The planetary nebula population of M33 and its metallicity gradient: A look into the galaxy's distant past

Abstract

The Planetary Nebula (PN) population of M33 is studied via multi-fiber
spectroscopy with Hectospec at the MMT. In this paper we present the spectra of
102 PNe, whereas plasma diagnostic and chemical abundances were performed on
the 93 PNe where the necessary diagnostic lines were measured. About 20% of the
PNe are compatible with being Type I; the rest of the sample is the progeny of
an old disk stellar population, with main sequence masses M<3M${_\odot}$ and
ages t$>$0.3 Gyr.
By studying the elemental abundances of the PNe in the M33 disk we were able
to infer that: (1) there is a tight correlation between O/H and Ne/H, broadly
excluding the evolution of oxygen; (2) the average abundances of the
$\alpha$-elements are consistent with those of \hii regions, indicating a
negligible global enrichment in the disk of M33 from the epoch of the formation
of the PN progenitors to the present time; (3) the radial oxygen gradient
across the M33 disk has a slope of -0.031$\pm$0.013 dex kpc$^{-1}$, in
agreement, within the errors, with the corresponding gradient derived from HII
regions. Our observations do not seem to imply that the metallicity gradient
across the M33 disk has flattened considerably with time. We report also the
discovery of a PN with Wolf-Rayet features, PN039, belonging the class of late
[WC] stars

Full text

Page 1

arXiv:0901.2273v1 [astro-ph.SR] 15 Jan 2009
The planetary nebula population of M33 and its metallicity
gradient: A look into the galaxy’s distant past1
Laura Magrini1,2
Istituto Nazionale di Astrofisica, Osservatorio Astrofisico di Arcetri
Istituto Nazionale di Ottica Applicata, Firenze, I 50125, Italy
laura@arcetri.astro.it
Letizia Stanghellini3
National Optical Astronomy Observatories, Tucson, AZ 85719
lstanghellini@noao.edu
and
Eva Villaver4,5
Space Telescope Science Institute, Baltimore, MD 21218
Affiliated with the Hubble Space Telescope Division of the European Space Agency
villaver@stsci.edu
ABSTRACT
The Planetary Nebula (PN) population of M33 is studied via multi-fiber spectroscopy with
Hectospec at the MMT. In this paper we present the spectra of 102 PNe, whereas plasma diag-
nostic and chemical abundances were performed on the 93 PNe where the necessary diagnostic
lines were measured.
About 20% of the PNe are compatible with being Type I; the rest of the sample is the progeny
of an old disk stellar population, with main sequence masses M<3M⊙and ages t>0.3 Gyr.
By studying the elemental abundances of the PNe in the M33 disk we were able to infer
that: (1) there is a tight correlation between O/H and Ne/H, broadly excluding the evolution of
oxygen; (2) the average abundances of the α-elements are consistent with those of H ii regions,
indicating a negligible global enrichment in the disk of M33 from the epoch of the formation
of the PN progenitors to the present time; (3) the radial oxygen gradient across the M33 disk
has a slope of -0.031±0.013 dex kpc−1, in agreement, within the errors, with the corresponding
gradient derived from H ii regions. Our observations do not seem to imply that the metallicity
gradient across the M33 disk has flattened considerably with time. We report also the discovery
of a PN with Wolf-Rayet features, PN039, belonging the class of late [WC] stars
Subject headings: planetary nebulae: general, abundances, individual (M33 system) — galaxies: individ-
ual(M33), evolution, abundances
1Observations reported here were obtained at the MMT
Observatory, a joint facility of the Smithsonian Institution
and the University of Arizona.
1. Introduction
The galaxy M33 (NGC 598) is one of the closest
spiral galaxies of the Local Group. The measured
distance to M33 ranges from 730 kpc (Christian
1

