A&A 430, 751-759 (2005)
G. Paturel1 - I. Vauglin1 - C. Petit1 - J. Borsenberger 2 - N. Epchtein3 - P. Fouqué 4 - G. Mamon5
1 - CRAL - Observatoire de Lyon, 69561 Saint-Genis Laval Cedex, France
2 - Observatoire de Paris-Meudon, 5, place Jules-Janssen, 92195 Meudon, France
3 - LUAN, Université de Nice Sophia-Antipolis, CNRS - UMR 6525, Parc Valrose, Nice, France
4 - Observatoire Midi-Pyrénées, 14 avenue Edouard-Belin, 31400 Toulouse, France
5 - Institut d'Astrophysique de Paris, 98B Bd. Arago, 75014 Paris, France
Received 26 April 2004 / Accepted 17 September 2004
We present a catalog of near-infrared properties of LEDA galaxies, using the full resolution images from the DENIS survey. The fluxes are integrated in eight homothetic ellipses defined by their proper axis ratio, position angle and major axis (up to twice the blue diameter at the isophote 25 mag arcsec-2) extracted from the LEDA database. From the curves of growth in I, J and photometric bands, we estimated different apparent magnitudes and diameters ("total'', "Kron'' and "isophotal''). Isophotal parameters refer to the limiting surface brightnesses: 22.5 I mag arcsec-2, 21.0 J mag arcsec-2 and 20.0 mag arcsec-2 for the three photometric bands, respectively. The result is a catalog of objects (among which there are galaxies, probable galaxies and galaxies to be confirmed). The catalog gives about (the figures vary, depending on the considered magnitude or diameter): I-band magnitudes, J-band magnitudes, -band magnitudes and I-band diameters, J-band diameters, -band diameters.
The typical standard deviations for I, J and magnitudes are 0.14, 0.15 and 0.25, respectively, for magnitudes limited at I=16, J=15 and . The contamination by superimposed objects probably remains the major source of problems and could require future improvement. The completeness limits in magnitude are about: 15.5, 14.5 and 13 in I, J and , respectively.
Key words: galaxies: general - catalogs
Today many such large surveys have been undertaken in different wavelength bands (2MASS, SLOAN, HIPASS, FIRST). The 2MASS project was conducted on the whole sky with similar purposes: a very large sky infrared survey. It uses J, H and bands (Jarrett et al. 2000) with two dedicated telescopes. DENIS was conducted in a more difficult way, using a telescope rented to ESO. For the extragalactic domain we produced preliminary catalogs using low resolution images (Vauglin et al. 1999; Paturel et al. 2003b; Rousseau et al. 2000; Vauglin et al. 2002).
In this paper we present an exploitation of the DENIS survey using full resolution images. This leads to a catalog of galaxies with I, J, magnitudes. This is still a provisional catalog because the final DENIS catalog (Mamon et al. 2004, in preparation) will not be restricted to LEDA objects. It will be deeper, more accurate and more homogeneous. Nevertheless, the present catalog should be useful for preliminary studies, especially when I-band data is required. The present catalog is about ten times larger than the one we produced, for I-band only, with low resolution images (Paturel et al. 2003b).
Many primary or secondary photometric standards were measured at least twice every 180 images
(strip). From this analysis one of us (JB) derived the I, J and
zero points needed to transform
intensities to absolute fluxes.
The adopted zero points are given for each night (the strips from the same night have thus
the same zero-point). The variations along the
survey are shown in Fig. 1. Some abrupt changes correspond to cleaning of mirrors or
modifications of the camera. When it was not possible to measure zero-points, we adopt
the mean between the previous and next values, instead of using a default value.
In such cases a flag is put on the corresponding data.
This flag is designated as
The statistics of flagged objects is given in Table 1.
|Figure 1: Variation of adopted zero-points with strip number (i.e. with time) for I, J and photometric bands.|
|Open with DEXTER|
Table 1: Statistics of the different flags, for I, J and photometric bands.
During the DENIS survey there was no automatic monitoring to check the quality of the sky.
So, we use the stability of the sky background to test the reliability
as showed in a previous paper (Vauglin et al. 2002).
The sky background may vary along a strip, e.g. when the strip crosses the Milky Way.
However, the standard deviation of the sky background calculated for
an elementary image should be more stable along a strip if the sky is clear. On the contrary,
if clouds cross the field, the pixel intensities will vary at random and the standard deviation
of the sky background will increase. Of course, one cannot exclude that
unresolved stars also affect the standard deviation of the sky background in crowed fields,
but this will also produce poorer data.
