A&A 369, 742-749 (2001)
DOI: 10.1051/0004-6361:20000437
S. A. Cellone1 - A. Buzzoni2, 3
1 -
Facultad de Ciencias Astronómicas y Geofísicas, Universidad
Nacional de La Plata, Paseo del Bosque,
1900 La Plata,
Argentina
2 - Telescopio Nazionale Galileo, Roque de los Muchachos Astronomical Obs.,
A.P. 565,
38700 S/Cruz de La Palma (TF), Spain
3 - Osservatorio Astronomico di Brera, Via Bianchi 46, 23807 Merate (Lc),
Italy
Received 15 February 2000 / Accepted 4 December 2000
Abstract
Conflicting evidence has been recently raised in order to use surface
brightness profiles of dwarf galaxies as a distance indicator. In this
paper we discuss in some detail the main error sources in profile
fitting procedures for galaxies with more than one physical component,
showing their impact on the resulting shape parameters.
The apparent tight coupling between shape parameter and (pseudo)
scalelength in the Sérsic law is especially dealt with,
demonstrating that this relationship is mostly a mathematical
artifact, thus throwing doubts on its usefulness as a distance
indicator. Galaxies departing from the luminosity-shape
relation are shown to exhibit different kinds of intrinsic
peculiarities, thus prompting for better securing the empirical
constraints to conform observational samples.
As a relevant example in this sense, new observations of the galaxy
N50 in the NGC 5044 Group are presented. We show that
this object may be at an intermediate evolutionary stage between blue
compact dwarfs (BCDs) and dwarf ellipticals (dEs).
Key words: galaxies: clusters: NGC 5044 Group - galaxies: distances and redshifts, elliptical and lenticular, fundamental parameters, photometry, structure
The shape of the surface brightness profiles of elliptical galaxies,
quantified by the parameter N in the Sérsic (1968) law
The use of this L-N relationship as a distance indicator for dwarf
elliptical galaxies was first investigated by YC94, while the
correlation between the Sérsic parameters N and
was
subsequently used (Young & Currie 1995, YC95) to
derive the distances to 64 dwarf galaxies in the Virgo Cluster. The
quest about the real usefulness of the L-N and
relations
to determine extragalactic distances has received opposite arguments
in favour (Young & Currie 1998) or against (Binggeli & Jerjen 1998,
BJ98).
In a recent paper, Cellone (1999, C99) presented CCD surface photometry for a small sample of dwarf and intermediate luminosity galaxies in the NGC5044 Group, showing that, at least in that case, the L-N correlation partially fails due to a few relatively bright galaxies with "convex'' profiles (i.e. N>1) strongly deviating from the standard relationship. In addition, the important intrinsic scatter in the L vs. N trend seems to drastically limit its practical use for obtaining the distance to the group.
A re-elaboration of a subset of those data, led Young & Currie (2001, YC00) to conclude that the galaxy population in the NGC5044 Group displays on the contrary "a tight scalelength-shape relationship'' revealing therefore to be "an excellent distance indicator''.
As we will show in the present paper, there are reasons to believe
that most of the YC00 conclusions rest in fact on a
misinterpretation of the data (Sect. 2), while their
claimed apparent tightness of the
relation seems likely an
artifact of the parameter mathematical coupling in
Eq. (1). This would therefore throw serious doubts in using
this method to derive extragalactic distances (Sect. 4.1).
We will also turn back here to the relevant case of galaxy N50 in the NGC5044 Group, one of those "outliers'' escaping the standard L-Nrelation in Cellone's (1999) analysis. In spite of its ostensibly normal dE photometric properties, new accurate observations (Sect. 3) indicate indeed that this is a quite peculiar and interesting object sharing most of the characters of dwarf ellipticals (dE) and blue compact dwarf (BCD) galaxies.
Suitable fitting of radial profiles in low surface brightness galaxies is a delicate task which, like an art, requires good skills and the knowledge of the appropriate technique. Every error sources, from seeing and photon noise, as well as any uncertainty in the sky cleaning must be accounted for, and we have to pay attention at the same time also to preserve astrophysical self-consistency of our output.
