The relation between surface star formation rate density and spiral arms in NGC 5236 (M83)

Abstract

For a long time the consensus has been that star formation rates are higher
in the interior of spiral arms in galaxies, compared to inter-arm regions.
However, recent studies have found that the star formation inside the arms is
not more efficient than elsewhere in the galaxy. Previous studies have based
their conclusion mainly on integrated light. We use resolved stellar
populations to investigate the star formation rates throughout the nearby
spiral galaxy NGC 5236. We aim to investigate how the star formation rate
varies in the spiral arms compared to the inter-arm regions, using optical
space-based observations of NGC 5236. Using ground-based H\alpha images we
traced regions of recent star formation, and reconstructed the arms of the
galaxy. Using HST/ACS images we estimate star formation histories by means of
the synthetic CMD method. Arms based on H\alpha images showed to follow the
regions where stellar crowding is higher. Star formation rates for individual
arms over the fields covered were estimated between 10 to 100 Myr, where the
stellar photometry is less affected by incompleteness. Comparison between arms
and inter-arm surface star formation rate densities (\Sigma$_{SFR}$) suggested
higher values in the arms (\sim0.6 dex). Over a small fraction of one arm we
checked how the \Sigma$_{SFR}$ changes for the trailing and leading part. The
leading part of the arm showed to have a higher \Sigma$_{SFR}$ in the age range
10-100 Myr. Predictions from the density wave theory of a rapid increase in the
star formation at the edge where the stars and the gas enter the density wave
are confirmed. The \Sigma$_{SFR}$ presents a steep decrease with distance from
the center of the arms through the inter-arm regions.

