1. The Large, Oxygen-Rich Halos of Star-Forming Galaxies Are a Major
Reservoir of Galactic Metals
J. Tumlinson, et al.
Science 334, 948 (2011);
DOI: 10.1126/science.1209840
This copy is for your personal, non-commercial use only.
If you wish to distribute this article to others, you can order high-quality copies for your
colleagues, clients, or customers by clicking here.
Permission to republish or repurpose articles or portions of articles can be obtained by
following the guidelines here.
The following resources related to this article are available online at
Downloaded from www.sciencemag.org on November 27, 2011
www.sciencemag.org (this infomation is current as of November 27, 2011 ):
Updated information and services, including high-resolution figures, can be found in the online
version of this article at:
http://www.sciencemag.org/content/334/6058/948.full.html
Supporting Online Material can be found at:
http://www.sciencemag.org/content/suppl/2011/11/16/334.6058.948.DC1.html
A list of selected additional articles on the Science Web sites related to this article can be
found at:
http://www.sciencemag.org/content/334/6058/948.full.html#related
This article cites 32 articles, 2 of which can be accessed free:
http://www.sciencemag.org/content/334/6058/948.full.html#ref-list-1
This article has been cited by 1 articles hosted by HighWire Press; see:
http://www.sciencemag.org/content/334/6058/948.full.html#related-urls
This article appears in the following subject collections:
Astronomy
http://www.sciencemag.org/cgi/collection/astronomy
Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the
American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright
2011 by the American Association for the Advancement of Science; all rights reserved. The title Science is a
registered trademark of AAAS.

2. ribosomal protein elements have coevolved to 17. D. Benelli et al., Nucleic Acids Res. 37, 256 (2009). 46. D. Tu, G. Blaha, P. B. Moore, T. A. Steitz, Cell 121, 257
form architectural complexes that mediate long- 18. M. Gartmann et al., J. Biol. Chem. 285, 14848 (2010). (2005).
19. T. F. Menne et al., Nat. Genet. 39, 486 (2007). 47. R. Beckmann et al., Cell 107, 361 (2001).
distance tertiary interactions. The structure also 20. K. Y. Lo et al., Mol. Cell 39, 196 (2010). 48. T. Lacombe et al., Mol. Microbiol. 72, 69 (2009).
offers insights into the regulatory mechanisms 21. See supporting material on Science Online. 49. A. W. Johnson, S. R. Ellis, Genes Dev. 25, 898 (2011).
of the eukaryotic ribosome, the maturation of the 22. A. Nakao, M. Yoshihama, N. Kenmochi, Nucleic Acids Res. 50. A. J. Finch et al., Genes Dev. 25, 917 (2011).
large ribosomal subunit, principles of antibiotic 32 (database issue), D168 (2004). 51. B. Hartzoulakis et al., Bioorg. Med. Chem. Lett. 17,
23. B. S. Strunk, K. Karbstein, RNA 15, 2083 (2009). 3953 (2007).
specificity and of the mechanism of translation Acknowledgments: All data were collected at the Swiss
24. H. T. Gazda et al., Am. J. Hum. Genet. 83, 769 (2008).
inhibition by cycloheximide, and the structural ba- 25. J. B. Moore 4th, J. E. Farrar, R. J. Arceci, J. M. Liu, Light Source (SLS, Paul Scherrer Institut, Villigen). We
sis of extraribosomal roles of various eukaryotic- S. R. Ellis, Haematologica 95, 57 (2010). thank T. Tomizaki, M. Müller, V. Olieric, G. Pompidor,
specific ribosomal proteins. As such, it provides 26. J. Zhang et al., Genes Dev. 21, 2580 (2007). and A. Pauluhn for their outstanding support at the
27. J. A. Chao, J. R. Williamson, Structure 12, 1165 (2004). SLS; J. Rabl for advice on cell growth and ribosome
a starting point for future biochemical, genet- purification as well as a clone of eIF6; T. Maier for
ic, and structural studies of protein synthesis in 28. S. Macías, M. Bragulat, D. F. Tardiff, J. Vilardell, Mol. Cell
30, 732 (2008). advice on data collection; T. Bucher for the preparation
eukaryotes. 29. M. Halic, T. Becker, J. Frank, C. M. Spahn, R. Beckmann, of crystals; J. Erzberger and T. Bucher for critically
Nat. Struct. Mol. Biol. 12, 467 (2005). reading the manuscript; and all members of the Ban
