Dust formation in a galaxy with primitive abundances.

Abstract

Interstellar dust plays a crucial role in the evolution of galaxies. It governs the chemistry and physics of the interstellar medium. In the local universe, dust forms primarily in the ejecta from stars, but its composition and origin in galaxies at very early times remain controversial. We report observational evidence of dust forming around a carbon star in a nearby galaxy with a low abundance of heavy elements, 25 times lower than the solar abundance. The production of dust by a carbon star in a galaxy with such primitive abundances raises the possibility that carbon stars contributed carbonaceous dust in the early universe.

Full text

Page 1

DOI: 10.1126/science.1165626
, 353 (2009);
323
Science
et al.G. C. Sloan,
Abundances
Dust Formation in a Galaxy with Primitive

www.sciencemag.org (this information is current as of January 23, 2009 ):
The following resources related to this article are available online at
http://www.sciencemag.org/cgi/content/full/323/5912/353
version of this article at:
including high-resolution figures, can be found in the online
Updated information and services,
http://www.sciencemag.org/cgi/content/full/323/5912/353#otherarticles
, 1 of which can be accessed for free:
cites 47 articles
This article
http://www.sciencemag.org/cgi/collection/astronomy
Astronomy
:
subject collections
This article appears in the following
http://www.sciencemag.org/about/permissions.dtl
in whole or in part can be found at:
this article
permission to reproduce
of this article or about obtaining
reprints
Information about obtaining
registered trademark of AAAS.
is aScience2009 by the American Association for the Advancement of Science; all rights reserved. The title
CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience
on January 23, 2009
www.sciencemag.org
Downloaded from

