view of . Images were acquired using the CTIO ARCON operating in Quad mode (http://www.ctio.noao.edu/instruments/arcon/ arcon.html). The gain was set at 3.2 e-/adu and the readout noise was determined to be 4.0 e-.
Besides the cluster fields, a number of standard star fields (Landolt 1983, 1992) were observed for calibration purposes. For the bias level and flat field corrections, zero second exposures and blank sky exposures in all filters were acquired each night. In the 1998 run, several short and long dome exposures were obtained with the purpose of creating a mask for the correction of shutter effects.
Photometry was performed using the IRAF/ DAOPHOT (Stetson et al. 1990) package. Standard stars were measured via aperture photometry with the APPHOT task. A 26 pix ( ) radius aperture was adopted since it included virtually all the stellar flux in all images as indicated by a growth curve analysis (Howell 1989). Magnitudes in the cluster fields were obtained following the standard procedures for PSF determination and fitting within IRAF/DAOPHOT. Due to the large size of the CCD chip, a quadratically variable PSF had to be used. About 60-70 well distributed PSF stars were selected by hand in each frame and were also used in the aperture correction determinations. The variable PSF was not able to adequately model the external regions of the images, so measurements for stars separated less than 150 pix from the edges were not used, which limited the useful field of view to . Stars with a goodness of fit parameter, , greater that 2.5 and with error estimates greater than 0.1, as output from ALLSTAR, were also dropped out. At the end of this process a list of aperture corrected PSF photometry was obtained for each image. In total, 4.095 standard star aperture measurements and 2.096.414 cluster field PSF measurements were performed.
Traditionally, extinction is determined assuming Bouguer's law,
Mi=Mi0 + KiX, where Mi is the i band magnitude measured at
airmass X, Mi0 is the magnitude one would measure outside the
atmosphere, and Ki is the extinction coefficient for band i.
Following this model, the extinction coefficient is usually determined
as the slope yielded
by a simple linear least squares fit of Mi vs. X. Ignoring the
contribution of higher order extinction coefficients (Young 1974),
the main inconvenience of this approach is that the slope
determination should be based on many measurements
of a single star. Since Mi0 will be different from star to star
one can not use different stars on a simultaneous
determination of Ki in a direct manner. The use of an average
determined from a few measurements of different stars is not a good alternative
because the errors of individual determinations can be very large,
producing an uncertain Ki.
One way to use different stars in a simultaneous determination of the
extinction coefficients is to employ Bouguer's law in a
modified fashion. If we consider two measurements of a star at two
different airmasses, we can write
difference of the star's magnitudes measured at different airmasses and
is the difference of the airmasses (the i index is dropped
for simplicity). Since the M0 term is no longer present,
this equation is independent of the star under consideration and
therefore measurements from different stars can be used in a
simultaneous determination of the extinction coefficient.
When there are N measurements at different airmasses for
a certain star the question of which of the possible N(N-1)/2
should be used arises, since only
N-1 of them are independent.
One of the most natural
solutions would be to use the lowest airmass measurement as a
reference to be subtracted from all the other measurements so that
the range in
would be maximised.
The drawback of this choice is that any error in the reference's
measurements will be introduced in a systematic way in all the
differences. Since in principle any of the possible
could be used to determine the extinction
coefficient, we have decided to use all the possible differences
together in the same plot expecting that a bad measurement should
produce outliers distinguishable from the dense cloud of good points.
|Figure 1: Top: linear fits for the 1994 run extinction determinations. Bottom: the same for the 1998 run|
|Open with DEXTER|
Figure 1 shows the linear fits
of the extinction determinations for the 1994 and
1998 runs. The fits were performed with the origin fixed to zero.
A per night analysis showed that within each run the extinction
coefficients remained constant (with a precision better that 0.01 mag), and
therefore data from different nights were used together to determine
average extinction coefficients for each run.
The results are summarised in Table 2, where a noticeable
decrease in extinction of about 0.03 mag is appreciated from 1994 to 1998.
As previously mentioned, photometry for
several stars from Landolt (1992)
was obtained in order to determine the transformations between the
instrumental and the standard system.