References and Notes laboratory for suggestions and discussions. Supported
30. L. Li, K. Ye, Nature 443, 302 (2006).
1. T. M. Schmeing, V. Ramakrishnan, Nature 461, 1234 (2009). by the Swiss National Science Foundation (SNSF), the
31. N. Kondrashov et al., Cell 145, 383 (2011).
2. G. Kramer, D. Boehringer, N. Ban, B. Bukau, Nat. Struct. National Center of Excellence in Research (NCCR)
32. J. L. Houmani, C. I. Davis, I. K. Ruf, J. Virol. 83, 9844
Mol. Biol. 16, 589 (2009). Structural Biology program of the SNSF, and European
(2009).
Research Council grant 250071 under the European
Downloaded from www.sciencemag.org on November 27, 2011
3. D. Bulkley, C. A. Innis, G. Blaha, T. A. Steitz, Proc. Natl. 33. M. L. DeLabre, J. Kessl, S. Karamanou, B. L. Trumpower,
Acad. Sci. U.S.A. 107, 17158 (2010). Community’s Seventh Framework Programme (N.B.)
Biochim. Biophys. Acta 1574, 255 (2002).
4. B. T. Wimberly et al., Nature 407, 327 (2000). and by EMBO and Human Frontier Science Program
34. C. B. Kirn-Safran et al., Dev. Dyn. 236, 447 (2007).
5. N. Ban, P. Nissen, J. Hansen, P. B. Moore, T. A. Steitz, fellowships (S.K.). Coordinates and structure factors have
35. T. Schneider-Poetsch et al., Nat. Chem. Biol. 6, 209 been deposited in the Protein Data Bank (accession
Science 289, 905 (2000). (2010).
6. M. Selmer et al., Science 313, 1935 (2006). codes for molecule 1: 4A1E and 4A18; molecule 2, 4A17
36. H. M. Fried, J. R. Warner, Nucleic Acids Res. 10, 3133 (1982). and 4A19; molecule 3, 4A1A and 4A1B; molecule 4,
7. T. M. Schmeing et al., Science 326, 688 (2009). 37. T. V. Pestova, C. U. Hellen, Genes Dev. 17, 181 (2003).
8. S. Petry et al., Cell 123, 1255 (2005). 4A1C and 4A1D). ETH Zürich has filed a patent
38. T. M. Schmeing, P. B. Moore, T. A. Steitz, RNA 9, 1345 application to use the crystals and the coordinates
9. R. Bingel-Erlenmeyer et al., Nature 452, 108 (2008). (2003).
10. V. G. Panse, A. W. Johnson, Trends Biochem. Sci. 35, of the 60S ribosomal subunit for developing compounds
39. D. R. Stevens, A. Atteia, L. G. Franzén, S. Purton, Mol. that can interfere with eukaryotic translation.
260 (2010). Gen. Genet. 264, 790 (2001).
11. J. P. Armache et al., Proc. Natl. Acad. Sci. U.S.A. 107, 40. N. F. Käufer, H. M. Fried, W. F. Schwindinger, M. Jasin,
19748 (2010). Supporting Online Material
J. R. Warner, Nucleic Acids Res. 11, 3123 (1983).
12. J. P. Armache et al., Proc. Natl. Acad. Sci. U.S.A. 107, www.sciencemag.org/cgi/content/full/science.1211204/DC1
41. G. Gürel, G. Blaha, T. A. Steitz, P. B. Moore, Antimicrob.
19754 (2010). Materials and Methods
Agents Chemother. 53, 5010 (2009).
13. A. Ben-Shem, L. Jenner, G. Yusupova, M. Yusupov, SOM Text
42. S. J. Schroeder, G. Blaha, P. B. Moore, Antimicrob. Agents
Science 330, 1203 (2010). Figs. S1 to S21
Chemother. 51, 4462 (2007).