Thus, the fluctuations of the standard deviation of the sky background along
a strip is used to estimate the quality q of the strip. The definition is thus:
|Figure 2: Variation of quality estimated from the fluctuations of the sky background along a strip as a function of strip number for the I photometric band.|
|Open with DEXTER|
Further, because the coordinates in LEDA are relatively good (the standard deviation is typically a few arcseconds according to Paturel et al. 1999) this solves the remaining problem of DENIS coordinates. We will keep only objects for which a good agreement exists between both coordinate determinations.
From LEDA we also extracted the diameter D25, axis ratio R25 and position angle for all selected galaxies. The dimensions refer, on average, to the blue limiting surface brightness 25 mag arcsec-2. Diameters and position angle are used to integrate the flux in eight homothetic ellipses having diameters regularly distributed between D25/4 and 2D25 (i.e., 2D25/i, i down from eight to one).
During this integration process we searched for the maximum of the pixel intensity within the ellipse of diameter 2D25/5. This defines the adopted center of the galaxy. The central flux and the surface area of the corresponding ellipse is used to estimate the object type (star or galaxy) when it is not confirmed from LEDA (see Sect. 4). We calculate the DENIS coordinates for the pixel of maximum central intensity. When the discrepancy with the LEDA coordinates exceeds twice the minor axis or is greater than , a flag is put on the measurement. Objects having a problem of coordinates are rare (a few percent). They are simply removed.
In order to remove superimposed stars we use the following procedure:
when a pixel intensity is higher than the maximum found
at the center of the galaxy, we replace the surface area corresponding to the Point Spread
Function (PSF) around this pixel by a constant intensity. This constant intensity is chosen
in the vicinity of the considered pixel (typically at a distance corresponding to the PSF radius).
When the total correction exceeds 0.3 mag a flag is put on the measurement.
This flag will be designated as f*.
The statistics of flagged objects are given in Table 1.
As expected, the number of objects that have this flag increases towards low galactic latitude
Note that the presence of a flag does not mean that the magnitude should be rejected. It simply means
that a correction for superimposed stars has been applied.
|Figure 3: Fraction of objects having a flag f* as a function of absolute value of the galactic latitude. The dashed line represents the mean fraction.|
|Open with DEXTER|
The eight fluxes within the eight homothetic ellipses are transformed into magnitudes using
proper zero points (previous section) and constant atmospheric extinction coefficients:
0.02 in I, 0.08 in J and 0.11 in .
Because the sky background
has been reduced to zero, on average, the magnitudes do not need, in principle, a sky background subtraction. In practice, the local sky background is estimated from external ellipses.
We give an example of a curve of growth (magnitude vs. surface area) for a typical galaxy arbitrarily
chosen (Fig. 4).
In the next section, we use the surface brightness profile for each galaxy.
|Figure 4: The curve of growth of a typical galaxy (NGC 7808). The points (filled circles) and dashed line represent the raw measurements obtained by integrating within the eight homothetic ellipses defined in the text. The open circles and solid line represent the measurements corrected for the local sky-background calculated from the two largest ellipses. The error bars are obtained from the standard deviation of the curve of growth.|
|Open with DEXTER|
The final test of quality is made by flagging galaxies that are near the edge of the frame. If the marginal projections of the largest ellipse have more than five percent length outside the limit of the frame, a flag is put on the measurements for this object. Table 1 summarizes the statistics of all flags.
After this first step we obtain three catalogs. They contain: measurements for objects in I, measurements for objects in J and measurements for objects in . Many objects are measured several times because of the overlap of the images along a strip and the overlap of adjacent strips. Thus, we have to calculate the mean curve of growth for each object.
The galaxy profile is calculated as follows:
fi = 10-0.4 mi, i=1 to 8) be the fluxes within the eight ellipses of surface
The surface brightness Bi at an equivalent radius
|Figure 5: Fit of a Sersic function to the brightness profile of a typical galaxy (NGC 7808). The error bars result from the standard deviation of the fit.|
|Open with DEXTER|
Using the standard deviation (rms residuals) of the fit, we calculated the weight of each individual magnitude as the inverse of the square of the standard deviation ( ). Then, we calculated the weighted mean magnitudes within the eight homothetic ellipses and their actual error (Paturel et al. 1997). This actual error takes into account both the internal uncertainty given by the standard deviation of the fit and the external uncertainty resulting from the comparison of independent measurements. An object is rejected if, at a given level, the magnitude cannot be calculated. Further, when the standard deviation of the fit is larger than a given limit, all measurements are also rejected. The adopted limits are: 0.6, 0.8 and 1.0 mag for I, J and , respectively.