We assume that the instrumentation was stable enough so that besides zero
point variations, the transformation coefficients did not change within an
relations between the reference night and the standard system were
taken to be of the form of Eqs. (1) to (5).
|(Eq. (1))||(Eq. (2))||(Eq. (3))||(Eq. (4))||(Eq. (5))||(Eq. (7))|
The photometry lists obtained as described in Sect. 3.1 were corrected from atmospheric extinction using the coefficients in Table 2 and transformed to the system of the reference night. Since several measurements per band were available for each cluster, the final instrumental photometry is an average of the individual measurements weighted by the internal errors output by the ALLSTAR task. The final internal errors assigned to each star were taken to be the error of the average. When only one measurement was available the error was taken to be the one output by ALLSTAR. The average magnitudes were then transformed to the standard system using the coefficients from Table 3.
As previously mentioned, nights 6, 7 and 10 of the 1998 run were considered non photometric due to occasional cloud coverage. During the clear periods of these nights, long exposures of several open clusters were acquired (NGC 2571, NGC 2311, NGC 2343, NGC 2432, NGC 2635, and Rup 18). The photometry for these frames was then transformed to the instrumental system by comparison with shorter exposures taken on photometric nights.
The final catalogue of calibrated data which includes photometry and error
estimates for 64.619 stars in at least three of the five bands will be made available electronically at the CDS.
Table 5 summarises the contribution in number
of measurements, photometric depth and field of view of this study
relative to previous works.
In Table 5, N is the number of stars measured in this
and Field are the number of
stars, limiting magnitude and the field of view
covered in previous works that used the technique presented in the
Data column (pe - photoelectric; pgr - photographic; ccd -
CCD). The column Other indicates the existence of other
studies performed using other techniques that, due their lower
precision or number of stars, were not included in the comparison.
work the limiting magnitude is
(except for Haf 18
and Haf 19 where
and the field of view
Table 5, the clusters marked with the same symbol have
small angular separations and appear in the same image. For these
clusters the number of stars refers to the whole image and is
indicated only once. All the other empty fields represent unavailable data.
The data from previous studies was obtained from the WEBDA (http://obswww.unige.ch/webda/)
open cluster database of
Mermilliod (1988, 1992).
The WEBDA database has allowed this and other analysis throughout this
work to be performed in a reasonable time.
|Haf 18b||2304||50||17||ccd||pe, pgr|
|Haf 19b||70||18||ccd||pe, pgr|
|NGC 2439||3477||120||18.5||ccd||pe, pgr|
|NGC 2453||2605||356||19||ccd||pe, pgr|
In the process of arriving to the instrumental magnitudes, CCD
images were processed, PSFs were fit, aperture and extinction corrections
and night to night photometric zero point offsets were
The whole process is highly complex and full of subjective decisions,
which makes any formal step by step error treatment virtually impossible.
Nevertheless, we can analyse how the these processes affect the
photometric precision by analysing the dispersion of measurements.
|Figure 2: Standard deviation of the instrumental magnitudes for stars of NGC 2571 observed 10 times or more|
|Open with DEXTER|
Figure 2 shows the standard deviation (not the error
of the mean value) of the instrumental magnitudes for stars
in the field of NGC 2571 which have 10 or more measurements and
therefore represent the error of an individual measurement.
Table 6 summarises the data in Fig. 2
for a number of magnitude ranges.
|14 - 16||0.033||0.021||0.017||0.016||0.018|
|16 - 18||0.042||0.038||0.032||0.026||0.031|
Because NGC 2571 was observed in four different nights, the dispersion
also includes the effects of uncertainties in the night to night
variations of the transformation zero point.
The errors of the averaged data are represented by the error of the
mean which are smaller than values in Table 6,
depending on the number of measurements used. Most stars in the
catalogue have three measurements per band, although there are cases like
NGC 2571 and Rup 18 that have been observed up to seventeen times per band and
whose errors are considerably smaller (about one fourth of the
The total photometric errors for a single measurement have been estimated by quadratically adding the internal errors from Table 6 to obtain the errors in each colour and then quadratically adding these errors to the residuals of the standard transformation from Table 3. The total errors are presented in Table 7. The values in this table should be regarded as upper error limits for our photometry since they refer to individual measurements. Also, when constructing the colour indexes some contributions to the magnitude errors tend to cancel out instead of increasing the uncertainty, as we have assumed by quadratically adding the errors in each band. Finally, because only one long exposure per band was acquired for NGC 2571, the error analysis has been performed using photometry from short exposure frames, therefore leading to overestimated errors in the faint end.
In Tables 8a and 8b, is the difference between our photometry and previous one in the sense of this work - previous work, is the standard deviation of , N is the number of stars used in the comparison, and Data is the kind of compared data: Pe - photoelectric, Pgr - photographic, and CCD - CCD. Because of the large number of studies involved, we have limited the analysis to the average differences and their dispersions. Some stars that presented large discrepancies possibly due to bad identifications, variability, or contamination from neighbouring stars (affecting non-psf studies) were not used.