Page 2

& Schommer 1987) to 910 kpc (Kim et al. 2002),
with the most recent estimates of Sarajedini et
al. (2006) and Bonanos et al. (2006) in the mid-
range. The proximity of M33, together with its
large angular size (optical size 53’×83’, Holmberg
1958), and its intermediate inclination (i=53◦), al-
lows detailed studies of its stellar populations and
ionized nebulae.
M33 is a galaxy rich in both PNe and H ii re-
gions. The population of PNe in M33 was early
investigated using an objective-prism survey by
Ford (1983) and Lequeux et al. (1987). The advent
of wide-field CCD cameras allowed deeper surveys.
Magrini et al. (2000) identified 131 candidate PNe.
More recently, Ciardullo et al. (2004) confirmed
a large number of previously detected PNe and
identified new candidates, leading to the current
number of spectroscopically confirmed PNe: 138
in the M33 disk and 2 in its halo.
Due the closeness of M33 and the brightness
of the giant H ii regions, their spectroscopy was
obtained since the 70s. Searle (1971) presented
spectrophotometry of eight H ii regions, and fur-
ther spectroscopic studies were carried on by
Smith (1975), Kwitter & Aller (1981), and by
V´ ılchez et al. (1988). Several catalogs of the M33
H ii regions have been published, such as those
by Courtes et al. (1987), Calzetti et al. (1995),
Wyder et al. (1997), and Hodge et al. (1999). Re-
cently, the Local Group Census (LGC, Corradi
& Magrini 2006) and the Local Group Survey
(LGS, Massey et al. 2007) have provided deep
narrow- and broad-band ∼2 square degree cover-
age of M33, disclosing a conspicuous number of
new emission-line objects located at all galacto-
centric distances.
Given the wealth of information on its PN and
H ii region populations, M33 is indeed an ideal
candidate to test chemical evolution predictions.
The aim of the spectroscopic studies of PNe and
H ii regions is devoted both to the individual
studies of these objects, and to draw the radial
metallicity gradient across the disk. In particu-
lar, PNe and H ii regions have two very differ-
ent formation ages, and the comparison of these
two populations provides insight on the evolution
of the host galaxy. PN progenitors are low- and
intermediate-mass stars (LIMS), with masses be-
tween 1 and 8 M⊙, which must have been formed
between ∼3×107yr and 10 Gyr ago (Maraston
2005). Instead H ii regions are very young.
During their evolution, LIMS do not modify, at
least at zeroth approximation, the composition of
α- elements such as oxygen, neon, argon, and sul-
fur. These elements are produced mainly from the
nucleosynthesis of Type II supernovae and main-
tain their capability to testify their original pres-
ence in the interstellar cloud that gave birth to
the PN progenitor. While there are indications
that for extremely low metallicities both oxygen
and neon could be modified, as it has been ob-
served both in the SMC and in other dwarf galax-
ies (Leisy and Dennefeld 2006; Magrini et al. 2005;
Kniazev et al. 2008), at the metallicity of M33
one does not expect any nucleosynthesis activity
involving these elements (Marigo 2001).
On the other hand, the helium, nitrogen, and
carbon abundances measured in PNe do not cor-
respond to those at the time of the progenitor’s
formation, since these elements are synthesized in
LIMS. These elements hence give information on
LIMS evolution accordingly to their initial mass
and metallicity, and, at given metallicity, are help-
ful to constrain the PN progenitor mass and age.
The M33 metallicity gradient derived from α-
element abundances has been the focus of sev-
eral studies already, both involving PNe and H ii
regions. The advantage of M33 with respect to
our Galaxy, as a playground for gradients, is that
PNe in M33 have well determined galactocentric
distances (with relative errors within 5%) com-
pared to the large indetermination of Galactic PN
distances (Stanghellini et al. 2008). Magrini et
al. (2004) using PN spectroscopy in M33 deter-
mined the elemental abundances of 11 PNe, and
Magrini et al. (2007b) (hereafter M07b) derived
an oxygen radial gradient of ∆(O/H) / ∆R = -
0.11±0.04 dex kpc−1, where R is the galactocen-
tric distance. The sample of 11 PNe was too small
for a definite answer to the PN gradient of M33,
especially given the paucity of PNe observed at
large galactocentric distances.
The metallicity gradient of M33 using H ii
regions has been obtained by many authors,
with broadly different results.
ies by Smith (1975), Kwitter & Aller (1981),
and V´ ılchez et al. (1988) agreed on a steep oxy-
gen gradient. Garnett et al. (1997) also ob-
tained a steep gradient, with ∆(O/H) / ∆R =
-0.11±0.02 dex kpc−1by homogeneously com-
The first stud-
2

Page 3

piling published data.
such as those by Crockett et al. (2006), Ma-
grini et al. (2007a), and Rosolowsky et al. (2008)
(hereafter M07a and RS08, respectively) seem to
converge to a much shallower gradient, respec-
tively deriving ∆(O/H) / ∆R = -0.012±0.011,
-0.054±0.011, and -0.027±0.012 dex kpc−1. The
most recent result by Rubin et al. (2008), based
on Ne/H and S/H, gives -0.058±0.014 and -
0.052±0.021 dex kpc−1, respectively. The RS08
sample is the largest to date, and should be used
preferentially, since the results from small samples
might emphasize the scatter rather than the slope.
Metallicity gradients in M33 have also been
estimated form young giant stars (Herrero et
al. 1994, McCarthy et al. 1995, Venn et al. 1998,
Monteverde et al. 1997, 2000, Urbaneja et al. 2005),
and from AGB (Cioni et al. 2008), RGB stars
(Stephens et al. 2002, Kim et al 2002, Galletti et
al. 2004, Tiede et al. 2004, Brooks et al. 2004,
Barker et al. 2006), and Cepheids (Beaulieu et
al. 2006).
One of the most discussed questions about the
metallicity gradient in disk galaxies is how it
evolves with time. Chemical evolution models (see
e.g. M07b for a review of M33 models) predict dif-
ferent temporal behaviors of the metallicity gra-
dient depending on assumptions such as gas in-
flow and outflow rate, and star and cloud forma-
tion efficiencies. Observations are needed to con-
strain these theoretical scenarios, but so far they
have been insufficient, especially for the old popu-
lations. Comparing different sets of results for the
young stellar populations, such as H ii regions, and
the old population, such as AGB and RGB stars,
is also delicate, since the techniques of observing
and analyzing nebulae and stars are very different,
each with its own collection of uncertainties.
The idea behind the observations leading to
this paper is to study the chemical and physical
properties of a large number of PNe and H ii re-
gions using the same set of observations, the same
data reduction and analysis techniques, and iden-
tical abundance determination methods, in order
to avoid all biases due to the stellar vs. nebu-
lar analysis. The chemical properties of PNe and
H ii regions will provide us with snapshots at two
epochs in the life-time of M33, in particular of its
metallicity gradient. In the present paper we show
the results obtained from the PN data. We have
Recent determinations,
obtained the first sizable sample of M33 PNe with
uniformly derived abundances in order to build the
first sound gradient determination from PNe. In
a forthcoming paper we will present our own H ii
region results, and we will compare them with the
PN results presented here.
This paper is organized as follows: in § 2 we
present the observations and data reduction, in
§ 3 we discuss the plasma diagnostics used and
how we determine the abundances of the PNe, and
in § 4 we describe the PN properties. In § 5 we
compare the results with the PN populations in
other galaxies. The chemical abundance gradients
and their implications in the evolution of M 33
are discussed in § 6 and in § 7. Finally in § 8 we
summarize our results, and present the conclusions
of this work.
2. Observations and data reduction
We obtained spectra of 102 PNe and 48 H ii
regions in M33 using the MMT Hectospec fiber-
fed spectrograph (Fabricant et al. 2005) which is
equipped with an Atmospheric Dispersion Correc-
tor. The spectrograph was used with a single
setup: 270 mm−1grating at a dispersion of 1.2
˚ A pixel−1. The resulting total spectral coverage
ranged from approximately 3600˚ A to 9100˚ A, thus
including the basic emission-lines necessary for the
determination of their physical and chemical prop-
erties. The instrument deploys 300 fibers over a
1◦diameter field of view and the fiber diameter is
∼ 1.5′′(6 pc using a distance of 840 kpc to M 33).
The 102 PNe were selected from the catalog of
Ciardullo et al. (2004), including four new PNe
discovered therein at large galactocentric radii.
In Table 1 we present the list of the observed
PNe. Column (1) gives the identification number
from Ciardullo et al. (2004) except for PNe 153
to 156 which are from LGC observations (Corradi
& Magrini 2006); columns (2) and (3) give the
equatorial coordinates, RA and DEC at J2000.0,
and column (4) gives the [O iii] magnitude from
Ciardullo et al. (2004) following the definition by
Jacoby (1989).
The H ii regions of our sample were chosen ei-
ther based on their [O iii] brightness, or among
those whose chemical abundances had already
been measured by others so we can use them as
a control sample. In this paper we only use the
3