Page 2

Dust Formation in a Galaxy with
Primitive Abundances
G. C. Sloan,1* M. Matsuura,2,3A. A. Zijlstra,4E. Lagadec,4M. A. T. Groenewegen,5
P. R. Wood,6C. Szyszka,4J. Bernard-Salas,1J. Th. van Loon7
Interstellar dust plays a crucial role in the evolution of galaxies. It governs the chemistry and
physics of the interstellar medium. In the local universe, dust forms primarily in the ejecta from
stars, but its composition and origin in galaxies at very early times remain controversial. We report
observational evidence of dust forming around a carbon star in a nearby galaxy with a low
abundance of heavy elements, 25 times lower than the solar abundance. The production of dust by
a carbon star in a galaxy with such primitive abundances raises the possibility that carbon stars
contributed carbonaceous dust in the early universe.
D
by galaxies, and fragments collapsing molecular
cloudsintointermediate-andlow-massstars.The
detectionofdustintheearlyuniverse(1,2)raises
questions about its composition and origin. Pri-
mordialdustmaydiffersubstantiallyfromdustin
theMilkyWaybecauseitformedwhentheabun-
dancesofrefractoryelementswerelow.Polycyclic
aromatichydrocarbons(PAHs),acarbon-richdust
component, have been detected as far back as a
redshift of 2.7, or 2.5 billion years after the Big
Bang (3),butgalaxieswithlowmetallicities(that
is, with little enrichment from elements heavier
thanhelium)haveadeficitinPAHemission(4,5).
The 2200 Å bump seen in ultraviolet extinction
curves and associated with carbonaceous inter-
stellar dust is also largely absent in both metal-
poor galaxies (6) and distant quasars (7). The
extinction curves of these galaxies still require
carbon-rich dust(8),but the natureof thatcarbon-
rich dust is changing with metallicity.
The main sources of dust are the winds of
evolvedstarsandsupernovae(SNe).IntheMilky
Way,stars on theasymptotic giant branch (AGB)
dominate the injection of dust into the interstellar
medium (ISM) (9). The chemistry of this dust
depends on the C/O ratio in the photosphere of
the star, which in turn depends on how much
carbon produced by triple-a burning in the inte-
rior of the star has been dredged to the surface.
ust is an important astrophysical constit-
uent. It regulates the cooling of the inter-
stellar medium, attenuates light emitted
The stable CO molecule will form until either C
or O is exhausted. Solids form from the remain-
der, resulting in either oxygen-rich dust domi-
nated by silicates or carbonaceous dust. Both
observations (10, 11) and theory (12) show that
the fraction of AGB stars that become carbon-
rich increases in more metal-poor systems.
There are reasons to expect reduced dust pro-
duction at low metallicity. Radiation pressure on
dust grains cannot drive the winds from oxygen-
rich AGB stars at metallicities below 0.1 of the
solar value, which could lead to higher SN rates
inprimitivegalaxies(13)andmakeSNethedom-
inant source of dust in the early universe (14).
However,observationalevidencethatSNeproduce
sufficient dust, and produce more than they de-
stroy, remains controversial (15, 16).
Studies with the Infrared Spectrograph (IRS)
(17) on the Spitzer Space Telescope (18) of the
Large Magellanic Cloud (LMC) and Small Mag-
ellanic Cloud (SMC) (19–21) and the Fornax
DwarfSpheroidal(22)haverevealeddust-forming
carbon stars at metallicities as low as ~0.1 that of
the Sun. As the metallicity of the sample de-
creases, the amount of dust produced by oxygen-
rich AGB stars decreases; but for carbon stars,
the amount of dust remains unchanged (23, 24).
Other properties of the outflows from carbon
stars do depend on metallicity. As the metallicity
of the sample drops, carbon stars produce less
trace dust components such as SiC and MgS,
eventhoughtheproductionofamorphouscarbon
doesnotchange.Thereisevidenceforincreasing
abundances of acetylene (C2H2) at lower metal-
licity.Acetyleneisthefundamentalbuildingblock
of larger carbon-rich structures such as PAHs (25).
Thus, mass loss and dust formation in carbon stars
do not appear to depend on the abundances with
which the star formed. Instead, free carbon (after
the formation of CO) neededby these starstoform
dust is produced by the stars themselves and then
dredgedtothesurface.Itfollowsthateventhemost
metal-poor carbon stars should produce dust.
To test this hypothesis, we observed a carbon
star in the direction of the Sculptor Dwarf Sphe-
roidalgalaxywiththeIRSonSpitzer.TheSculptor
Dwarf is a satellite of the Milky Way, with a met-
allicity only 0.04 that of the Sun (26). A study of
carbon stars in the Galactic Halo detected a can-
didate, MAG 29, in the field of the Sculptor
Dwarf (27). This study assumed that the absolute
magnitude in the narrowK filter (Ks)of MAG 29
was −6.9, but noted that it could be brighter
by half a magnitude or more. The apparent Ks
magnitude of MAG 29 is 11.60 (28). Thus, the
assumed absolute Ksmagnitude (MK) implies a
distance of 50 kpc, in the foreground of the
REPORTS
1Department of Astronomy, Cornell University, Ithaca, NY
14853-6801, USA.2National Optical Astronomical Observa-
tory of Japan, Osawa 2-21-1, Mitaka, Tokyo 181-8588, Japan.
3Department of Physics and Astronomy, University College
London, Gower Street, London WC1E 6BT, UK.4School of
Physics and Astronomy, University of Manchester, Post Office
Box 88, Manchester M6O 1QD, UK.5Royal Observatory of
Belgium, Ringlaan 3, B-1180 Brussels, Belgium.6Research
School of Astronomy and Astrophysics, Australian National
University, Cotter Road, Weston Creek Australian Capital
Territory 2611, Australia.7Astrophysics Group, Lennard-Jones
Laboratories, Keele University, Staffordshire ST5 5BG, UK.
*To whom correspondence should be addressed. E-mail:
sloan@isc.astro.cornell.edu
Fig. 1. The IRS spec-
trumofMAG29.Fn,flux
density per unit of fre-
quency; l, wavelength.
The three points at 1.2
to2.2mmrepresentolder,
ground-based 2MASS ob-
servations (28). (A) The
Manchester method de-
terminesthe[6.4]−[9.3]
and[16.5]−[21.5]colors
(boxes),fitslinesegments
over the C2H2bands at
7.5 and 13.7 mm and
under the SiC emission
feature at ~11 mm (solid
lines), and extrapolates a
Planck function from the
[16.5] − [21.5] color un-
der any possible MgS
emission in the 30-mm
vicinity (dashed line). (B)
Aradiativetransfermodel
of the dust continuum
(solid curve).
A
B
www.sciencemag.org
SCIENCE
VOL 32316 JANUARY 2009
353
on January 23, 2009
www.sciencemag.org
Downloaded from