In a fit of digitized plots of the NGC5044 galaxy sample from Fig. 1
in C99, Young & Currie (2001) obtained, for each galaxy, a new
set of Sérsic parameters. Like C99, they worked in the
surface brightness domain, that is by using Eq. (1) in the
form
One of the questioned objects in the C99 sample is galaxy N42,
for which C99 indicates N=1.43 to be compared with N=0.60in the YC00 fit. At least two evident weak points emerge, in
our opinion, from the YC00 analysis. Contrary to the C99
fitting procedure, that only relied on the S/N>1 portion of galaxy
surface brightness profile (i.e. with
), the new
fit extends much farther from galaxy centre. In the outermost regions,
photon noise and statistical uncertainty in the sky subtraction begin
to dominate causing the output profile to artificially level off at
large radii. By itself, this effect works in inducing nominally
"concave'' (i.e., N<1) surface brightness profiles throughout in
the YC00 fit (see Andredakis et al. 1995 for similar conclusions dealing
with bulge deconvolution in spiral galaxies).
![]() |
Figure 1:
V band surface brightness profiles of galaxies N42
(top) and N29 (bottom). Sérsic law fits are shown as
solid lines. Small ticks show the inner and the two outer (
![]() ![]() |
Open with DEXTER |
In addition, even at first glance (cf. Fig. 1) the N42
surface brightness profile reveals at least two distinct components:
an inner bulge and a main body extending out to
.
This galaxy would therefore need a multi-component scheme (e.g.:
Papaderos et al. 1996) to properly decompose its profile.
Cellone's (1999) model for the N42 main component provided an
integrated magnitude
mag. After subtraction, this
leaves the inner bulge component, with
,
and
extending out to
,
as evident from Fig. 1.
Although this decomposition scheme might probably
be not unique, it shows however that the main morphological component
of the galaxy, providing about 3/4 of the total V luminosity, is in
fact suitably fitted by the original "convex'' profile. Our choice
is also supported by a
test on the fitting residuals confirming
that a simple Sérsic fit can be ruled out at a
confidence level.
As a comparison, Fig. 1 (lower panel) also shows
the profile of the bright dwarf N29. In this case, no change in slope
is evident, and a single Sérsic law (with N=0.54) fits
nicely this profile all along its useful range, as confirmed again
by the
statistics.
The case of N51 (the second galaxy disputed by YC00) is similar
to that of N42, although not so extreme, while for the third object,
the previously uncatalogued galaxy N95A, C99 reported an
exceedingly low surface brightness (
mag arcsec-2)
that definitely prevented any reliable fit. For this reason this
galaxy was not included in any subsequent analysis.
While statistical tests support in our case both the choice of a two-component fit for N42, and a simple Sérsic law for N29, more generally any suitable correction for the bulge contribution in dwarf galaxies may be a non-univocal task. Seeing conditions and other internal bias sources (e.g. ongoing star formation) act in facts in the sense of disturbing galaxy morphology making any fitting procedure somewhat dependent on galaxy apparent size and on environment conditions as well.
In spite of any standard criterion to single out the bulge component, it is clear however that by simply neglecting the problem one would more likely tend to predict too "spiked'' galaxy profiles preferring lower values of N (Andredakis et al. 1995). We will turn back on this point and its impact on the L-N relationship in Sect. 4.
![]() |
Figure 2:
Top:
![]() ![]() |
Open with DEXTER |
The galaxy N50 was originally classified as dEpec, N/BCDring by Ferguson & Sandage (1990). Because of poor spatial resolution, C99 failed to detect any BCD feature in this object eventually appearing as a normal dE, based also on its (B-V)0=0.76 (cf. Caldwell & Bothun 1987). In any case, this would make N50 an interesting object because it is, along with N42, one of the two brightest dEs with convex profile (i.e., N>1) in the C99 sample.