Page 4

Table 1
Observed PNe
Id
(1)
RA J2000.0 DEC
(2)
M [OIII]
(4) (3)
PN001
PN002
PN003
1:32:09.04
1:32:26.54
1:32:38.03
30:22:05.700
30:25:49.800
30:24:00.603
23.51
23.53
21.82
Note.—(1) Identification number from Ciardullo
et al. (2004), except for PNe 153 to 156, which
are from LGC observations (Corradi & Magrini
2006); (2), (3) galactic coordinates at J2000.0; (4)
[O iii] magnitude following the definition by Ja-
coby (1989).Table 1 is published in its entirety
in the electronic edition of the Astrophysical Jour-
nal. A portion is shown here for guidance regarding
its form and content.
data of the H ii regions to test our procedures by
comparing our fluxes with those in the literature.
We used a large number of fibers to take sky
spectra for sky-subtraction.
mize the sky-subtraction we selected low diffuse-
emission areas on the face of M33 from the INT
Hα and [O iii]-continuum frames (Magrini et al.
2000). Since the continuum from the PN central
star can not be detected at the M33 distance any
continuum emission in the spectra has to be aris-
ing from an underlying unresolved stellar popu-
lation. However, the continuum emission in our
spectra is in most cases negligible, and therefore
the Balmer absorption lines coming from the stel-
lar background are unimportant.
Five 1800 s exposures were taken on the night of
October 13, 2007, and three additional 1800 s ex-
posures of the same field were obtained on Novem-
ber 12, 2007. The airmass during the observa-
tions ranged between 1.07 and 1.4 during the night
of October 13, and from 1.3 to 1.5 on November
12. The seeing was 1.2′′and 1.7′′, respectively.
Several dome-flat and sky-flat exposures were ob-
tained during the nights of observations to perform
the data reduction. Arc exposures with the cali-
bration lamp He-Ne-Ar were taken for wavelength
calibration.
The spectra were reduced using the Hectospec
package. All observations were bias subtracted,
overscan corrected, and trimmed. The science ex-
posures were flat-fielded and combined together
to eliminate cosmic rays, and the one-dimensional
spectra were extracted and wavelength calibrated.
In order to opti-
The relative flux calibration was done observing
the standard star Hiltm600 (Massey et al. 1988)
during the nights of October 15 and November
27. The standard star was observed with airmass
∼1.1. The emission-line fluxes were measured with
the package SPLOT of IRAF2. Errors in the fluxes
were calculated taking into account the statisti-
cal error in the measurement of the fluxes, as well
as systematic errors of the flux calibrations, back-
ground determination, and sky subtraction.
The observed line fluxes were corrected for the
effect of the interstellar extinction using the ex-
tinction law of Mathis (1990) with RV=3.1.
We derived c(Hβ), the logarithmic nebular ex-
tinction, by using the weighted average of the
observed-to-theoretical Balmer ratios of Hα, Hγ,
and Hδ to Hβ (Osterbrock & Ferland 2006).
Table 2 gives the results of our line measure-
ments and extinction corrections.
gives the PN name; column (2) gives the nebular
extinction coefficient c(Hβ) with its error; columns
(3) and (4) indicate the emitting ion and the rest-
frame wavelength in˚ A; columns (5), (6), and (7)
give the measured line fluxes (Fλ), their absolute
errors (∆(Fλ), and finally the extinction corrected
fluxes (Iλ).Both Fλ and Iλ are normalized to
Hβ=100.
In order to confirm the goodness of our spec-
Column (1)
2IRAF is distributed by the National Optical Astronomy
Observatory, which is operated by the Association of Uni-
versities for Research in Astronomy (AURA) under coop-
erative agreement with the National Science Foundation
4