Full text

Page 1

arXiv:1111.1249v1 [astro-ph.CO] 4 Nov 2011
Astronomy & Astrophysics manuscript no. paper-17432
November 8, 2011
c ? ESO 2011
The relation between surface star formation rate density and
spiral arms in NGC 5236 (M83)
E. Silva-Villa and S. S. Larsen
Astronomy Institute, University of Utrecht, Princetonplein 5, 3584 CC, Utrecht, The Netherlands
e-mail: [e.silvavilla,s.s.larsen]@uu.nl
Preprint online version: November 8, 2011
ABSTRACT
Context. For a long time the consensus has been that star formation rates are higher in the interior of spiral arms in
galaxies, compared to inter-arm regions. However, recent studies have found that the star formation inside the arms
is not more efficient than elsewhere in the galaxy. Previous studies have based their conclusion mainly on integrated
light. We use resolved stellar populations to investigate the star formation rates throughout the nearby spiral galaxy
NGC 5236.
Aims. We aim to investigate how the star formation rate varies in the spiral arms compared to the inter-arm regions,
using optical space-based observations of NGC 5236.
Methods. Using ground-based Hα images we traced regions of recent star formation, and reconstructed the arms of the
galaxy. Using HST/ACS images we estimate star formation histories by means of the synthetic CMD method.
Results. Arms based on Hα images showed to follow the regions where stellar crowding is higher. Star formation rates
for individual arms over the fields covered were estimated between 10 to 100 Myr, where the stellar photometry is
less affected by incompleteness. Comparison between arms and inter-arm surface star formation rate densities (ΣSFR)
suggested higher values in the arms (∼0.6 dex). Over a small fraction of one arm we checked how the ΣSFR changes for
the trailing and leading part. The leading part of the arm showed to have a higher ΣSFR in the age range 10-100 Myr.
Conclusions. Predictions from the density wave theory of a rapid increase in the star formation at the edge where the
stars and the gas enter the density wave are confirmed. The ΣSFR presents a steep decrease with distance from the
center of the arms through the inter-arm regions.
Key words. galaxies: Individual – NGC 5236 galaxies: Star formation
1. Introduction
Studies of OB stars (see e.g. Morgan et al. 1953; McGruder
1975; Muzzio 1979; Kaltcheva 2009), show that the con-
centration of these stars is higher in the Sagittarius arm
of the Galaxy. This suggests an active star formation in
the arms of spiral galaxies. The creation of spiral arms is
a problem that has been studied since the late 60’s (e.g.
Lin & Shu 1964; Roberts 1969). The density wave theory
explains how the spiral arms can be formed and can re-
main stable over time for an isolated galaxy. The theory
predicts more active star formation in the arms, where
the gas compression induced by the density waves triggers
the process (see e.g. Lin & Shu 1964; Bash & Visser 1981;
Knapen et al. 1992, 1996; Kurtz et al. 2002; Grosbøl et al.
2006). Alternative theories to explain the spiral arms in
galaxies have been proposed. In the modal theory, inward-
moving waves reflect or refract off at the center of a galaxy
(even in the galaxies with a bar in the center), and then
the wave comes back out as leading or trailing spiral arms
(e.g.Mark 1974; Lau & Mark 1976; Bertin et al. 1989;
Elmegreen et al. 1992, for theoretical and observational ap-
proaches). Another theory, know as the Stochastic Self-
Propagating Star Formation (Mueller & Arnett 1976), sug-
gests that episodes like supernovae, shock wave or gravita-
tional interactions are responsible to propagate and trigger
star formation (e.g. Gerola & Seiden 1978; Seiden & Gerola
1979; Feitzinger et al. 1981, for theoretical and observa-
tional approaches). This theory is very good at explain-
ing flocculent galaxies, while not grand-design ones. It is
possible that the whole process is a combination of these
theories. However, in this paper, we will test only results
expected from the density wave theory.
If the density wave theory is correct, the process of
gas being compressed by the waves should lead to many
observable effects, e.g. different star formation rates or
color-gradients across the arms. Assuming a constant an-
gular velocity of the spiral density wave and an approx-
imately flat rotation curve of the stellar component, in-
side the corotation radius the gas will overtake the density
wave, which will produce an increase in the star forma-
tion when compressed. Outside the corotation radius the
wave overtakes the gas. Consequently, the stars will drift
and age, creating a color gradient that can be observed.
Mart´ ınez-Garc´ ıa et al. (2009) have studied the color gradi-
ent across spiral arms of 13 spiral galaxies. In their work,
a reddening free index was used to study this process, con-
cluding that azimuthal color gradients are common in spiral
arms of disk galaxies. Previous works have tried to relate
the spiral arms with the star formation (see e.g. Allen et al.
1986; Tilanus & Allen 1993; Stedman & Knapen 2001;
Knapen et al. 2010; S´ anchez-Gil et al. 2011). These stud-
ies have shown the relation between the star formation and
the spiral structure using different gas components over dif-
1