To better assess its evolutionary properties, we collected new
observations of N50 with the EFOSC2 camera at the ESO 3.6 m telescope
in La Silla, Chile, on the nights of April 16 and 17, 1999, as a part
of a study of the low surface brightness galaxy population of the
NGC5044 Group. A detailed description of the observations and image
processing will be given in a forthcoming paper (Buzzoni et al. 2001). Direct
images of this galaxy were obtained under sub-arcsec seeing conditions
in the g, r, i, and z bands of the Gunn system
(Schneider et al. 1983). Data reduction has been accomplished using the
IRAF package achieving a
mag
internal error, while external magnitude uncertainty from standard
zero points amounted to
mag. Surface brightness profiles
have been obtained in the four bands down
mag arcsec-2.
Direct imaging has been complemented also with long-slit spectroscopy between 4300 and 6300 Å at 6 Å FWHM wavelength resolution. Supplementary spectroscopic observations have been also carried out in the range 3500-5400 Å at 8 Å FWHM resolution with a Boller & Chivens spectrograph at the 2.15 m telescope of the CASLEO observatory in San Juan, Argentina on April 9, 1997.
Figure 2 (top panel) shows a
gband contour plot of N50 (the stellar PSF FWHM is
). The
central
(
400 h0-1 pc) region shows several knots
surrounding a central cusp. These features are better seen after
subtraction of a Sérsic model (bottom left panel), and even a
probable dust lane can be appreciated west of the nucleus.
An enlarged map that identifies the central knots is also reported
in the figure (bottom right).
Table 1 reports a full summary of the galaxy photometry,
including aperture magnitudes and detailed measurements of the single
knots. For the latter features, we tried different clean-up
procedures to subtract the smooth galaxy contribution; however, a
plain subtraction of the local "background'' measured around each
source eventually revealed the best choice. The internal photometric
error amounted in this case to mag in each band.
Aperture magnitudes* | ||||
radius [''] | g | g-r | g-i | g-z |
![]() |
17.554 | 0.436 | 0.697 | 0.610 |
![]() |
16.384 | 0.417 | 0.666 | 0.582 |
![]() |
15.628 | 0.416 | 0.669 | 0.582 |
![]() |
15.347 | 0.416 | 0.667 | 0.582 |
![]() |
15.250 | 0.413 | 0.661 | 0.574 |
![]() |
15.213 | 0.412 | 0.654 | 0.575 |
Knot photometry* | ||||
g | g-r | g-i | g-z | |
K-1 | 19.272 | 0.482 | 0.757 | 0.652 |
K-2 | 19.786 | 0.360 | 0.572 | 0.517 |
K-3 | 19.862 | 0.374 | 0.609 | 0.520 |
K-4 | 19.779 | 0.381 | 0.614 | 0.554 |
K-5 | 19.950 | 0.398 | 0.628 | 0.554 |
K-6 | 19.797 | 0.416 | 0.646 | 0.590 |
K-7 | 20.029 | 0.429 | 0.682 | 0.598 |
K-8 | 19.820 | 0.474 | 0.755 | 0.662 |
K-9 | 19.750 | 0.399 | 0.635 | 0.551 |
K-10 | 19.913 | 0.409 | 0.647 | 0.564 |
(*) Internal mag uncertainty is 0.001 for aperture photometry and
0.004 for knot magnitudes. Zero-point external error is 0.03 mag throughout.
![]() |
Figure 3: Top: the g-band surface brightness profile of N50. Error bars account for photometric errors and uncertainties in the sky level. Bottom: the resulting g-i colour profile of the galaxy |
Open with DEXTER |
The g surface brightness profile of N50 is shown in
Fig. 3. Error bars including photon noise
and sky level uncertainty have been taken into account in the plot.
The bottom panel shows the g-i colour profile (after slightly
degrading the i frame to the g PSF to consistently sample surface
luminosity). Note in the figure the blue colour bump about
from the centre due to the knotty ring. A smooth colour gradient is
also evident along galaxy radius with the outermost regions sensibly
bluer than the centre. This colour gradient affects the value of the
parameter N, which increases monotonically from N=1.39 in g to
N=1.63 in z, sampling the range
.