Page 5

Table 2
Observed and de-reddened fluxes.
Id
(1)
c(Hβ)
(2)
Ion
(3)
λ (˚ A)
(4)
Fλ
∆(Fλ)Iλ
(7)(5)(6)
PN0010.323±0.026 HI
HeII
HI
[OIII]
[OIII]
HI
HI
HI
[OIII]
[OIII]
HI
[OII]
HI
HeI
[NeIII]/HI
HI
HI
[OIII]
HeI
HI
[OIII]
[OIII]
[NII]
HeI
[NII]
HI
[NII]
HeI
[SII]
[SII]
HeI
[ArIII]
4340
4686
4861
4959
5007
6563
4340
4861
4959
5007
6563
3727
3835
3889
3968
4100
4340
4363
4471
4861
4959
5007
5755
5876
6548
6563
6584
6678
6717
6731
7065
7135
47.6
58.0
100.0
304.2
917.8
362.2
42.3
100.0
304.1
953.0
368.3
101.8
7.7
9.1
7.6
17.3
37.5
7.4
4.4
100.0
178.5
540.1
3.7
22.4
43.5
384.8
128.2
5.2
7.2
10.3
9.7
14.8
3.7
4.4
4.8
7.4
12.2
8.0
3.9
4.8
7.5
12.
52.4
60.1
100.0
298.9
894.0
289.0
46.9
100.0
298.5
927.04
289.70
137.2
10.1
11.7
PN002 0.343±0.030
8.
PN003 0.460±0.014 1.6
0.8
0.8
0.6
0.8
1.2
0.7
0.7
1.8
2.0
3.0
0.7
1.0
1.0
2.4
1.6
0.7
0.6
0.8
0.8
0.8
9.6
21.1
43.1
8.5
4.9
100.0
174.2
520.2
3.0
18.0
31.6
280.0
92.7
3.7
5.1
7.3
6.5
9.9
Note.—(1) PN name; (2) nebular extinction coefficient c(Hβ) with its error;
(3) emitting ion; (4) rest-frame wavelength in˚ A; (5) measured line fluxes; (6)
absolute errors on the measured line fluxes; (7) extinction corrected line fluxes.
Both Fλ (5) and Iλ (7) are expressed on a scale where Hβ=100. Table 2 is
published in its entirety in the electronic edition of the Astrophysical Journal.
A portion is shown here for guidance regarding its form and content.
5

Page 6

troscopic calibration we compare our emission-line
flux measurements of both H ii regions and PNe
with previously published fluxes of the same ob-
jects. In Figure 1 we plot our measured fluxes
against the ones from the literature, where the
two dashed lines mark differences of ±0.15 dex
between the two sets. We found a good agree-
ment between the sets, especially for bright emis-
sion lines, giving us confidence of a sound spectral
calibration.
In Figure 2 we show, as an example of the data
quality, the spectra of three PNe with different
excitation.
3.Plasma Diagnostics and Abundances
3.1. Electron Densities and Temperatures
We have used the extinction-corrected intensi-
ties to obtain the PN electron densities and tem-
peratures. In order to calculate the electronic
densities we used the doublet of the sulfur lines
[S ii]λλ6716,6731, while for the electron temper-
atures we used the ratios [O iii]λ4363/(λ5007 +
λ4959) and [N ii]λ5755/(λ6548 + λ6584).
performed plasma diagnostics by using the 5-level
atom model included in the nebular analysis pack-
age in IRAF/STSDAS (Shaw & Dufour 1994).
We have also used the standard forbidden line
diagnostics IRAF routines in nebular to deter-
mine the electron temperatures. We have deter-
mined the low- and medium-excitation tempera-
tures from the [O iii] and the [N ii] line ratios,
respectively (see also Osterbrock & Ferland 2006,
§5.2). The temperature uncertainties have been
estimated by formal error propagation of the ab-
solute errors on the line fluxes (see also Table 4).
The average relative uncertainties in the determi-
nation of Te[O III] and Te[N II] are of the order
of 5% and 15%, respectively.
The [O iii] and [N ii] temperature diagnostics
are available, respectively, for 32 and 8 PNe of our
sample. Only for six PNe we can use both temper-
ature diagnostics. In order to improve the size of
the PN sample with temperature determinations
we looked for correlations between the electron
temperature and various diagnostics of the neb-
ular excitation, similarly to the approach followed
by Kaler (1986) for a Galactic PN sample.
In Figure 3 we plot the [O iii] electron tem-
We
Fig. 1.— Emission-line flux measurements of both
H ii regions (filled circles) and PNe (empty cir-
cles) are compared to previously published fluxes.
The fluxes of the H ii regions have been taken
from Kwitter & Aller (1981); Vilchez et al. (1988);
Crockett et al. (2006), M07a; RS08, and the PN
ones are from Magrini et al. (2003).
Fig. 3.— [O iii] electron temperature vs. Iλ4686
line intensity, scaled for IHβ=100 and corrected for
extinction. The solid line is the fit done with the
fitexy routine, shown in Eq. (1).
6

Page 7

Fig. 2.— Spectra of three PNe with different excitation selected from our sample in order to show the data
quality.
7