The comparisons in Tables 8a and 8b
show that in general our photometry does agree with the one from other
studies. Cases were the agreement is not so good usually correspond to
comparisons with photographic data (ex. Cz 29 and Haf 10).
There are significative differences relative to the photoelectric photometry of
Seggewiss (Seggewiss 1971), but there is good agreement with
other data for the same clusters.
|References for Tables 8a and 8b|
|1||-- Moffat & Vogt (Moffat & Vogt 1975)||16||-- Subramaniam & Sagar (Subramaniam & Sagar 1999)|
|2||-- FitzGerald & Moffat (FitzGerald & Moffat 1980)||17||-- Hassan (Hassan 1984)|
|3||-- Vogt & Moffat (1972)||18||-- White (White 1975)|
|4||-- FitzGerald & Moffat (FitzGerald & Moffat 1974)||19||-- Ramsay & Pollaco (Ramsay & Pollacco 1992)|
|5||-- Moffat & FitzGerald (Moffat & FitzGerald 1974)||20||-- Becker et al. (Becker et al.(1976)Becker, Svolopoulos, & Fang)|
|6||-- Munari et al. (Munari et al. 1998)||21||-- Moffat & FitzGerald (Moffat & FitzGerald 1974)|
|7||-- Labhart et al. (Labhart et al. 1992)||22||-- Moffat & FitzGerald (Moffat & FitzGerald 1974)|
|8||-- Munari & Carraro (Munari & Carraro 1996)||23||-- Mallik et al. (Mallik et al. 1995)|
|9||-- Clariá (Clariá 1973)||24||-- Lindoff (Lindoff 1968)|
|10||-- Seggewiss (Seggewiss 1971)||25||-- Eggen (Eggen 1974)|
|11||-- Clariá (Clariá 1972)||26||-- Havlen (Havlen 1976)|
|12||-- FitzGerald et al. (FitzGerald et al. 1990)||27||-- Jorgensen & Westerlund (Jorgensen & Westerlund 1988)|
|13||-- Hoag et al. (Hoag et al. 1961)||28||-- Clariá (Clariá1976)|
|14||-- Clariá (Clariá1974)||29||-- Kilambi (Kilambi 1978)|
|15||-- Pedreros (Pedreros(1984))|
In the case of Haf 19 we find a great difference in (U-B) relative to the values of Munari & Carraro (Munari & Carraro 1996) ( ). On the other hand we find that our (U-B) photometry is comprised between the data of Moffat & FitzGerald (Moffat & FitzGerald 1974) and FitzGerald & Moffat (FitzGerald & Moffat 1974), although the dispersion in these comparisons is quite large ( ). The photometry of NGC 2635 also presents large deviations from the one obtained by Vogt & Moffat (1972). However the comparison was performed with only five stars, and three of them have very close bright neighbours, which could be the explanation of the higher brightness of the photoelectric data of Vogt & Moffat (1972) relative to our PSF magnitudes.
Regarding the comparisons, we find good agreement with the CCD photometry of Munari & Carraro (Munari & Carraro 1996) and Munari et al. (Munari et al. 1998), and somewhat large deviations with respect to the (V-I) colours from the other CCD studies. The largest deviations occur with the photoelectric data of Jorgensen & Westerlund (Jorgensen & Westerlund 1988) where the difference is better described by a significative colour term ( with ; and with ).
Turner (Turner 1989) in an empirical study of the fields of six open clusters determined a mean value of for the reddening slope, although values ranging from at least 0.62 to 0.80 were found from one region to another. One of the regions analysed by Turner (Turner 1989) was the field of NGC 2439, which is also one of the clusters in our sample, and for which he finds a value of E(U-B)/E(B-V) = 0.75.
To investigate the reddening law in the region of our open cluster sample we have searched the WEBDA database for MK spectral types of O, B, A dwarfs observed in this study in order to derive their intrinsic colours. The intrinsic colours were determined by interpolation over the tables given by Schmidt-Kaler (Schmidt-Kaler 1982) which relate MK types to the indexes and were then used to compute the colour excesses.
Figure 3 shows the linear fit to the slope of the colour
excess data. The fit yielded a
with an 0.07 rms dispersion
of the residuals, which despite its low accuracy
is in excellent agreement with the standard slope of 0.72.