14. J. Rabl, M. Leibundgut, S. F. Ataide, A. Haag, N. Ban, Tables S1 and S2
43. A. Yonath, Annu. Rev. Biochem. 74, 649 (2005).
Science 331, 730 (2011). References (52–72)
44. S. Zaman, M. Fitzpatrick, L. Lindahl, J. Zengel, Mol.
15. M. Ceci et al., Nature 426, 579 (2003). Microbiol. 66, 1039 (2007). 14 July 2011; accepted 5 October 2011
16. C. M. Groft, R. Beckmann, A. Sali, S. K. Burley, Nat. 45. M. G. Lawrence, L. Lindahl, J. M. Zengel, J. Bacteriol. Published online 3 November 2011;
Struct. Biol. 7, 1156 (2000). 190, 5862 (2008). 10.1126/science.1211204
REPORTS
The Large, Oxygen-Rich Halos of medium (CGM)—loosely defined as gas surround-
ing galaxies within their own halos of dark mat-
ter (out to 100 to 300 kpc)—lies at the nexus of
Star-Forming Galaxies Are a Major accretion and outflow, but the structure of the
CGM and its relation to galaxy properties are
Reservoir of Galactic Metals still uncertain. Galactic outflows are observed
at both low (2–4) and high (5–7) redshift, but it
J. Tumlinson,1* C. Thom,1 J. K. Werk,2 J. X. Prochaska,2 T. M. Tripp,3 D. H. Weinberg,4 is unclear how far they propagate, what level
M. S. Peeples,5 J. M. O’Meara,6 B. D. Oppenheimer,7 J. D. Meiring,3 N. S. Katz,3 R. Davé,8 of heavy-element enrichment they possess, and
A. B. Ford,8 K. R. Sembach1 whether the gas escapes the halo or eventually
returns to fuel later star formation. Models of
The circumgalactic medium (CGM) is fed by galaxy outflows and accretion of intergalactic gas,
but its mass, heavy element enrichment, and relation to galaxy properties are poorly constrained 1
by observations. In a survey of the outskirts of 42 galaxies with the Cosmic Origins Spectrograph Space Telescope Science Institute, Baltimore, MD 21218, USA.
2
University of California Observatories–Lick Observatory, Santa
onboard the Hubble Space Telescope, we detected ubiquitous, large (150-kiloparsec) halos of Cruz, CA 95064, USA. 3Department of Astronomy, University of
ionized oxygen surrounding star-forming galaxies; we found much less ionized oxygen around Massachusetts, Amherst, MA 01003, USA. 4Department of
galaxies with little or no star formation. This ionized CGM contains a substantial mass of heavy Astronomy, Ohio State University, Columbus, OH 43210, USA.
5
elements and gas, perhaps far exceeding the reservoirs of gas in the galaxies themselves. Our data Department of Physics and Astronomy, University of Cali-
fornia, Los Angeles, CA 90095, USA. 6Department of Chemistry
indicate that it is a basic component of nearly all star-forming galaxies that is removed or and Physics, Saint Michael’s College, Colchester, VT 05439,
transformed during the quenching of star formation and the transition to passive evolution. USA. 7Leiden Observatory, Leiden University, NL-2300 RA
Leiden, Netherlands. 8Steward Observatory, University of
alaxies grow by accreting gas from the sions release gas enriched with heavy elements
G
Arizona, Tucson, AZ 85721, USA.
intergalactic medium (IGM) and convert- [or metals (1)], some of which is ejected in *To whom correspondence should be addressed. E-mail:
ing it to stars. Stellar winds and explo- galactic-scale outflows (2). The circumgalactic tumlinson@stsci.edu
948 18 NOVEMBER 2011 VOL 334 SCIENCE www.sciencemag.org

3. REPORTS
galaxy evolution require efficient outflows to the Hubble Space Telescope to directly map the Sun. The QSO sightlines probe projected radial
explain observed galaxy masses and chemical CGM by absorption-line spectroscopy, in which distances to the galaxies (i.e., impact parameters)
abundances and to account for metals observed a diffuse gas is detected by its absorption of of R = 14 to 155 kpc. We used the COS data to
in the more diffuse IGM (8, 9). The CGM may light from a background source. Our background measure the O VI column densities (NOVI in cm−2),
also reflect the theoretically predicted transition sources are ultraviolet-bright quasi-stellar objects line profiles, and velocities with respect to the
from filamentary streams of cold gas that feed (QSOs), which are the luminous active nuclei of target galaxies (Fig. 1) (21). We measured the
low-mass galaxies to hot, quasi-static envelopes galaxies lying far behind the galaxies of interest. precise redshift, star formation rate (SFR in M◉
that surround high-mass galaxies (10, 11). Both We focus on the ultraviolet 1032, 1038 Å doublet of year−1), and metallicity for each of our sample
outflow and accretion through the CGM may be O VI (O+5), the most accessible tracer of hot and/or galaxies by means of low-resolution spectrosco-
intimately connected to the observed dichotomy highly ionized gas at redshift z < 0.5. O VI has py from the Keck Observatory Low-Resolution
between blue, star-forming, disk-dominated gal- been used to trace missing baryons in the IGM Imaging Spectrograph (LRIS) and the Las Campanas
axies and red, passively evolving, elliptical galaxies (13–16), the association of metals with galaxies Observatory Magellan Echellette (MagE) spec-
with little or no star formation (12). However, the (17–19), and coronal gas in the Milky Way halo (20). trograph (21, 22).