Page 8

perature against the intensity of the Iλ4686 line,
scaled for IHβ=100 and corrected for extinction.
Both quantities are plotted with their formal er-
ror bars. Planetary nebulae in this plot are those
hot enough for helium to be doubly ionized. We
find a clear correlation between the [O iii] electron
temperature and the intensity of the Iλ4686 line,
an effect that is expected as the central stars heat
the medium-high excitation nebulae. In order to
use this correlation in our temperature calculation
we have applied the routine fitexy in Numerical
Recipes (Press et al. 1992) to fit the relation be-
tween the two variables, taking into account their
errors, and minimizing χ2.
For Iλ4686>0 we find:
Te[O III] = (63±8) Iλ4686+(11,000±200), (1)
where Te[O III] is given in K and Iλ4686is scaled
to Hβ=100.Note that using a different fitting
routine, such as sixlin (Isobe et al. 1990), we de-
termine temperatures that agree with the ones of
Eq. (1) at the 1% level. We infer that we are able
to estimate the temperatures to better than 10%
with the above formulation.
If Iλ4686=0 we assume a constant value of Te[O
III], as in Kaler (1986). The average Te[O III] for
Iλ4686=0 is 12100 ±1500 K, which is compatible
with the continuity of Eq. (1). Note that both this
value and Eq. (1) give temperatures that are con-
sistently higher than those found by Kaler (1986)
in Galactic PNe, an effect that is probably due to
the lower metallicity that characterizes the M33
stellar populations (see Stanghellini et al. (2003)
for a discussion on how metallicity affects the neb-
ular temperatures).
In Figure 4 we show Te[N II] against Iλ4686,
where both quantities are plotted with their for-
mal errors. The line represents the mean value of
Te[N II]= 13600 ±2100 K, which has been derived
using the (inverse squared) relative temperature
errors as weights. Once again, while the physi-
cal behavior of the temperatures is similar to that
found in Kaler (1986), the actual values are higher,
possibly reflecting the lower efficiency of the cool-
ing process at low metallicity. For Iλ4686 >0 we
adopt Te[N II]=13,600 K.
If Iλ4686=0, we expect Te[N II] to decrease with
the stellar temperature. The only way to frame
this decline is to use the ratio between the fluxes
of the singly-ionized to the doubly-ionized oxygen.
By correlating the oxygen line strengths, corrected
for extinction, to the electron temperatures for the
PNe with measured [N ii] lines we find:
Te[N II] = (11,700±2800)−(2800±4000)log
Iλλ3727−29
Iλ4959+ Iλ5007.
(2)
Note that in our spectra the λλ3727 − 29
lines are always blended.
are not available, and Iλ4686=0 we assume Te[N
II]=14,100 ±2800 K, which, in this case, is the
weighted mean of the temperatures.
If the oxygen lines
3.2.Chemical abundances
We computed the ionic abundances using the
nebular analysis package in IRAF/STSDAS (Shaw
& Dufour 1994). The elemental abundances were
then determined by applying the ionization correc-
tion factors (ICFs) following the prescriptions by
Kingsburgh & Barlow (1994) for the case where
only optical lines are available. As discussed in
the previous §, the [O iii] λ4363 emission line was
measured with a sufficiently high signal to noise
ratio in 32 PNe, affording the direct determina-
tion of the Te[O III]. We could also directly mea-
sure Te[N II] in 8 PNe. For the other targets we
adopted the temperatures derived from the rela-
tions we have described in the previous Section. In
the abundance analysis we have used Te[N II] for
the calculation of the N+, O+, S+abundances,
while Te[O III] was used for the abundances of
O2+, S2+, Ar2+, He+, and He2+.
We calculated the abundances of He i and He ii
using the equations of Benjamin et al. (1999) in
two density regimes, i.e.
≤1000 cm−3. The Clegg’s collisional populations
were taken into account (Clegg 1987).
the effect of double collisions not being properly
corrected, the He+/H ionic abundance from the
λ7065 emission-line is quite different from the
ionic abundance computed from λλ4471, 5876,
and 6678. We thus omit the abundance derived
from the former line from the average. The com-
puted individual ionic and elemental abundances
of each PN are shown in Table 3, where column (1)
gives the PN identification name; columns (2) and
(3) show the plasma diagnostic and abundances
available for each PN, where in column (2) we la-
bel each diagnostic, and in column (3) we give the
ne >1000 cm−3and
Due to
8

Page 9

Table 3
Plasma Diagnostics and Abundances
Id
(1) (2) (3)
PN001
Te(OIII)
Te(NII)
HeII/H
He/H
OIII/H
ICF(O)
O/H
14650b
13600b
0.063
0.063
7.586 10−05
1.000
7.586 10−05
PN002
Te(OIII)
Te(NII)
OIII/H
ICF(O)
O/H
14650b
13600b
7.884 10−05
1.000
7.884 10−05
PN003
Te(OIII)
Te(NII)
Te(OIII)
Te(NII)
Ne(SII)
HeI/H
He/H
OII/H
OIII/H
ICF(O)
O/H
NII/H
ICF(N)
N/H
ArIII/H
ICF(Ar)
Ar/H
SII/H
ICF(S)
S/H
14000a
15200a
12060b
14120b
2540
0.125
0.125
1.590 10−05
6.500 10−05
1.000
8.090 10−05
7.168 10−06
5.087
3.646 10−05
4.47 10−07
1.245
8.359 10−07
1.736 10−07
1.276
1.665 10−06
acomputed from electron tempera-
ture diagnostic lines.
bderived from the relations in §3.
Note.—(1) identification name; (2)
label of each plasma diagnostic and
abundancesavailable;
value obtained from our analysis. Ta-
ble 3 is published in its entirety in the
electronic edition of the Astrophysical
Journal. A portion is shown here for
guidance regarding its form and con-
tent.
(3)relative
9

Page 10

Table 4
Typical Errors in dex
Magnitude
(1)
Te
(2)
∆(He/H)
(3)
∆(O/H)
(4)
∆(N/H)
(5)
∆(Ne/H)
(6)
∆(Ar/H)
(7)
∆(S/H)
(8)
N. PNe
(9)
20.60-21.00l
c
l
c
l
c
l
c
l
c
l
c
c
c
c
0.03
-
0.04
0.09
0.04
0.08
0.08
0.11
0.13
0.17
0.09
0.21
0.11
-
-
0.025
0.17
0.03
0.20
0.03
0.21
0.04
0.21
0.09
0.20
0.07
0.23
0.21
0.45
0.5
0.06
0.40
0.06
0.34
0.04
0.34
0.07
0.40
0.16
0.40
0.09
0.40
0.34
-
-
0.05
0.43
0.06
0.43
0.08
0.43
0.15
0.43
0.16
0.43
0.13
0.43
0.43
-
-
0.06
0.34
0.11
0.30
0.09
0.34
0.17
0.40
0.17
0.43
0.18
0.43
0.43
-
-
0.09
0.43
0.11
0.30
0.09
0.30
0.11
0.30
0.11
0.30
0.16
0.30
0.43
-
-
8
1
9
8
5
8
6
16
4
9
1
12
5
2
1
21.01-21.50
21.51-22.00
22.01-22.50
22.51-23.00
23.51-24.00
24.01-24.50
24.51-25.00
25.01-25.50
Note.—(1) Magnitude range; (2) method of the electron temperature determination, where l means that
the temperature has been measured based on an emission line and c means that we used the correlations
described in §3; (3-8) typical errors in dex of the total abundances; (9) number of PNe available in each bin.
Fig. 4.— Same as Figure 3 but for the Te[N II]
electron temperature. The solid line represents the
mean value of Te[N II]= 13600 ±2100K, which has
been derived using the relative temperature errors
as weights.
relative value obtained from our analysis.
The formal errors on the ionic and total abun-
dances were computed taking into account the un-
certainties in the observed fluxes, in the electron
temperatures and densities, and in c(Hβ). The
errors (in dex) of the final abundances are given
in Table 4. Typical errors are given for PNe in
different [O iii] magnitude ranges, taking also into
account whether the electron temperatures were
derived from emission-line diagnostic or from the
relations described in § 3.1. In the latter case, we
assumed a percentage error on the electron tem-
perature of ∼10%, which is the dispersion of the
relation used to derive the temperature. The first
column of Table 4 gives the magnitude range; col-
umn (2) gives the method used for the electron
temperature determination, where l means that
the temperature has been measured based on an
emission line and c means that we used the cor-
relations described in §3; columns (3) to (8) give
the typical errors in dex of the total abundances.
The last column gives the number of PNe on each
bin.
4. The PN population
Most of the PNe discovered in M33 belong to
its disk, with only two PNe having radial velocities
compatible with the halo population (Ciardullo et
10