We calculated the mean of each flag. The integer division gives the
final flags, each being 0 or 1.
We also calculated the confidence in the recognition of an
object as a galaxy.
Many objects are confirmed galaxies. Hence, this parameter is useful only
for objects not yet confirmed. This parameter is defined in magnitude as:
|Figure 6: Parameter G for two populations: confirmed galaxies (full line) or unconfirmed objects near the galactic plane (dashed line). Below the limits drawn with a vertical doted line an unconfirmed object should not be considered as a galaxy without verification.|
|Open with DEXTER|
We plot (Fig. 7) the fraction of not confirmed galaxies versus the default magnitude in I-band. The fraction is almost constant up to the completeness limit ( mag) found in Sect. 8, but rises quickly above this limit. This shows that much work is still needed to confirm the extragalactic nature of many objects in LEDA. Nevertheless, the DENIS magnitudes will be available from this paper, when these objects will be confirmed.
We calculated four kinds of magnitude for each band:
|Figure 7: Fraction (in percentage of the total number) of not confirmed galaxies versus the default magnitude, . The dashed line shows the completeness limit (dashed line) found in Sect. 8. Above this limit the fraction of not confirmed galaxies raises rapidly.|
|Open with DEXTER|
We calculated also the colours I-J,
to reveal discrepant magnitudes. Indeed, when a magnitude is involved in two discrepant
colors (differing from the mean at the 3 level) it is removed because it results
from an uncertain measurement. The mean values are the following:
|Figure 8: Colours I-J, and , calculated from total magnitudes.|
|Open with DEXTER|
In order to have an external estimate of the standard deviation of our DENIS magnitudes, we make a comparison of our I magnitudes with the I-band ones of the survey by Mathewson et al. (1992). Our J and magnitudes are compared with the J and magnitudes of the 2MASS survey (Jarrett et al. 2002). For 2MASS we use the "total'' magnitudes (Cols. 41, 44 and 47 in the 2MASS extended source catalog).
We first made a cross-identification with these surveys. Each galaxy of our catalog is searched in other catalogs within a radius of . The cross-identification is accepted when there is only one candidate within r and when the agreement between coordinates is better than .
A cut at a limiting magnitude (16, 15 and 14 for I, J and , respectively) avoids the classical bias that distorts the relationship (Paturel et al. 1991) when the catalogs involved in the comparison have different depths. The effect of the choice of these limits is studied below (Table 4). Further, in order to obtain the nominal standard deviation, we first made the comparison for confirmed galaxies, with no flags and with an internal actual error less than 0.2 mag. This last limitation practically does not affect the results. Indeed, we repeated the calculation with an internal error as large as 0.9 mag without having a significant change in the final standard deviation (see Table 5). The case of less accurate objects (those with a flag) is studied in next section.
Table 2: Comparison of zero-points of different kinds of apparent magnitude ("default'', "Kron'' and "isophotal'') with the "total'' magnitude.
The different results of the two-by-two comparisons are summarized in Table 3. Examples of the comparison are given in Fig. 9 for the default magnitudes: , and .
For the comparison with Mathewson's sample, we considered the direct regression (because Mathewson's magnitudes are assumed to be much more accurate than our magnitudes), while for the comparison with 2MASS we used the mean regression, because we assume that both systems have nearly the same accuracy (this is a first estimate, probably leading to an underestimation of our standard deviations). The slopes for I-band magnitudes are not significantly different from one. For J and -bands, the small departure from one, always leads to a shift in magnitude smaller than half the standard deviation, over a range of 10 mag. Thus, we will provisionaly consider only the zero-point differences.
The zero-point shift (Table 3) is reasonably small (less than a few percents) when considering "default'' magnitudes. Our total magnitudes are brighter than those from other surveys, but the zero-point shifts never exceed 0.1 mag. In another paper we will make more detailed comparisons, in order to convert magnitudes to homogeneous systems.
To estimate the standard errors of our magnitudes we assumed that all the error in I-band comes from our own I-magnitudes and that the errors in J and are the same than those of 2MASS J and magnitudes. In other words, the standard deviation of the I-band regression gives directly the standard error on our I-magnitudes while the standard error on J- and -magnitudes is obtained by dividing the standard deviation of the regression by (same error on both axes). The results for different limiting magnitudes are summarized in Table 4. Let us note again that the standard errors on J and magnitudes of 2MASS are probably smaller than ours. Thus our standard errors may be underestimated.