Page 2

E. Silva-Villa and S. S. Larsen: ΣSFR and Arms over NGC 5236
ferent parts of a large set of galaxies (e.g. M83, M51, M100,
M101, among others).
Compression of the gas across the galaxy is expected to
be observed as regions with enhanced star formation. The
(common) components used to estimate the star formation
efficiencies (SFE) in a galaxy are the gas and the stellar
population, the latter being commonly measured through
their integrated light. When H2 is use to trace star form-
ing regions (e.g. Foyle et al. 2010), there seemed to be no
specific correlation between the SFE and the spiral arms.
However, studies that estimate the SFE as the fraction of
Hα to CO and/or HI maps do find a correlation between
the SFE and the spiral arm (e.g.
Cepa & Beckman 1990; Knapen et al. 1992). Nevertheless,
it is important to note that most of the studies of star for-
mation (either efficiency or rate) and their relation with
the arms in spiral galaxies have been done using gas com-
ponents, namely HI, H2, CO, etc combined with SFR trac-
ers (e.g. Hα), and unresolved stellar populations, mea-
sured through their integrated light (e.g. Knapen et al.
1996; Leroy et al. 2008). Tracers based on gas components
are assumed to indicate the regions where SFE is high,
which could lead to erroneous estimations, as stated by
Foyle et al. (2010).
As a new approach, we present in this paper the use of
resolved stellar populations as a tool to study how the sur-
face star formation rate density (ΣSFR) varies in the arms
and inter-arm areas of the nearby spiral galaxy NGC 5236.
This galaxy has been selected because it is one of the clos-
est face-on, nearby, grand-design spiral galaxies, where spa-
tial resolution allows a study of this kind. The low inclina-
tion in the line-of-sight reduces the effect of internal red-
dening. Previous estimations of the ΣSFR show values of
∼ 13 × 10−3M⊙yr−1Kpc−2(e.g. Larsen & Richtler 2000;
Calzetti et al. 2010; Silva-Villa & Larsen 2011), which is
high compared to other normal spiral galaxies (e.g.
Larsen & Richtler 2000, see table 2). A large number of
studies have been done in the center of this galaxy, study-
ing different properties through different wavelengths, all
of them suggesting an increasing activity in the star for-
mation during the last ∼ 10Myr (e.g. Ryder et al. 1995;
Harris et al. 2001; Houghton & Thatte 2008; Knapen et al.
2010). Activity in the center of the galaxy combined with
the high levels of ΣSFRin the disk, show that NGC 5236 is
still actively forming stars, making it ideal to study differ-
ences in the star formation and the possible relations with
environment and location.
Kinematic studies of NGC 5236 were presented by
Lundgren et al. (2004) based on gas (H2+HI) measure-
ments. The estimation of the gas surface density in the
arms done by Lundgren et al. is higher than the Toomre’s
value for stability (Q ∝ Σ−1
gas). This is possibly causing in-
stabilities in the arms of the galaxy, potentially leading to
star formation. However, their estimation of the gas sur-
face density in the inter-arm regions do not show the same
high values. In a further work, Lundgren et al. (2008) used
far-UV, B and Hα integrated light to estimate star forma-
tion rates, while CO was used for gas maps. Lundgren et
al. conclude that the star formation presents higher levels
in the nuclear regions, close where the bar ends, and in the
arms of the galaxy. The authors also found an increased
SFE along the arms of this galaxy.
This paper is structured as follows. We introduce the
observations, optical and Hα, used to study the field stellar
Lord & Young 1990;
population and the regions of recent star formation, respec-
tively, in Sect. 2. Section 3 is devoted to review how the
optical bands were used to run the photometry of the field
stars. Using Hα (i.e. tracing recent star formation), we will
introduce in Sect. 4 the method used to re-construct the
arms of the galaxy. Using the photometry from Sect. 3, we
estimated the star formation history of different groups of
field stars in Sect. 5. Finally, we present our discussion and
conclusions in Sects. 5 and 6, respectively.
2. Observations
We used images from Hubble Space Telescope and Cerro
Tololo observatory, covering different optical wavelengths
and Hα.
2.1. BVI observations and data reduction
We use observations of NGC 5236 taken by the Hubble
Space Telescope (HST), using the Advanced Camera for
Surveys (ACS). The instrument has a resolution of 0.′′05
per pixel. With a distance modulus of 28.27 (∼ 4Mpc,
Thim et al. 2003), 1 pixel corresponds to ∼1 pc in our im-
ages.
The images of this galaxy have been taken in the optical
bands F435W (∼ B), F555W (∼ V ), and F814W (∼ I),
with exposure times of 680 sec for the bands B and V, and
430 sec for the band I. Our observations were taken in 2004
as part of Cycle 12, centered at α : 13 : 37 : 00 and δ : −29 :
49 : 38 and α : 13 : 37 : 06 and δ : −29 : 55 : 28 (J2000)
for the first and second field observed, respectively. Figure
1 presents a DSS image indicating the regions observed.
The standard STScI pipeline was used for the initial
data processing. ACS images were drizzled using the mul-
tidrizzle task (Koekemoer et al. 2002) in the STSDAS pack-
age in IRAF using the default parameters, but disabling
the automatic sky subtraction. Object detection for field
stars was performed on an average B, V, and I image, us-
ing daofind in IRAF.
2.2. Hα observations
We
Observatory (CTIO) 1.5m telescope image of NGC 5236,
taken in 2006 as part of the Survey for Ionization in
Neutral Gas Galaxies (SINGG, Meurer et al. 2006)1. The
survey used the Hα and R bands, with a resolution of
0.′′4 per pixel. The total exposure time of the observation
for NGC 5236 was 1800 sec, with center coordinates at
α : 13 : 37 : 02 and δ : −29 : 52 : 06 (J2000). The image
used in this paper is based only on the Hα filter. For
details on the observations and calibration of the image
see Meurer et al. (2006).
use an archivalCerro TololoInter-American
3. Field stars photometry
Details of our method of analysis can be found in
Silva-Villa & Larsen (2010). Below we will reiterate the
main points of the procedures we used to carry out pho-
tometry on our data.
1SINGG is a subsample of the HI Parkes All Sky Survey
(HIPASS), Meyer et al. (2004)
2