The location of N50 in a g-r vs. g-i colour diagram is shown in Fig. 4. In the main panel of the figure we compared galaxy integrated colours with the locus of Main Sequence stars (based on the Vilnius spectral atlas of Straizys & Sviderskiene 1972), as well as with the theoretical colours for 15 Gyr template galaxies of different morphological type according to the three-zone synthesis models of Buzzoni (1998, 2000). As expected, N50 colours are slightly bluer than high-mass ellipticals, and intermediate between E and Sa Hubble types.
A more detailed match of the population synthesis predictions with the N50 colour profile and with the nuclear knotty features is attempted in the insert panel of Fig. 4. Buzzoni's (1989) simple stellar population (SSP) models, computed for a Salpeter IMF, red horizontal branch morphology, and different metallicity ([Fe/H] = -0.25, 0.0, and +0.30) are reported, tracking evolution from 5 to 15 Gyrs.
![]() |
Figure 4:
( Main panel) - Two-colour diagram comparing N50
integrated g-r and g-i (big open dot) with the stellar Main
Sequence from the Vilnius spectral atlas (star markers) and with
Buzzoni's (1998, 2000) 15 Gyr galaxy models for different
Hubble types, as labelled top left. ( Insert panel) -
N50 aperture photometry (solid dots) and individual colours of the
visible knots in the galaxy central region (open dots with
identification labels according to Table 1). Aperture
photometry is for circular spots at
![]() ![]() ![]() |
Open with DEXTER |
N50 aperture photometry at
,
,
and
is
displayed together with individual photometry of the visible knots
according to the identification number in Table 1 (see
also the reference map in Fig. 2). A substantial
agreement seems to exist between theoretical models and observations
within the zero-point uncertainty in the magnitude scale. An old
(10-15 Gyr) stellar population with slightly sub-solar metallicity
([Fe/H]
)
appears to be the main component in N50 but a
mild [Fe/H] radial gradient might also exist inducing the blueing
colour drift along galaxy radius.
Quite interestingly, nuclear knots reveal a much larger (and
statistically significant) spread in colour. Knot #1 (the nucleus?)
appears indeed even redder than the galaxy core, as do those lying
close to the apparent dust lane (#8) visible one arcsec west of the
nucleus (cf. Fig. 2). Dust reddening might be "patchy''
on the N50 central region, with the west area (corresponding to knots
# 1, 6, 7, near the dust lane # 8) slightly more obscured [
]
than the east side (i.e., about knots # 3 and 4).
Data in Fig. 4 are not corrected for our own Galaxy
reddening (which however should not exceed
according to Burstein & Heiles 1982). In addition, one should also consider a
little blue shift of all the galaxy data by
and
to take into account for k-correction.
Even correcting for these effects, it seems likely however that the
whole stellar population in N50 should consist of stars older than 5
Gyr, and only a much enhanced (super-solar) metallicity should be
invoked to predict a younger age.
The spread in age among the galaxy stellar population is even more
evident from the analysis of the integrated spectrum, shown in
Fig. 5. Galaxy spectral energy distribution has been
obtained by matching the CASLEO data in the range
Å with ESO observations (
Å),
adding then the monochromatic fluxes from the integrated g, r,
i, and z magnitudes. Gunn photometry also set the absolute flux
calibration reproducing galaxy energy distribution within the central
aperture.
![]() |
Figure 5:
Composite spectral energy distribution of galaxy N50. The
CASLEO spectrum, in the range
![]() ![]() ![]() |
Open with DEXTER |
Two SSP models from the Buzzoni (1989) data set, with [Fe/H] = -0.25 and age 5 and 15 Gyrs, are superposed on the plot. The older model fits well along the optical and red wavelength while it lacks UV luminosity below 4500 Å. On the contrary, the 5 Gyr stellar population provides a good fit to the ultraviolet but it would predict too blue Gunn colours. Accounting for SSP luminosity evolution, in a simple interpretative scheme assuming a mix of these two main stellar components, this allows us to estimate that the old (15 Gyr) population comprises about 3/4 of the total mass of the galaxy.