low density of the CGM makes it extremely dif- The high sensitivity of COS enables a QSO Our systematic sampling of galaxy properties
ficult to probe directly; thus, models of its structure absorption-line survey of halos around galaxies allows us to investigate the connection between
and influences are typically constrained indirect- with a predetermined set of properties. We have galaxies themselves and the CGM. The O VI de-
ly by its effects on the visible portions of galaxies, selected 42 sample galaxies (tables S1 and S2) tections extend to R = 150 kpc away from the
not usually by observations of the gas itself. that span redshifts zgal = 0.10 to 0.36 and stellar targeted galaxies, but the whole sample shows no
Downloaded from www.sciencemag.org on November 27, 2011
We have undertaken a large program with the masses [log(M*/M◉)] = 9.5 to 11.5, where M* is obvious trend with radius R (Fig. 2). The strong
new Cosmic Origins Spectrograph (COS) aboard the galaxy stellar mass and M◉ is the mass of the clustering of detections within T200 km s−1 of the
3
Observed Wavelength (Å)
Fig. 1. An illustration of our sampling technique and data. (A) An SDSS composite magnitude of 18.1. (C and D) The redshifted O VI 1032, 1038 Å doublet for
image of the field around the QSO J1016+4706 with two targeted galaxies, galaxies G1 (C) and G2 (D). (E and F) The full sample showing the locations of
labeled G1 and G2, which are both in the star-forming subsample. (B) The all sightlines in position angle and impact parameter R with respect to the
complete COS count-rate spectrum (counts s−1) versus observed wavelength. targeted galaxies, for the star-forming (E) and passively evolving (F)
This QSO lies at redshift zQSO = 0.822 and has an observed far-ultraviolet subsamples. The circles mark R = 50, 100, and 150 kpc.
www.sciencemag.org SCIENCE VOL 334 18 NOVEMBER 2011 949

4. REPORTS
A B
Downloaded from www.sciencemag.org on November 27, 2011
Impact parameter [kpc] log (Mhalo /M )
Fig. 2. O VI association with galaxies. (A) O VI column density, NOVI, versus R respect to galaxy systemic redshift for O VI detections, versus inferred dark-
for the star-forming (blue) and passive (red) subsamples. Solid and open matter halo mass. The range bars mark the full range of O VI absorption for
symbols mark O VI detections and 3s upper limits, respectively. The each system. The inset shows a histogram of the component velocities. The
detections in the star-forming galaxies maintain log NOVI ≈ 14.5 to R ≈ dashed lines mark the mass-dependent escape velocity at R = 50, 100, and
150 kpc, the outer limit of our survey. (B) Component centroid velocities with 150 kpc from outside to inside.
Fig. 3. O VI correlation A B
with galaxy properties. -9
(A) O VI column density
versus sSFR (≡ M*/SFR).
Star-forming galaxies
log (sSFR [yr -1])
log (NOVI [cm-2])
-10
are divided from passive-
ly evolving galaxies by
sSFR ≈ 10−11 year−1; our
-11
detection limit is sSFR ≈
5 × 10−12 year−1. (B) The
galaxy color-magnitude
-12
diagram (sSFR versus M*)
for SDSS+GALEX galaxies Star-forming galaxies
from (23). Passive galaxies
-13
9 10 11
sSFR [yr -1] log (M /M )
galaxy systemic velocities indicates a close phys- and passive subsamples overlap, rejects at >99% the hit rate correction fhit computed separately
ical and/or gravitational association. confidence the null hypothesis that they draw in three 50-kpc annuli (Figs. 1 and 2). This mass
CGM gas as traced by O VI reflects the un- from the same parent distribution of NOVI (fig. of oxygen is strictly a lower limit because we
derlying bimodality of the general galaxy popu- S2). We therefore conclude that the basic dichot- have scaled to the maximum fOVI = 0.2 (Fig. 4).