Page 3

E. Silva-Villa and S. S. Larsen: ΣSFR and Arms over NGC 5236
Fig.1. M83 DSS image. Blue lines delineate the ACS point-
ings used in this paper.
Fig.2. Left column: Color-magnitude diagrams for the Arm
1, Arm 2, and inter-arm regions. Right column: Histogram
of colors of main sequence stars inside the boxes marked
over the CMDs. Errors are Poissonian. Vertical dashed lines
represent the peak of the Gaussian distributions.
Due to the crowding, we performed PSF photometry for
field stars. Using a set of bona-fide stars visually selected
in our images, measuring their FWHM with imexam, we
construct our point-spread function (PSF) using the PSF
task in DAOPHOT. This procedure is employed in the same
manner for each band (i.e. B, V, and I). The PSF stars are
selected individually in each band, in order to appear bright
Fig.3. Upper panel: Blurred image used for the detection of
pronounced Hα regions. Lower panel: Original CTIO image
of NGC 5236. Overplotted in each panel are the estimated
locations of the arms. In the lower panel the two fields ob-
served with the HST/ACS are marked as yellow regions.
Green dot-dashed line is the corotation radius located at
170” (Lundgren et al. 2004). Dash-dotted lines in blue de-
note annuli at 200.′′0, 140.′′0 and 110.′′0, see text for details.
and isolated. PSF photometry is done with DAOPHOT in
IRAF.
Our PSF-fitting magnitudes are corrected to a nominal
aperture radius of 0.′′5, following standard procedures. From
this nominal value to infinity, we apply the corrections in
Sirianni et al. (2005).
3