The redshift of N50, as derived from our spectra, amounts to
km s-1 (that is
). This
yields a distance to the galaxy of
23.9 h0-1 Mpc and a distance
modulus
.
From the model fit of
Fig. 5, a bolometric correction to g of
can be obtained, leading for N50 to
.
This is
or
.
Estimating a theoretical
from the relevant SSP models of
Buzzoni (1989) (once accounting for the whole stellar mass by
integrating the Salpeter IMF) then one obtains
as a fair estimate of the
total (stellar) mass of N50.
In its overall morphology, N50 is very reminiscent of the dwarf
(
MB=-16.71 mag) galaxy Markarian 996, which Thuan et
al. (1996) report to have smooth elliptical isophotes, with
several bright knots and dust patches in its central (400 pc)
region. However, contrary to what we observe in N50, Mrk 996 clearly
shows signs of active ongoing star formation in its centre, thus
fitting with a nE BCD (nuclear-elliptical blue-compact-dwarf)
classification.
The age we derive for the blue knots in N50 (see Fig. 6)
suggests that it may have looked very similar to Mrk 996 a few Gyrs ago.
Although now observed in its more quiescent evolutionary stage,
the H
and 3727 Å [OII] emission lines, prominent in our spectrum
of the galaxy, still witness the presence of a wealth residual gas.
In this sense, N50 probably represents an ideal link in the dE-BCD
connection (Thuan 1985; Evans et al. 1990; Meurer et al. 1992).
![]() |
Figure 6: Expected B-V and g-r star-burst colour evolution according to Buzzoni (1989, 1998). Three different values for [Fe/H] = -0.3, 0.0 (solid lines), and +0.3 are displayed, as labelled on the B-Vplot. The indicative age distribution of the N50 central knots is also reported in the g-r plot (assuming [Fe/H] = -0.25). Note that, in general, even any massive starburst episode older than 2 Gyr, superposed to a galaxy old stellar component would not be able to turn integrated colours bluer than (B-V) = 0.7 still allowing galaxy to be recognized as a quiescent "standard'' elliptical according to the YC00 colour selection criterium |
Open with DEXTER |
Quite importantly, it is also worth noting from Fig. 6 that even any massive intervening starburst activity older than 2 Gyr, superposed to the galaxy "quiescent'' old stellar component, would still maintain integrated colours redder than (B-V) = 0.7. According to the YC00 colour selection criterium, active dwarf ellipticals would therefore be recognized as "standard'' galaxies in the L-Nrelationship.
Despite such a negligible effect on the integrated colours, starburst episodes could however much more strongly affect galaxy surface brightness profiles. In the case of N50, for example, the presence of the bright knotty ring around the centre certainly modulates the Sérsic shape parameter leading to a higher fitting value for N.
Each of the three "outliers'' that depart from the L-N relation in the NGC5044 sample studied by C99 (namely N42, N49, and N50), displays a different kind of peculiarity. As shown in Sect. 2, N42 has a normal colour (B-V=0.75), but its morphology reveals the presence of a central bulge that should be accounted for in a multi-component fitting model.
On the other hand, N49 is a very blue (B-V=0.49) irregular galaxy (C99), and cannot therefore be included in our analysis of the dwarf elliptical population.
N50 takes only apparently the look of a standard dE object. In the previous section we showed that in spite of its quite normal B-V, this galaxy shows the signature of relatively fresh star formation in its centre.
Given a so wide range of morphological features it is difficult to firmly assess their systematic influence on the dE L-N relationship. We should certainly agree with YC00 that any simple colour argument such as that relying on the integrated B-V, is not sufficient alone to secure a fair sample selection.