lation (12, 23). We found a correlation of NOVI omy between star-forming (“blue-cloud”) and The corresponding total mass of circumgalactic
with specific star formation rate sSFR (≡ SFR/M*) passive (“red-sequence”) galaxies is strongly re- gas is
(Fig. 3). For the 30 galaxies with sSFR ≥ 10−11
flected in their gaseous halos, and that the CGM Z⊙
year−1, there were 27 detections with a typical out to at least 150 kpc either directly influences or Mgas ¼ 177 MO
Z
column density log NOVI = 14.5 (24) and a is directly affected by star formation. Z⊙ 0:2
high covering fraction fhit ≈ 0.8 to 1 maintained O VI is a fragile ionization state that never ¼ 2 Â 109 M⊙ ð2Þ
Z fOVI
all the way out to R = 150 kpc (Fig. 2). For the exceeds a fraction fOVI = 0.2 of the total oxygen
12 galaxies in the passive subsample (sSFR ≤ for the physical conditions of halo gas and is where Z is the gas metallicity, and the solar
10−11 year−1), there were only four detections with frequently much less abundant (Fig. 4). Our ob- oxygen abundance is nO/nH = 5 × 10−4 (26).
lower typical NOVI than the star-forming sub- servations imply a typical CGM oxygen mass Even for the most conservative ionization cor-
sample (25). Accounting for the upper limits in MO, for star-forming galaxies, of rection ( fOVI = 0.2), the OVI-traced CGM con-
NOVI and sSFR, we can reject the null hypothesis tains a mass of metals and gas that is substantial
0:2
that there is no correlation between NOVI and M O ¼ 5pR2 〈N OVI 〉mO fhit relative to other reservoirs of interstellar and cir-
sSFR at 99.9% confidence for the whole sample fOVI cumgalactic gas. If our sample galaxies lie on the
7 0:2
and 98% for each of the 50-kpc annuli shown in ¼ 1:2 Â 10 M⊙ ð1Þ mean trend of gas fraction for low-z galaxies
Fig. 1 (21). This effect remained even when we fOVI (27), they have interstellar medium (ISM) gas
controlled for stellar mass: A Kolmogorov-Smirnov where we have taken a typical mean column masses of MISM = 5 × 109 to 10 × 109 M◉ and
test over log M* 10.5, where the star-forming density 〈NOVI〉 = 1014.5 cm−2 and R = 150 kpc, and contain M O = 2 × 107 to 10 × 107 M◉ of
ISM
950 18 NOVEMBER 2011 VOL 334 SCIENCE www.sciencemag.org

5. REPORTS
A B metals retained in the ISM. Thus, the detected
oxygen could be the cumulative effect of steady
enrichment over the preceding several billion
years, the product of sporadic flows driven by
rapid starbursts and an active nucleus (33), or the
fossil remains of outflows from as early as z ≈
1.5 to 3 (7, 31). Although the exact origin of the
mass-metallicity relation of galaxies is not yet
known, models that explain it in terms of
preferential loss of metals imply that a substantial
fraction of the metals produced by star formation
must be ejected from the galaxy rather than
retained in the ISM (28). The CGM detected here
could be a major reservoir of this ejected ma-
terial, with important consequences for models of
galactic chemical evolution.
Fig. 4. CGM oxygen masses compared to galactic reservoirs. (A) The curves and the axis labels at right The O VI we observe arises in bulk flows of
show the fraction of gas-phase oxygen in the O VI ionization state fOVI as a function of temperature, for gas over 100 to 400 km s−1, but the relative
Downloaded from www.sciencemag.org on November 27, 2011
three overdensities relative to the cosmic mean, r/r. All values of r/ r ≥ 1000 track the black curve on velocities are usually below halo escape speeds
which collisional ionization dominates, whereas for lower values, photoionization by the extragalactic (Fig. 2), even when we take projection effects
background can increase fOVI at low T. For gas that traces dark matter, r/ r = 1000 is typical at R ≈ 100 into account (fig. S1). Thus, much of the mate-
kpc; r/ r = 50 to 100 for the outskirts of the halo. The pale green band shows the expected oxygen mass of
rial driven into the halo by star formation could
the galaxies’ ISM if they lie on the standard relation between MISM and M* and follow the mass-metallicity
eventually be reacquired by the galaxy in “re-
relation (MZR). The green dashed line shows the oxygen mass produced by 3 × 109 M◉ of star formation.