Page 3

Sculptor Dwarf, which is 87 kpc away (29).
However, calibrations of near- and mid-infrared
color magnitude relations (23) make MAG 29 a
likely member of the Sculptor Dwarf. The J − Ks
color (difference between apparent J and Ks
magnitudes) of MAG 29 is 3.24 (28). The linear
relation between J − Ksand MKimplies MK=
−7.90T 0.51 and a distance of 80 T 19 kpc. Mea-
suring the magnitude of MAG 29 at 6.4 and
9.3 mm ([9.3]) and applying the relation between
[6.4] − [9.3] color and M9.3provides a second
estimate of thedistance.The[6.4]− [9.3]color is
0.494 T 0.012, which gives an absolute magni-
tude at 9.3 mm (M9.3) = −11.38 T 0.59. The ap-
parentmagnitudeat9.3mmis8.35,implyingthat
the distance is 88 T 18 kpc. The two color-based
distances are consistent with each other and the dis-
tance to the Sculptor Dwarf, but both are incon-
sistent with the previous provisional estimate of
50 kpc. Combining the two new distances gives
84 T 13 kpc, confirming the association between
the star and the galaxy and making MAG 29 the
most metal-poor, carbon-rich AGB star studied
spectroscopically in the infrared.
Themid-infraredspectrumofMAG29shows
a continuum, with strong molecular absorption
bands (Fig. 1A). We analyzed the spectrum by
applyingtheManchestermethod(19),whichuses
the [6.4] − [9.3] color to diagnose the continuum
temperatureinthisspectralregion.Radiativetrans-
fer models show that this color varies linearly
withthelogarithmof the mass-loss rate (24).The
calibrationof[6.4]−[9.3]colortodustmass-loss
rate (23), which assumes an outflow velocity of
10kms−1,impliesthatMAG29isproducingdust
at the rate of 8 × 10−9solar masses (M⊙) year−1,
with an uncertainty of ~16%.
Wetestedthisconclusionbyfittingaradiative
transfer model (30) to the IRS spectrum and to
the 2MASS data of Skrutskie etal.(28) (Fig.1B).
The2MASSdatawerebestfittedwithamodelin
which the stellar temperature was4000 K,higher
than found for Galactic mass-losing stars. Metal-
poor AGB stars are bluer and hotter than their
metal-richcounterparts(31).Thedustshellaround
MAG 29 has a temperature of 1600 K at its inner
radius.Althoughbelowthecondensationtemper-
ature of amorphous carbon, it is higher than what
is typically found for Galactic carbon stars, in-
dicating that the dust in MAG 29 forms close to
the star and that graphite may be a component.
The fittedmodelimpliesthatMAG29isproduc-
ing dust at a rate of 2.5 × 10−8M⊙year−1. This
value is larger than that estimated with the Man-
chester method by a factor of ~3 because the
radiative transfer model leads to a higher outflow
velocity (35 versus 10 km s−1). Assuming a gas-
to-dust ratio of 200, the total mass-loss rate of
MAG 29 is in the range of 1.6 × 10−6M⊙year−1
to 4.9 × 10−6M⊙year−1. These results point to
substantial mass loss and dust production.
Inmostcarbonstars,amorphouscarbondom-
inates the dust, but trace amounts of SiC and
MgS produce emission features at ~11.3 and 26
to 30 mm, respectively. We used the Manchester
method to fit a line segment to the continuum on
either side of the SiCfeature.We found a possible
emission feature centered at 10.9 mm, similar to