Attempting a very summary analysis in this sense, one could expect bulge-enhanced systems to affect the L-N relation in a systematic way, especially at the bright tail (MV < -15) of dE luminosity function, by forcing a lower value of N (see Fig. 3 in C99). While in some cases this could even (artificially) improve the match with the standard L-Nrelationship, it does not help much for calibrating galaxy distance since in any case the Sérsic shape parameter poorly tracks galaxy luminosity at brighter magnitudes.
On the other side, intervening star formation, especially in case of clumpy features like those in N50, could dramatically affect galaxy surface brightness enhancing the spread in the profile fitting procedure. Again, this problem would more severly affect brighter dEs for which both "convex'' and "concave'' Sérsic profiles could result.
A fair estimate of the frequency of peculiar objects like N42 or N50 is in this regard the real concern to ultimately assess the reliability of the L-N relation as a distance indicator. M32-like galaxies seem rather rare objects (Ferguson & Binggeli 1994; Ziegler & Bender 1998) but, as a matter of fact, even a rather coarse sample like that of C99 resulted affected by over 20% of such "deviating'' objects.
In any case, it is clear from our results that any useful application of the L-N relation as a distance indicator should forcedly be pursued on a statistical basis mainly relying on non-nucleated (faint) dEs. Obviously, such a tight sampling constraint might be the most stricking drawback of this approach for extragalactic studies.
It is well known that reliable extragalactic distance indicators are
either based on the luminosity of a given standard candle (e.g.,
SNIa, globular clusters luminosity function, etc.), or on the
relation between two independently measured parameters of galaxies,
one depending on distance and the other distance-independent (e.g.,
the Tully-Fisher and
relations) (see for example
Jacoby et al. 1992 or Trimble 1997 and references therein). For the L-Nrelation to fulfill the condition of independence between both
parameters, the total apparent magnitude should be obtained
independently from N by means of aperture or growth-curve
photometry, instead of calculating it from the integration of
Eq. (1), i.e.,
The situation with the
relation is substantially different,
since it involves two free parameters obtained from the same fitting
formula. Worse, it is evident from Eq. (1) that the coupling
between
and N must be strong, i.e., any error in the
measurement of N will propagate (non linearly) to
.
This is
important because there is a sizable scatter among values of
N measured for the same galaxies by different researchers
(Ryden et al. 1999).
As an illustrative example, we have re-fitted from the C99 data
the two high-S/N profiles of dwarfs N29 (N = 0.54) and N83 (N =
0.93) after varying the adopted sky level by a
of
(this is a relative fluctuation
). In addition, based on the C99 observational setup, we
also generated an artificial profile for a faint N > 1 galaxy and
explored sky-fluctuation uncertainty likewise.
The results are displayed in Fig. 7 where solid bars
represent the ranges spanned by the fitted Sérsic parameters. It is
evident that variations in the adopted sky level cause the galaxies to
move along the
relation (the dotted line is the
polynomial fit for Fornax Cluster dwarfs given by YC00),
showing that observational errors at least partially contribute to the
relation. This feature is to some extent a straightforward
geometrical consequence of the fit: a more concave profile will
generally predict a more "spiked'' nucleus, that is a sharper pseudo
scalelength.
![]() |
Figure 7:
Shape parameter (N) vs. pseudo scalelength (![]() ![]() ![]() |
Open with DEXTER |
Also shown in Fig. 7 (dashed lines) are the loci for
galaxies with constant luminosity and
,
according to
BJ98. The three dashed lines correspond to
,
16, and 18
mag, respectively, from upper-left to lower-right. Even before
Sérsic law was first used to quantify the shape of dE profiles, it
was qualitatively known that brighter dwarfs tend to be larger (and,
at the same time, of higher surface brightness, and with more
"concave'' profiles; e.g., Ferguson & Binggeli 1994, and references
therein). Figure 7 shows, instead, that brighter galaxies
following the
relation tend to have smaller
(and N) values (this can also be seen from the data in YC95;
see also Jerjen & Binggeli 1997). This is actually why the parameter
in
the Sérsic law (Eq. (1)) can no longer be taken as a
physical scalelength. Our suggestion is therefore that special
caution should be deserved in using the
relationship to
infer galaxies distances.