The yellow band shows the expected oxygen mass for the extreme assumption that the typical host dark- cycled winds,” which may be an important source
matter halos (2 × 1011 to 1012 M◉) have the universal baryon fraction and solar metallicity. (B) The CGM of fuel for ongoing star formation (34). It is un-
oxygen masses compared with the interstellar oxygen mass as a function of M*. Points with range bars likely that the detected gas is predominantly fresh
show the CGM oxygen mass MO implied by Eq. 1 for fOVI = 0.2, calculated separately for star-forming material accreting from the IGM because models
(blue) and passive (red) galaxies according to the hit rates in four bins of stellar mass. The purple curves of “cold mode” accretion predict very low me-
show the calculated MISM for typical star-forming galaxies in the SDSS, accounting for the mean MZR in
O tallicity and low covering fractions fhit ≈ 10 to
the central curve and its uncertainties in the shaded region. The data points increase their mass in inverse 20% (35, 36), and “hot mode” accretion typically
proportion to fOVI. involves gas at temperatures T 106 K with
undetectably low fOVI.
oxygen, taking into account the observed corre- to produce a 1014.5 cm−2 column density within The passive galaxies in our sample once
lation between galaxy stellar mass and ISM the confines of a galactic halo, especially if the formed stars; thus, it follows that they would
metallicity (Fig. 4) (21). The minimum CGM metallicity is low (fig. S5). Thus, fOVI = 0.02 and once have possessed halos of ionized, metal-
oxygen mass is thus 10 to 70% of the ISM ox- Z = 0.1Z◉ are plausible conditions for the O VI– enriched gas visible in O VI. The relative paucity
ygen (Fig. 4 and fig. S4). The covering fractions traced gas, but it is unlikely that both conditions of O VI around these galaxies implies that this
and column densities we find for star-forming hold simultaneously. However, if either condition material was transformed by processes that plau-
galaxies are insensitive to M*, whereas the ISM holds, the CGM detected here could represent an sibly accompany the quenching of star formation
metal masses decline steeply with M* according important contribution to the cosmic budgets of (37), such as tidal stripping in group environ-
to the mass-metallicity relation. Thus, the ratio metals and baryons. In either case, Mgas is com- ments, reaccretion onto the galaxy in ionized
of CGM metals to ISM metals appears to increase parable to the total ~3 × 1010 M◉ inside R = 300 kpc form, or heating or cooling to a temperature at
for lower-mass galaxies (assuming constant fOVI), inferred from H I measurements at low redshift which O VI is too rare to detect. Our findings
perhaps indicating that metals more easily escape (19) and to the ~4 × 1010 M◉ inferred for the present a quantitative challenge for theoretical
from their shallower gravitational potentials. The CGM surrounding rapidly star-forming galaxies models of galaxy growth and feedback, which
implied total mass of circumgalactic gas Mgas is at z ≈ 2 to 3 (31). By generalizing our typical MO must explain both the ubiquitous presence of mas-
more uncertain because it can strictly take on any to all star-forming galaxies with M* 109.5 M◉, sive, metal-enriched ionized halos around star-
metallicity; for a fiducial solar metallicity, Eq. 2 we estimate that the halos of such galaxies con- forming galaxies and the fate of these metals after
implies a total CGM mass comparable to MISM tain 15% × (0.02/fOVI) of the oxygen in the uni- star formation ends.
and several times the total mass inferred for Milky verse and 2% × (0.02/fOVI) × (Z◉/Z) of the
Way “high-velocity clouds” (28, 29) or for low- baryons in the universe. References and Notes
ionization (Mg II) gas surrounding low-redshift The metals detected out to R ≈ 150 kpc must 1. In astronomical usage, metals are those elements heavier
galaxies to R = 100 kpc (30). have been produced in galaxies, after which they than hydrogen and helium; they are formed only by
For the densities typically expected at radii were likely transported into the CGM in some stellar nucleosynthesis.