that found in two carbon stars in the SMC (19),
with a strength of 3 T 1% that of the continuum.
However, it is possible that what appears to be
emissionissimplycontinuumbetweenasuspected
absorption band at 10 mm (20) and acetylene ab-
sorption between 11 and 16 mm. Therefore, the
measured SiC strength is an upper limit of 4%.
TheMgSemissionextendsbeyond37mm,outside
the wavelength coverage of the IRS. We extrapo-
lated the continuum under the MgS emission
feature by fitting a Planck function at 16.5 and
21.5mmandfoundnoevidenceofMgSemission.
The spectrum of MAG 29 shows a deep ab-
sorption band at 7.5 mm. It is substantially deeper
and broader than the acetylene bands in carbon
stars in the LMC (32). Thus, we modified the
Manchestermethodbyfittingacontinuumat6.14
to 6.44 mm and 9.12 to 9.42 mm. The band's
equivalent width is 0.79 T 0.02 mm, almost twice
as strong as that measured in any other carbon
star.Wecomparedittothatofatypicalcarbonstar
in the SMC and to a synthetic spectrum of acety-
lene with an excitation temperature of 1750 K
(Fig. 2). The synthetic spectrum reproduces the
general shape of the band, demonstrating that
acetylene is the primary absorber. The synthetic
spectrum does not match the details of the ab-
sorption core, but it is limited by the available
line lists, which do not include the higher-order
transitions that will be populated at these high
temperatures and column densities (33).
MAG 29 shows another acetylene feature
centered at 13.7 mm. The Manchester method is
not useful to study this feature because it mea-
surestheQbranchofthistransition,butnottheP
andRbranchesthatcanextendfrom~11to~16mm
(33).Matchesbetweenthedataandamodelofthe
continuum over this wavelength range show that
the band's equivalent width is 0.990 T 0.022 mm.
Onlyonecarbonstar,MSXSMC093intheSMC,
has a stronger band (34), whereas IRAS 04496-
6958intheLMC,anothercarbonstar,hasaband
of comparable strength (Fig. 3)
We compared the acetylene strength and the
SiC strength of MAG 29 with those observed
in Local Group samples (Fig. 4). We limited the
comparison sample to those sources with colors
similar to those of MAG 29; that is, those with
0.35 ≤ [6.4] − [9.3] ≤ 0.65. The data for the
7.5-mm feature and SiC strength are from previ-
ouslypublishedstudiesoftheMagellanicClouds
(19–21, 23) and new analysis of spectra from the
Fornax Dwarf (22). The 11- to 16-mm data are
takendirectlyfromarecentstudy(34).Weadopted
[Z/H]=0.00T 0.10 for the Galaxy (Z is the metal
Fig. 2. The 7.5-mm ab-
sorption band in the
spectrum of MAG 29,
compared to the spec-
trumofMSXSMC036,a
typical carbon star in
the SMC (19) (top) and
toasyntheticabsorption
spectrumfromacetylene
(C2H2) (bottom).
Fig. 3. The11-to16-mm
absorption band in the
spectrum of MAG 29,
plotted as transmission
by dividing by the con-
tinuumasestimatedfrom
theradiativetransfermod-
elinFig.1andcompared
to the IRS spectrum of
IRAS 04496−6958 (33)
(top)andasyntheticacet-
ylenespectrum(bottom).
16 JANUARY 2009VOL 323
SCIENCE
www.sciencemag.org
354
REPORTS
on January 23, 2009
www.sciencemag.org
Downloaded from