Page 11

Table 5
Average chemical abundances and abundance ratios
Sample
(1)
He/H
(2)
O/H (10−4)
(3)
N/H (10−4)
(4)
Ne/H (10−5)
(5)
Ar/H (10−6)
(6)
S/H (10−6)
(7)
N/O
(8)
Ne/O
(9)
M33 PNe, Disk (91)
M33 PNe, non-Type I (72)
M33 PNe, Type I (19)
M33 PNe, Halo (2)
M33 HII regions
Solar value(d)
LMC PNe
SMC PNe
Galactic PNe
0.118±0.074
0.114±0.083
0.135±0.027
0.120:
0.101±0.015(a)
0.085±0.02
0.103±0.026
0.113±0.022
0.123±0.042
1.96±1.27
1.73±0.93
2.70±1.88
2.62±0.31
2.04±0.75(b)
4.57±0.04
2.32±1.65
1.05±0.46
3.53±1.95
1.09±1.54
0.39±0.19
1.66±1.88
0.84:
1.28±0.57(a)
0.60±0.09
1.48±1.75
0.28±0.33
2.44±3.46
4.09±3.08
3.65±2.11
4.83±4.25
3.47±0.82
4.24±2.61(c)
6.9±1.0
4.04±3.60
1.77±1.32
9.68±7.98
1.19±0.57
1.12±0.50
1.38±0.68
1.24±0.28
1.31 ±0.45(a)
1.51±0.30
1.14±0.72
0.59±0.59
1.26±1.24
3.34±1.71
4.46±2.30
6.64±3.80
6.45:
5.85±2.28(a)
13.8±2.0
3.46±8.88
4.80±6.57
...
0.40±0.32
0.18±0.06
0.58±0.33
0.31:
0.06±0.02(a)
0.13±0.10
0.87±1.15
0.28±0.50
0.67±0.82
0.17±0.06
0.17±0.06
0.16±0.04
0.13±0.03
0.20±0.06(c)
0.15±0.05
0.17 ±0.09
0.17 ±0.08
0.25±0.10
Note.—(1) the PN sample; (2-7) average chemical abundances by number with their rms uncertainties; (8-9) <N/O> and <Ne/O>, the average of N/O and Ne/O
values for each object.
aChemical abundances computed using the sample of M07a, extended to all literature data with direct electron temperature measurement.
bO/H is calculated with the cumulative sample of M07a, which includes also all previous O/H with direct electron temperature measurement, and RS08.
cNe/H and Ne/O are from Crockett et al. (2006).
dSolar value from Asplund et al. (2005).
:Value from a single object.
11