2. S. Veilleux, G. Cecil, J. Bland-Hawthorn, Annu. Rev.
R ≈ 100 kpc, fOVI exceeds 0.1 only over a narrow form of outflow. However, these outflows need Astron. Astrophys. 43, 769 (2005).
temperature range 105.4−5.6 K, and it exceeds not be active at the time of observation; indeed, 3. M. D. Lehnert, T. M. Heckman, Astrophys. J. 462, 651
0.02 only over 105.2−5.7 K (Fig. 4). Either a large the large masses imply long time scales. Because (1996).
fraction of CGM gas lies in this finely tuned 1 M◉ of star formation eventually returns 0.014 M◉ 4. C. L. Martin, Astrophys. J. 621, 227 (2005).
5. D. S. Rupke, S. Veilleux, D. B. Sanders, Astrophys. J.
temperature range—a condition that is difficult to of oxygen to the ISM (32), at least 8.6 × 108 M◉ of Suppl. Ser. 160, 115 (2005).
maintain because gas cooling rates peak at T ≈ star formation is required to yield the detected 6. A. E. Shapley, C. C. Steidel, M. Pettini, K. L. Adelberger,
105.5 K—or the CGM oxygen and gas masses are oxygen mass. This is equivalent to ~3 × 108 years Astrophys. J. 588, 65 (2003).
much larger than the minimum values we have of star formation at the median SFR = 3 M◉ year−1 7. B. J. Weiner et al., Astrophys. J. 692, 187 (2009).
8. V. Springel, L. Hernquist, Mon. Not. R. Astron. Soc. 339,
quoted above. Lower-density photoionized gas of our star-forming sample, in the unlikely event 312 (2003).
can achieve high fOVI ≈ 0.1 over a wider tem- that all oxygen produced is expelled to the CGM, 9. B. D. Oppenheimer, R. Davé, Mon. Not. R. Astron. Soc.
perature range, but at these low densities it is hard and longer in inverse proportion to the fraction of 373, 1265 (2006).
www.sciencemag.org SCIENCE VOL 334 18 NOVEMBER 2011 951

6. REPORTS
10. D. Keres, N. Katz, D. H. Weinberg, R. Davé, Mon. Not. R. (14, 15, 38) and is higher than the mean value (14.0) 39. J. N. Bregman, E. D. Miller, A. E. Athey, J. A. Irwin,
Astron. Soc. 363, 2 (2005). measured by the Far Ultraviolet Spectroscopic Explorer Astrophys. J. 635, 1031 (2005).
11. A. Dekel, Y. Birnboim, Mon. Not. R. Astron. Soc. 368, through the halo of the Milky Way (20), which would Acknowledgments: We thank the anonymous reviewers
2 (2006). belong in our star-forming sample. for constructive comments. This work is based on
12. G. Kauffmann et al., Mon. Not. R. Astron. Soc. 341, 25. O VI emission is seen in elliptical galaxies (39), but observations made for program GO11598 with the
33 (2003). this gas is most likely associated with the ISM and not NASA/ESA Hubble Space Telescope, obtained at the
13. T. M. Tripp, B. D. Savage, E. B. Jenkins, Astrophys. J. the CGM. Space Telescope Science Institute, operated by AURA
534, L1 (2000). 26. M. Asplund, N. Grevesse, A. J. Sauval, P. Scott, Annu. Rev. under NASA contract NAS 5-26555, and at the
14. C. W. Danforth, J. M. Shull, Astrophys. J. 679, 194 Astron. Astrophys. 47, 481 (2009). W. M. Keck Observatory, operated as a scientific
(2008). 27. M. S. Peeples, F. Shankar, Mon. Not. R. Astron. Soc. 417, partnership of the California Institute of Technology, the
15. C. Thom, H.-W. Chen, Astrophys. J. 683, 22 (2008). 2962 (2011). University of California, and NASA. The Observatory was
16. J. N. Bregman, Annu. Rev. Astron. Astrophys. 45, 28. M. E. Putman, Astrophys. J. 645, 1164 (2006). made possible by the generous financial support of the
221 (2007). 29. N. Lehner, J. C. Howk, Science 334, 955 (2011); W. M. Keck Foundation. The Hubble data are available
17. J. T. Stocke et al., Astrophys. J. 641, 217 (2006). 10.1126/science.1209069. from the MAST archive at http://archive.stsci.edu.