Page 4

E. Silva-Villa and S. S. Larsen: ΣSFR and Arms over NGC 5236
Fig.4. Surface density of stars for the ACS fields. Red lines
represent the path of the arms.
HST zero-points2were applied to the PSF magnitudes
after applying aperture corrections. The zero-points used
in this work are ZPB = 25.77, ZPV = 25.72 and ZPI =
25.52 magnitudes. Typical errors of our photometry do not
change dramatically from the ones in Silva-Villa & Larsen
(2010, see its Fig. 2).
The final color-magnitude diagrams (CMD) for the stars
in the arms and the inter-arm regions (see Sect. 5 for defi-
nition of the stars that belong to the arms and in the inter-
arm region) are presented in Fig. 2, left column. In the same
figure we investigate whether there are any indications of
differences in the mean extinction from one region to an-
other. We created histograms of the main sequence stars,
defined to be stars in the color range −0.3 ≤ (V −I) ≤ 0.7
and in the magnitude range −10 ≤ MV ≤ −5, as indicated
by the red boxes over the CMDs (see Fig. 2, right column).
The histograms were made with variable bin widths using
100 stars per bin. The number of stars is normalized to the
width of the bin. We fitted Gaussians to the observed distri-
butions to estimate the maximum of the distributions. We
do not observe a large variation among the Arm 2 and the
inter-arm areas, where the peak of the distribution is close
the same value (∼ 0.10±0.02). However, the Arm 1 presents
a slight shift in the peak of the distribution (∼0.13±0.02)
compared to Arm 2 and inter-arm regions. This difference
is not large (≤ 0.03, and errors overlap), but we note that
the Arm 1 is closer to the galactic center, where extinction
can be affecting the observations. There is also an apparent
increase in the distributions close to (V − I) ≈0.5. Inside
our photometric errors, separating the main sequence stars
from the blue He burning phases is not straightforward, but
the count of stars can give an indication of the presence of
these stars, as seen in Fig. 2, where at (V − I) ≈ 0.5 the
distribution of stars clearly deviates from the Gaussian fit.
4. Defining the spiral arms over NGC 5236
To identify the arms of NGC 5236, we followed the method
described by Scheepmaker et al. (2009), using Hα as indica-
tor. The CTIO image of the galaxy was cropped to remove
as much background as possible. Over the new (sub)image,
we use a Gaussian kernel (with a 20 pixels sigma) to blur
the image, enhancing the regions where Hα appears to be
more concentrated. Because of the high concentration of
gas in the center of the galaxy, we manually mask this part
2www.stsci.edu/hst/acs/analysis/zeropoints/#tablestart
of the image, which will allow better analysis of the re-
gions outside of the center. We analyzed the blurred image
with Daofind in IRAF to find the places where Hα is more
concentrated (minimum data value of 400 counts over the
background). The coordinates retrieved by IRAF were vi-
sually inspected to remove unnecessary detections, i.e. de-
tections that could affect the estimations of the arm’s path
and that are not part of the ACS field-of-view. We used
a cubic spline interpolation to fit the arms of NGC 5236.
Figure 3 depicts the CTIO images. The upper panel is the
cropped and blurred image, overplotted with the estimated
location of the arms. The lower panel shows the original
CTIO image, overplotted with the location of the arms (red
lines) and the two fields covered by the ACS observations
(yellow squares). Conversion of the coordinates from the
CTIO coordinate system to the ACS coordinate system was
done using wcsctran in IRAF. First, using the header of the
CTIO image, we converted the CTIO coordinate system to
WCS coordinates, and then, using the same procedure, we
moved finally to the ACS coordinate system. The different
circumferences in Fig. 3 are marking the distances 200.′′0,
140.′′0, 170.′′0 and 110.′′0 used to estimate the velocity of a
particle at different radii (see Sect. 6).
We create the density plot of the resolved stellar pop-
ulations over the ACS fields with the arm’s path cre-
ated using the Hα. After overplotting the arms, we ob-
serve that the arms follow the region where the density
of stars is higher, as seen in Fig. 4. From the same fig-
ure, it is observed that other regions have high density
of stars, e.g. in field 2 there is a possible “feather” be-
tween the two arms, and in field 1 we observe many re-
gions with high densities of stars, which is expected, due
to the proximity to the center of the galaxy, where the
concentration of gas is higher (e.g. Crosthwaite et al. 2002;
Lundgren et al. 2004). There is large observational evi-
dence related with “feathers” and/or “spurs” close (or at-
tached) to spiral arms, which is supported by theoretical
studies (see e.g. Scoville et al. 2001; Kim & Ostriker 2002;
Shetty & Ostriker 2006; La Vigne et al. 2006).
For the rest of this paper we will refer as Arm 1 to
the arm that is fully inside the corotation radius (see green
dotted line in Fig. 3), while the Arm 2 has one part inside
the corotation radius and one part outside of it. Figure 3
presents the two arms and the corotation radius, estimated
by Lundgren et al. (2004) to be located at 170”.
5. Selection of stars and star formation histories
We aim to study here how the star formation varies from
the center of the spiral arms to the inter-arm regions. We
use the resolved stellar population detected over our ACS
fields and estimate their star formation histories at a fixed
distance from the center of the arm.
5.1. The arms
Stars were selected every 0.2 Kpc, assuming that the arms
presented above mark the ”center” of the distribution.
Figure 5 shows the distribution of the stars which follow
the arms at the distances 0-0.2, 0.2-0.4, and 0.4-0.6 Kpc,
represented with the colors yellow, green, and blue, respec-
tively. In our second field the distance between the two
arms would allow to cover a larger area, however, we kept
4