Page 4

abundance), −0.45 T 0.15 for the LMC, and
−0.85 T 0.15 for the SMC (35). We adopted
[Z/H] = −1.33 T 0.20 for the Sculptor Dwarf (26)
and [Z/H] = 0.93 T 0.30 for Fornax [based on
[Fe/H] = −1.0 T 0.3 (36) and assuming the same
[a/Fe]asfortheSculptorDwarf].Theweaktrend
of increasing acetylene absorption with decreas-
ing metallicity is more readily apparent when the
data are examined over a larger color range (23).
The spectrum of MAG 29 shows that the pre-
viously observed trends extend to metallicities as
low as that of the Sculptor Dwarf.
IntheGalacticenvironment,radiationpressure
on carbon-rich dust can drive high mass loss from
carbon stars (37). A carbon star self-produces its
carbon via the triple-a reaction;ifitcandredge up
enough excess carbon, compared to oxygen, to
allow the formation of carbonaceous dust, this
high–mass-loss phase becomes inevitable (38).
The spectrum of MAG 29 shows that metallicity
is not a hurdle for the formation of dust around
carbon stars. Metal-poor stars have an increased
carbondredge-upefficiencyandlacktheintrinsic
oxygenthatoffsetsdredged-upcarbon,leadingto
a higher amount of excess carbon after the for-
mation of CO (39, 40).
Once enough time has elapsed after a galaxy
hasformedforstarstoreachtheAGBanddredge
upcarbon,itsISMshouldbeseededwithcarbon-
rich dust. Dust has been detected in galaxies to a
redshift of 6.4 (1), or only 870 million years after
the Big Bang. Recent modeling (41) shows that
as soon as trace amounts of metals (10−5of the
solar amount) appear in the ISM, intermediate-
mass stars can form. Thus, precursors of carbon
stars will form almost immediately after the for-
mation of galaxies, which should have occurred
by a redshift of ~10, or ~480 million years after
the Big Bang (42). Thus, carbon stars have ~390
millionyearstoevolvefromtheirbirthtotheAGB.
Models of stars at 0.01 solar metallicity (43) be-
come carbon-rich over the full mass range studied
(2.5 to 5 M⊙). Similarly, all models from 2 to
6 M⊙at half that metallicity also become carbon-
rich(44).Thesemodelswillevolvefromthezero-
age main sequence to the AGB in only 70 million
years for the most massive stars. The 3-M⊙mod-
els require only 280 to 310 million years. These
timescalesarelessthanthe500-million-yeartime
scale used previously to rule out carbon stars as
contributors to the dust at these redshifts (14, 45).
How much dust carbon stars can contribute at
these redshifts depends on variables that are more
difficulttoquantify.Theamountofcarbonformed
and dredged up differs by an order of magnitude
in the models referred to above. The star forma-
tion rate and the initial mass function will deter-
minehowmanystarsformintheappropriatemass
range. The star formation rates needed for carbon
stars to explain the dust observed at high redshift
may be consistent with observed luminosities
(46). All of these quantities need to be better un-
derstood before we can quantify how much dust
carbon stars contributed in the early universe.
The relative contributions of SNe and carbon
stars to dust in the early universe is a problem of
great interest. Core-collapse SNe appear well be-
fore the first carbon stars, whereas type Ia SNe
require white dwarfs, which appear only after
AGB stars have evolved. Core-collapse SNe can
produce both silicates and carbonaceous dust; it
is difficult to determine which would dominate
(47). Current measurements of the dust around
observed SNe fall short of what is needed to
account for the dust observed at high redshifts
(45), but we lack direct observations of SNe at
low metallicity. A recent study of dust extinction
at a redshift of 6.2 found evidence for carbona-
ceous dust (48). In the light of the observations
reported here, the presence of such dust could be
explained, at least in part, by the mass loss from
carbon stars in the early universe.
References and Notes
1. F. Bertoldi et al., Astron. Astrophys. 406, L55 (2003).
2. D. C. Hines et al., Astrophys. J. 641, L85 (2006).
3. H. I. Teplitz et al., Astrophys. J. 659, 941 (2007).