The relatively high dispersion in the
residuals may be due to errors in the spectral classifications
(as suggested by the dispersion of the NGC 2453 data) and do
not necessarily reflect to cluster to cluster variations of the
In view of these results we find that the reddening law in the
direction of our sample follows the mean Galactic law and we therefore
adopt the standard value
E(U-B)/E(B-V)=0.72 for subsequent
reddening analysis. For the other colour-colour combinations,
the standard slopes given by Straizys (Straizys 1992)
We do note however that the available data is not sufficient for
a rigorous analysis, so that cluster to cluster variations
up to 0.07 in the reddening slope cannot be discarded.
|Figure 3: Reddening slope defined by the data of the 1998 run. The straight line is a fit of with an 0.07 rms dispersion of the residuals|
|Open with DEXTER|
As in the case of the reddening slope, it is not possible to perform a rigorous study of the ratio of total to selective absorption, RV=AV/E(B-V), where AV is the absorption in the V band. The same kind of proportionality is also defined for the other combinations of bands and colours. Several authors (Sherwood 1975; Crawford & Mandwewala 1976; Turner 1976) have shown that the ratio of total to selective absorption has a typical value of for a diffuse interstellar medium. Denser molecular clouds can give rise to higher values of RV around 4 - 6 (Mathis 1990; Turner 1994).
For cases in which there is evidence of variable extinction across a cluster
field, where all cluster members are assumed to be at a common distance,
a plot of V - MV versus E(B-V) should show a correlation of
slope RV. Evidently, this method commonly known as the
variable extinction method requires the knowledge of
the absolute magnitude, MV, and the colour excess, E(B-V), for
each star. Both MV and E(B-V) can be derived from spectral types.
Turner (1976) applied this method to determine RV for
51 open clusters and obtained an average value of
Three of the open clusters studied by Turner (1976)
were NGC 2323, NGC 2343 and Trumpler 9, which lie in the same Galactic
longitude range of our sample, and for which he obtained the values
2.85, 3.09 and 2.75 respectively.
|Figure 4: The linear least squares fit of the slope yields . If the deviated point of NGC 2453 (marked with a circle) is eliminated the fit yields|
|Open with DEXTER|
We have also investigated the ratio of total to selective absorption using the same photometric and spectroscopic data as in the study of the reddening slope. Because there are not enough stars with known spectral types in each cluster, the variable extinction method cannot be used in a direct form. Instead, a differential approach, similar to the one in Sect. 3.2 for the determination of the extinction coefficients, was followed. If a certain cluster has at least two stars with known spectral types we can write which is distance independent and where RV is the only unknown. Another advantage of this approach is that a greater range in E(B-V) may be achieved than in the traditional one-cluster method. However, the sensitivity to cluster-to-cluster variations is lost and only an average value can be determined.
Figure 4 shows a plot of versus . A value of was found from a linear least squares fit of the slope (i.e. for ) which despite its large uncertainly is consistent with the standard RV=3.1. Furthermore, if the deviating point of NGC 2453 in the lower left of Fig. 4 is eliminated, then the fit yields which is practically identical to the mean value of RV=3.08 given by Turner (1976). Once again, we take the value obtained in this analysis more as an indication that the region under study follows a standard extinction law than that an actual determination of RV, and will use the standard RV=3.1 value in the upcoming analysis of our open cluster photometry.
We have obtained homogeneous photometry in the fields of 30 open clusters between and using data gathered at the same telescope, following the same reduction procedures, and calibrating all the data to the same standard system. These data have resulted in a precise and deep (up to ), photometric catalogue of approximately 65000 stars.
We have compared our data to photometry from other sources finding that in general there is good agreement (at a 0.03 mag level) although some studies present large deviations. The differences between our photometry and that from other studies have been presented in a table which can be used to put all the measurements on the common scale defined by this study.
Since we intend to use this catalogue for cluster reddening, distance and age determinations, we have performed a rough analysis of the reddening slope and of the ratio of total to selective absorption for this region of the Galaxy and found that the typical Galactic values E(U-B)/E(B-V) = 0.72 and are consistent with our data.
The author wishes to thank E. J. Alfaro and A. J. Delgado for providing the 1994 images. The author would also like to thank J.-C. Mermilliod and J. Alves for many useful comments and help. This work was financially supported by FCT (Portugal) through the grant PRAXIS XXI BD/3895/94 and the YALO project. Most of the work was done at the IAA-CSIC (Spain) as part of the author's Ph.D. research. This research made use of the NASA Astrophysics Data System, and of the Simbad database operated at the Centre de Données Stellaires - Strasbourg, France.