Page 12

al. 2004). Therefore, we have the opportunity to
analyze in detail a pure disk PN population by ex-
cluding from our analysis the two suspected halo
PNe (namely PN067 and PN024). We have also
excluded from our analysis the PN039 since its
physical conditions do not allow us to obtain abun-
dance diagnostics. PN039 has a WR-nucleus and
will be discussed in detail in the Appendix.
4.1. The He/H vs N/O diagram
The plot of the N/O ratio vs the He/H one is a
classical diagnostic diagram used to discriminate
PNe of different types, i.e.
progenitors. The N/O ratio provides information
about the stellar nucleosynthesis during the AGB
phase of LIMS. On the one hand, nitrogen is pro-
duced in AGB stars in two ways: by neutron cap-
ture, during the CNO cycle, and by hot-bottom
burning. Hot-bottom burning produces primarily
nitrogen but occurs only if the base of the con-
vective envelope of the AGB stars is hot enough
to favor the conversion of12C into14N. Thus ni-
trogen is expected to be mostly enriched in those
PNe with the most massive progenitors, i.e. with
turnoff mass larger than ∼3 M⊙ (van der Hoek
& Groenewegen 1997; Marigo 2001). The He/H
ratio gives also an indication of the initial mass
of the progenitor, the nebula is enriched progres-
sively for more massive stars, it reaches a plateau
between 3 and 4 M⊙, and then it increases again
toward the higher masses (Marigo 2001).
In order to make a selection of the PNe with
high-mass progenitors, in Figure 5, we plot the
M33 disk PNe on the He-N/O plane. Peimbert &
Torres-Peimbert (1983) found that Galactic PNe
that were nitrogen- and helium-enriched also had
bipolar shape, and were located closer to the
Galactic plane. It looked like these PNe formed
a younger population, which they defined as those
with log (N/O) ≥ −0.3 and He/H≥∼0.125, and
called them Type I PNe.
low (1994) analyzed a much larger sample and re-
define the Galactic Type I PNe as those having
log(N/O)≥-0.1, with 18% of the PNe in their sam-
ple being of Type I. Dopita (1991) analyzed LMC
PNe and noted that the Type I definition needed
a revision to allow for the lower metallicity of the
LMC with respect to the Galaxy, setting the limit
of Type I PNe to log(N/O)≥-0.5 for the LMC.
Leisy & Dennefeld (2006) basically confirmed this
with different mass
Kingsburgh & Bar-
limit. Magrini et al. (2004) also noted that the
definition of Type I PNe depends on the metallic-
ity, because the amount of nitrogen that can be
produced by hot bottom burning is dependent on
the amount of carbon present, and also because
the oxygen abundance depends on the metallicity
of the galaxy. Since the oxygen metallicity of M33
is very similar to that of the LMC, we use the Type
I limits as in Dopita (1991), namely, Type I PNe
are those with log(N/O)>-0.5, independent of he-
lium abundance. With this definition we found 19
PNe being Type I in the disk of M33 (∼20% of the
whole disk population) and one in the halo.
4.2. The Ne/H and S/H vs O/H diagrams
The study of the chemical evolution of galaxies
needs strong constraints regarding the past com-
position of the ISM, and PNe can supply them,
in particular with their soundly-determined oxy-
gen abundances. The questions is whether oxygen
(and neon) is modified in LIMS while in their AGB
phase.
A possible way to verify this is to study the
relationship between neon and oxygen. These el-
ements derive both from primary nucleosynthesis,
mostly in stars with M>10 M⊙. If the O/H and
Ne/H abundances are really independent of the
evolution of the PN progenitors through the AGB
phase, a tight correlation between their abundance
should be observed.
In Figure 6 we show O/H against Ne/H for the
55 M33 PNe where both abundances were avail-
able. The slope of the correlation is close to unity,
0.90±0.11, with a correlationcoefficient Rxy=0.81,
pointing at a locked variation of these two ele-
ments. Recent results from Wang & Liu (2008)
indicate that oxygen and neon could be manufac-
tured with similar yields also in LIMS, but only at
very low metallicities, 12 + log(O/H) < 8. These
results are however still not explained by cur-
rent nucleosynthesis theories that predict different
channels, and thus different yields, for Ne and O
production in LIMS at low metallicity (cf., e.g.,
Karakas & Lattanzio 2003, Marigo et al. 2003).
Our findings confirm that in M33, as in LMC, and
the Galaxy, there is no evidence of enhancement
of oxygen and neon in PNe.
Similar correlations are expected also for sul-
phur versus oxygen, since sulphur is manufactured
12

Page 13

by massive stars.
fur versus oxygen abundances of the 38 M33 PNe
where these abundances are available. Their rela-
tionship has a slope of 0.97±0.22 and a correlation
coefficient of 0.51, showing moderately good corre-
lation. We refrain from showing and analyzing the
argon to oxygen relation. As seen in Stanghellini
et al. (2006), these quantities might not correlate
due to their different origin.
ments are manufactured in massive stars, argon is
synthesized in very different α -element processes
than oxygen, thus a lockstep of these elements is
not completely expected.
In Figure 7 we plot the sul-
Even if both ele-
5. Comparison of abundances of different
stellar populations, and of PN popula-
tions of other galaxies
The average chemical abundances of our M33
PNe are reported in Table 5 where the first col-
umn describes the selected sample, columns (2)
to (7) give the average total abundance of He/H,
O/H, Ne/H, Ar/H, and S/H, whereas the last
two columns, (8) and (9), give the mean values
of N/O and Ne/O. The average abundances were
computed excluding those abundances from up-
per limit detections.When the emission lines
for temperature diagnostics were not detected, we
adopted the electron temperatures discussed in
§3.1. Their larger uncertainties were taken into
account by larger formal errors in the total chem-
ical abundances as described in Table 4.
We group the PNe in M33 into four popula-
tions: i) the whole PN disk population; ii) the
disk population excluding those PNe which are be-
lieved to be Type I; iii) the Type I PNe, and iv)
the halo PN population as classified by Ciardullo
et al. (2004). For the sake of comparison we add
to the Table the average chemical abundances of
the M33 H ii regions, the solar abundances, and
the average abundances of the PNe in LMC, SMC
(Stanghellini 2008) and the Galaxy (Stanghellini
et al. 2006). Note, however, that the population of
the Type I and non-Type I classes remains some-
what uncertain, since about a dozen PNe could
belong to either class based on their formal abun-
dance errors. However, the chemical abundances
and distributions of Type I and non-Type I PNe
in M33 are very similar, so that the displacement
of a few PNe from a group to the other would not
Fig. 5.— log(N/O) versus 12+log(He/H). Type
I PNe and non-type I PNe are plotted as filled
circles and empty circles respectively, according to
the definition of Dopita (1991). The continuous
line mark log(N/O)>-0.5 that, independently of
helium abundance, defines the area of Type I PNe.
Fig. 6.— The relationship between oxygen and
neon abundances. Symbols are as in Figure 5. The
continuous line is the weighted least square fit to
the complete sample of PNe.
13