4. C. W. Engelbracht et al., Astrophys. J. 628, L29 (2005).
5. Y. Wu et al., Astrophys. J. 639, 157 (2006).
6. K. D. Gordon, G. C. Clayton, K. A. Misselt, A. U. Landolt,
M. J. Wolff, Astrophys. J. 594, 279 (2003).
7. K. M. Pitman, G. C. Clayton, K. D. Gordon, Publ. Astron.
Soc. Pac. 112, 537 (2000).
8. J. S. Mathis, Astrophys. J. 497, 824 (1998).
9. R. Gehrz, in Proceedings of IAU Symposium 135,
L. J. Allamandola, A. G. G. M. Tielens, Eds. (Kluwer
Academic, Dordrecht, Netherlands, 1989), pp. 445–453.
10. B. M. Blanco, V. M. Blanco, M. F. McCarthy, Nature 271,
638 (1978).
11. V. M. Blanco, B. M. Blanco, M. F. McCarthy, Astrophys. J.
242, 938 (1980).
12. A. Renzini, M. Voli, Astron. Astrophys. 94, 175 (1981).
13. G. H. Bowen, L. A. Willson, Astrophys. J. 375, L53 (1991).
14. E. Dwek, Astrophys. J. 501, 643 (1998).
15. B. E. K. Sugerman et al., Science 313, 196 (2006).
16. W. P. S. Meikle et al., Astrophys. J. 508, 106 (2007).
17. J. R. Houck et al., Astrophys. J. Suppl. 154, 18 (2004).
18. M. W. Werner et al., Astrophys. J. Suppl. 154, 1 (2004).
19. G. C. Sloan et al., Astrophys. J. 645, 1118 (2006).
20. A. A. Zijlstra et al., Mon. Not. R. Astron. Soc. 370, 1961
(2006).
21. E.Lagadecetal.,Mon.Not.R.Astron.Soc.376,1270(2007).
22. M. Matsuura et al., Mon. Not. R. Astron. Soc. 382, 1889
(2007).
23. G. C. Sloan et al., Astrophys. J. 686, 1056 (2008).
24. M. A. T. Groenewegen et al., Mon. Not. R. Astron. Soc.
376, 313 (2007).
25. M. Frenklach, E. D. Feigelson, Astrophys. J. 341, 372 (1989).
26. C. Babusiaux, G. Gilmore, M. Irwin, Mon. Not. R. Astron.
Soc. 359, 985 (2005).
27. N. Mauron, M. Azzopardi, K. Gigoyan, T. R. Kendall,
Astron. Astrophys. 418, 77 (2004).
28. M. F. Skrutskie et al., Astron. J. 131, 1163 (2006).
29. J. Kaluzny et al., Astron. Astrophys. 112, 407 (1995).
30. M. Elitzur, Ž. Ivezić, Mon. Not. R. Astron. Soc. 327, 403
(2001).
31. E. M. Levesque et al., Astrophys. J. 628, 973 (2005).
32. M. Matsuura et al., Mon. Not. R. Astron. Soc. 371, 415
(2006).
33. A. K. Speck et al., Astrophys. J. 650, 892 (2006).
34. J. M. Leisenring, F. Kemper, G. C. Sloan, Astrophys. J.
681, 1557 (2008).
35. J. Th. van Loon, Astron. Astrophys. 354, 125 (2000).
36. E. Tolstoy et al., Mon. Not. R. Astron. Soc. 327, 918 (2001).
37. P. Woitke, Astron. Astrophys. 460, L9 (2006).
38. E. Lagadec, A. A. Zijlstra, Mon. Not. R. Astron. Soc. 390,
L59 (2008).
39. J. Th. van Loon, A. A. Zijlstra, M. A. T. Groenewegen,
Astron. Astrophys. 346, 805 (1999).
40. M. Matsuura et al., Astron. Astrophys. 434, 691 (2005).
41. P. C. Clark, S. C. O. Glover, R. S. Klessen, Astrophys. J.
672, 757 (2008).
42. T. H. Greif, J. L. Johnson, R. S. Klessen, V. Bromm,
Mon. Not. R. Astron. Soc. 387, 1021 (2008).
43. P. Ventura, F. D'Antona, I. Mazzitelli, Astron. Astrophys.
393, 215 (2002).
44. F. Herwig, Astrophys. J. Suppl. 155, 651 (2004).
45. E. Dwek, F. Galliano, A. P. Jones, Astrophys. J. 662, 927
(2007).
46. H. L. Morgan, M. G. Edmunds, Mon. Not. R. Astron. Soc.
343, 427 (2003).
47. R. Schneider, A. Ferrara, R. Salvaterra, Mon. Not. R.
Astron. Soc. 351, 1379 (2004).
48. R. Maiolino et al., Nature 431, 533 (2004).
49. Support for G.C.S. was provided by NASA through
contract number 1257184 issued by the Jet Propulsion
Laboratory, California Institute of Technology, under
NASA contract 1407. The Australian Research Council
provided support to P.R.W., and M.M. was supported by a
Japan Society for the Promotion of Science fellowship.
8 September 2008; accepted 2 December 2008
10.1126/science.1165626
Fig. 4. The metallicity
dependenceof theacet-
ylene equivalent widths
at 7.5 mm (top) and 11
to 16 mm (middle) and
the SiC/continuum flux
ratio(bottom)forsources
with0.35≤[6.4]−[9.3]≤
0.65. The error bars are
thestandarddeviationof
thepopulations,exceptfor
MAG 29. We have con-
sideredtheSiCstrengthin
MAG 29 to be 0 T 4%of
thecontinuum,although
it could be as large as
3 T 1%.
www.sciencemag.org
SCIENCE
VOL 32316 JANUARY 2009
355
REPORTS
on January 23, 2009
www.sciencemag.org
Downloaded from