Table 3: Comparison of zero-points of our DENIS magnitudes, with those of other samples ("Mathewson'' and "2MASS'').
Table 4: Standard error on our magnitudes as a function of the limiting magnitudes.
A comparison of our I magnitudes with J ones from 2MASS gives a standard deviation . Assuming again that the error on both axes is the same, one deduces that the uncertainty on our I-band magnitude is about , i.e., . This is in good agreement with the value estimated from the comparison with Mathewson's sample around the same limit (0.14 at mag). These results are obtained after one cycle of rejection. Without any rejection, they give larger standard errors. For instance, for limiting magnitudes 16, 15 and 14 in I, J and respectively, we obtain 0.20 (instead of 0.14), 0.18 (instead of 0.14) and 0.28 (instead of 0.25).
We compared our I-band diameters at the isophote 22.5 I-mag arcsec-2 with those measured by Mathewson at the isophote 23.5 I-mag arcsec-2. The comparison is shown in Fig. 10. The shift between both diameter systems is 0.066 with a standard deviation . This shows that, on average, the accuracy of these diameters is as good as, e.g. Nilson's diameters ( according to Paturel et al. 1991).
We estimate the quantity for the I-band around the brightness mag arcsec-2. For comparison, in B-band, around 25 mag arcsec-2, this quantity is 0.09, for a typical spiral galaxy (Fouqué & Paturel 1985).
The standard deviations, calculated as previously, are given in Table 5. The calculation is made for the limiting magnitudes 16, 15 and 14 for I, J and , respectively. The most severe influence comes from f* (star contamination).
Table 5: Standard error on total magnitudes as a function of the flag. The first two lines give the result when there is no flag but for actual error limited to either 0.2 mag or 0.9 mag.
Table 6: Description of the structure of the catalog for each galaxy. In the electronic form available at CDS, only the numerical data are given (the designation of columns are given here only for clearness). Further, the four lines are all written on a single line to make the reading easier. The catalog is sorted in Right Ascension for the Equinox 2000.
|Figure 9: Comparison of DENIS I-, J-, -bands "default'' magnitudes with respectively the I-band magnitudes from Mathewson's sample and the J and magnitudes from the 2MASS survey.|
|Open with DEXTER|
|Figure 10: Comparison of DENIS I-band diameter at the isophote 22.5 I-mag arcsec-2 with those measured by Mathewson at the isophote 23.5 I-mag arcsec-2.|
|Open with DEXTER|
The typical standard deviations on total magnitudes are 0.14, 0.15 and 0.25, for I-, J- and -band and limits in magnitude of 16, 15 and 14, respectively.
In order to estimate the mean completeness limit for each band, we plotted
(Fig. 11), for the three bands.
The slopes differ significantly from the canonical value 0.6, expected
for a uniform distribution of galaxies. The slopes are: 0.52, 0.54 and 0.58,
for I-, J- and -band respectively. This variation seems to be correlated
with the uncertainty: the higher the uncertainty on magnitude, the closer to
the canonical value 0.6. This could be an illustration of the
Eddington effect recently analyzed by Teerikorpi (2004).
|Figure 11: Completeness curves for I, J and magnitudes. Black circles and solid line represent the completeness curves for the present DENIS catalog. Open circles and dashed line represent the completeness curves for the 2MASS extended source catalog.|
|Open with DEXTER|
From Fig. 11 one sees that the completeness limits are about : 15.5, 14.5 and 13.5 in I, J and , respectively. For comparison we give the same plot for J and of the 2MASS extended source catalog (open circles), limited to the southern hemisphere. 2MASS is deeper in (14 mag instead of 13.5). In J, the deepness is almost the same for both survey. The 2MASS catalog has more objects than the DENIS one but it presents an excess of bright objects, while DENIS presents a lack of bright objects (in J). On the other hand, due to the selection of our objects from LEDA (essentially built from blue plates), our catalog is probably biased towards blue objects.
In conclusion, we want to warn any potential user that the contamination by superimposed objects (stars or companion galaxies) remains the major cause of problems. In using of individual measurements it is advisable to check that there is no contamination within . This could be a way to improve the database in the future.
We thank all astronomers and technicians who participated to the measurements in La Silla. A part of the calculation (I-band) was made on a computer managed by Ph. Prugniel; we thank him for that. We thank the referee, T. H. Jarrett, for very good suggestions that helped us to improve the paper.