Page 14

change our results significantly.
An inspection of Table 5 affords several inter-
esting clues on the chemical evolution of galaxies.
First, if we compare the average abundances of
the α-elements of the M33 disk PNe with those of
the H ii regions, we do not note significant enrich-
ment in the H ii regions, leading to the conclusion
that these elements are not substantially modified
in the LIMS evolution. We substantially found
a flat age-metallicity relationship in M33 all the
way to the time of PN formation. Second, we in-
fer that the metallicity of the PNe in M33, based
on oxygen abundances, is sub-solar. If we consider
the entire sample of disk PNe in M33 we see that
the averages of α-elements are virtually identical
to those of PNe in the LMC, with PN metallicity
about 1/2 that of the Galaxy, and twice the SMC
on average, both in oxygen and neon.
From Table 5 we see that the average Ne/O ra-
tio in the general population of M33 PNe is iden-
tical to that of the LMC, but lower than the mean
Ne/O in Galactic PNe. This finding is right on the
mark with the recent results of Wang & Liu (2008).
In fact Wang & Liu (2008) found that the Ne/O
ratio increases with increasing oxygen metallicity
both in PNe and in H ii regions. This suggests
a different enrichment history of neon and oxygen
in the ISM and thus probably different produc-
tion mechanisms of these two α-elements in mas-
sive stars, in agreement with current theoretical
calculations by Kobayashi et al. (2006). While we
should expect the solar Ne/O ratio to be consistent
with that of Galactic disk PNe and H ii regions,
we note that the Asplund et al.’s (2005) value,
as well as Grevesse & Sauval’s (1998), is slightly
discrepant in this scenario. Both PNe and H ii re-
gions in the Galactic disk give a consistent Ne/O
ratio of ∼0.25, higher than the solar value. Our
results on M33 PNe seem thus to be consistent
with the recent suggestions of a revision of the so-
lar Ne/O ratio and the absolute neon (e.g. Wang
& Liu 2008, Rubin et al. 2008).
The N/O average shown in Table 5 for M33
PNe is lower than the Galactic and higher than the
SMC values. The comparison to the LMC average
has limited importance, giving the large range of
LMC N/O values.
6.The abundance gradients
In Figure 8, we show the oxygen abundance
as a function of galactocentric distance for our
sample of disk PNe. The galactocentric distances
were computed by adopting a distance to M33
840 kpc, as determined by Freedman et al. (1991),
an average and well-accepted distance estimate.
A weighted linear least-square fit to the complete
disk sample gives a gradient:
12+log(O/H) = −0.031(±0.013)RGC+8.44(±0.06),
(3)
where RGC is the de-projected galactocentric
distance in kpc, computed assuming an inclination
of 53◦, and a position angle of 22◦. In Figures 9,
and 10 we plot the radial metallicity gradients of
Ne/H and S/H. The solid lines are the weighted
linear fits to the complete sample of disk PNe (first
row of Table 6 for each element). In all these fig-
ures the two symbols refer to Type I (filled circles)
or non-Type I (empty circles) PNe.
In Table 6 we show the slopes (column 2) and
zero-points (chemical abundances in the centre of
the galaxy, column 3) of the metallicity gradients
considering different samples of PNe. The sample
used and the number of PNe included are given
in columns (4) and (5) respectively. For each el-
ement, the first row report the gradient obtained
considering the whole sample of disk PNe (thus
excluding the two possible halo PNe), the second
row gives the gradient computed using non-Type I
PNe, and the third row gives the gradient obtained
for Type I PNe.
The gradients are computed with a weighted
least mean square fit. For each element the slope
of the three different samples are consistent within
the errors. Note that the metallicity gradient of
an old stellar population could be affected by the
radial migration of stars during cosmic times. The
present-time location of a PN could not necessar-
ily be the place where it was born. Radial mixing
of stars is believed to be due to several mecha-
nisms, among them the diffusion of stars on their
orbits because of various irregularities in the galac-
tic potential (cf. Wielen et al. 1996), to the pas-
sage of spiral patterns (cf.
2006), to changes in the angular momentum due
to non-asymmetric forces due to molecular clouds
Minchev & Quillen
14

Page 15

(Spitzer & Schwarzzchild 1953). Upper limits to
the radial migration rate were estimated by De Si-
mone et al. (2004) and by Haywood (2008), giving
similar values around 1.5-3.7 kpc Gyr−1, limited
to a radial region ±2 kpc from the birth place. The
model of Sellwood & Binney (2002) show that spi-
ral waves in galaxy disks churn both the interstel-
lar medium and the stars, affecting their metallic-
ity gradients. The effect of stirring of the entire
disk due to the spiral waves is a flattening of the
metallicity gradient.
However, the fact that metallicity gradients sur-
vive and are well observed in disk galaxies seem
to indicate that the effectiveness of the migration
processes is moderate.
7.Discussion
The metallicity gradient of PNe allows to an-
alyze how was the metal distribution in the past
epochs of M33. In M33 we have to possibility to
compare the metallicity gradient of PNe with a
good number of other measurements of abundance
gradients, in particular from H ii regions. The ad-
vantage of comparing PNe to H ii regions is the
similarity of their emission-line spectra in spite of
their different evolutionary states that allows to
use the same observation techniques, analysis, and
abundance determinations.
In order to compare our PN results with the
leading samples of M33 H ii regions we build the
largest, most reliable sample available to date. As
already noted by e.g. RS08 and M07a, the size
and quality of the sample is fundamental to de-
rive reliable metallicity gradients. In H ii regions
there is a substantial intrinsic scatter of 0.11 dex
in the metallicity at any given distance from the
M33 center, which imposes a fundamental limit on
the accuracy of gradient measurements that rely
on small samples of objects (RS08). Also, in the
past, only giant H ii regions have been studied,
and those might have a steeper metallicity gradi-
ent than small and compact H ii regions.
We have re-calculated the H ii regions metallic-
ity gradients considering a cumulative sample, in-
cluding that by RS08, consisting of 61 H ii regions
with accurate abundance determinations, and that
by M07a that included new and literature spec-
troscopy performed to a set of 28 H ii regions.
The cumulative sample of 89 H ii regions gives,
Fig. 7.— The relationship between oxygen and
sulphur abundances. Symbols are as in Figure 5.
The continuous line is the weighted least square
fit to the complete sample of PNe.
Fig. 9.— The radial gradient of neon abundance.
Symbols are as in Fig. 5. The continuous line is the
weighted least square fit to the complete sample
of disk PNe. Slopes and zero-points are shown in
Table 6.
15