Page 5

E. Silva-Villa and S. S. Larsen: ΣSFR and Arms over NGC 5236
Fig.5. Selection of the resolved stellar population over the arms of NGC 5236. Red lines show the arms of the galaxy
based on Hα images. Yellow, green, and blue represent the selection of stars at different distances that belong to the
arms. Orange regions represent the inter-arms.
the same distances as in the first field in order to have com-
parable results among the fields.
For each of the selected group of stars, star forma-
tion histories (SFH) were calculated using the synthetic
CMD method (Tosi et al. 1991). A description and tests
of the IDL-based program used to estimate the SFH is
given in Silva-Villa & Larsen (2010). For more applications,
see Silva-Villa & Larsen (2011). We have updated the code
presented by Silva-Villa & Larsen (2010) allowing the use
of multiple metallicities. When using multiple metallicities,
the code models the Hess diagram as a weighted sum of all
isochrones in the input library. The final SFH is obtained
adding together the individual SFR over time for each as-
sumed metallicity.
The parameters used for the estimation of the SFH in
NGC 5236 are: Distance modulus of 28.27, solar and LMC-
like metallicities, a binary fraction of 0.5 with a mass ra-
tio between 0.1-0.9 (assuming a flat distribution and no
binary evolution), and using the color combination V-I.
We normalized our estimations by the areas covered, hav-
ing then the surface star formation rate density (ΣSFR
[M⊙yr−1Kpc−2]). The estimation of the different areas cov-
ered was carried out following a similar procedure as for the
selection of the stars. We create an image of the size of the
ACS fields and calculate the distance of each pixel to the
arms. Having the total amount of pixels under the desired
area, the areas were calculated following the relation:
Ai= Atotal×
NpixAi
NpixAtotal
,
(1)
where Aiis the area to be calculated, Atotalis the total area
covered by the ACS (21.3 Kpc2, each field), and NpixAiand
NpixAtotalare the number of pixels for the area i and the
total number of pixels, respectively.
Figure 6 illustrates the fits done for field 1 for the three
areas covering Arm 2. We note that the fits are far from
perfect. Trying to understand what could be the cause of
the mismatch between observed and modelled CMDs, we
repeated the fits varying the input parameters. As an exam-
ple, the right column in Fig. 6 shows the fits using only solar
metallicity (instead of solar and LMC). It is clear that solar
metallicity leaves a ”gap” between the main sequence stars
and the blue loop, which is not seen in our observed CMDs
(first column). The combination of both metallicities gives a
better fit to the data, as the blue loops extend to higher ef-
fective temperatures (bluer colours) for LMC-like metallic-
ity, and we can not reject the possibility that higher metal-
licities are present. Regardless of the improvement when
using a combination of metallicities, we note that neither
red nor blue He burning phases are well fitted by our pro-
gram, appearing bluer and redder, respectively, in our final
fits, when compared with the observations. Assumptions of
no binaries did not show any improvement. Following the
small change in extinction, as suggested in Sect. 2, we used
a time dependent extinction, but we did not see any im-
provement. The middle column presents the best fit which
combines the parameters described in the previous para-
graph.
In view of the difficulties reproducing the observed
CMDs, caution should clearly be exercised when interpret-
ing SFHs inferred from these fits.
For each field and each arm, between 10 and 100 Myr
(age range less affected by incompleteness), we estimated
the mean values for ΣSFR. We also combined results for
5