18. H.-W. Chen, J. S. Mulchaey, Astrophys. J. 701, 1219 30. H.-W. Chen et al., Astrophys. J. 714, 1521 (2010). M.S.P. was supported by the Southern California Center
(2009). 31. C. C. Steidel et al., Astrophys. J. 717, 289 (2010). for Galaxy Evolution, a multicampus research program
19. J. X. Prochaska, B. Weiner, H.-W. Chen, J. S. Mulchaey, 32. D. Thomas, L. Greggio, R. Bender, Mon. Not. R. funded by the UC Office of Research.
K. L. Cooksey, http://arxiv.org/abs/1103.1891 (2011). Astron. Soc. 296, 119 (1998).
20. K. R. Sembach et al., Astrophys. J. Suppl. Ser. 146, 33. T. M. Tripp et al., Science 334, 952 (2011).
165 (2003). 34. B. D. Oppenheimer et al., Mon. Not. R. Astron. Soc. 406, Supporting Online Material
21. See supporting material on Science Online. 2325 (2010). www.sciencemag.org/cgi/content/full/334/6058/948/DC1
22. J. K. Werk et al., http://arxiv.org/abs/1108.3852 35. K. R. Stewart et al., Astrophys. J. 735, L1 (2011).
Downloaded from www.sciencemag.org on November 27, 2011
SOM Text
(2011). 36. M. Fumagalli et al., http://arxiv.org/abs/1103.2130 Figs. S1 to S5
23. D. Schiminovich et al., Astrophys. J. Suppl. Ser. 173, (2011). Tables S1 and S2
315 (2007). 37. J. M. Gabor, R. Davé, K. Finlator, B. D. Oppenheimer, References (40–62)
24. The typical log NOVI = 14.5 to 15.0 for star-forming Mon. Not. R. Astron. Soc. 407, 749 (2010).
galaxies resembles the high end of the column-density 38. T. M. Tripp et al., Astrophys. J. Suppl. Ser. 177, 15 June 2011; accepted 27 September 2011
distribution seen in blind surveys of intergalactic clouds 39 (2008). 10.1126/science.1209840
the total column density and mass of the outflows
The Hidden Mass and Large Spatial are poorly constrained. Previous outflow obser-
vations were often limited to low-resolution spec-
Extent of a Post-Starburst Galaxy Outflow tra of only one or two ions (e.g., Na I or Mg II) or
relied on composite spectra that cannot yield precise
Todd M. Tripp,1* Joseph D. Meiring,1 J. Xavier Prochaska,2 Christopher N. A. Willmer,3 column densities. Without any constraints on hydro-
J. Christopher Howk,4 Jessica K. Werk,2 Edward B. Jenkins,5 David V. Bowen,5 Nicolas Lehner,4 gen (the vast bulk of the mass) or other elements
Kenneth R. Sembach,6 Christopher Thom,6 Jason Tumlinson6 and ions, these studies were forced to make highly
uncertain assumptions to correct for ionization,
Outflowing winds of multiphase plasma have been proposed to regulate the buildup of galaxies, elemental abundances, and depletion of species
but key aspects of these outflows have not been probed with observations. By using ultraviolet by dust. Lastly, galactic winds contain multiple
absorption spectroscopy, we show that “warm-hot” plasma at 105.5 kelvin contains 10 to 150 times phases with a broad range of physical conditions
more mass than the cold gas in a post-starburst galaxy wind. This wind extends to distances 68 (6), and wind gas in the key temperature range
kiloparsecs, and at least some portion of it will escape. Moreover, the kinematical correlation of between 105 to 106 K (where radiative cooling is
the cold and warm-hot phases indicates that the warm-hot plasma is related to the interaction of maximized) is too cool to be observed in x-rays;
the cold matter with a hotter (unseen) phase at 106 kelvin. Such multiphase winds can detection of this so-called “warm-hot” phase
remove substantial masses and alter the evolution of post-starburst galaxies. requires observations in the ultraviolet (UV).
To study the more extended gas around gal-
alaxies do not evolve in isolation. They in- galaxies (2) and eventually into elliptical-type axies, including regions affected by outflows, we
G teract with other galaxies and, more subtly,
with the gas in their immediate environ-
ments. Mergers of comparable-mass, gas-rich
galaxies with little or no star formation (3).
Mergers are not required to propel galaxy evo-
lution, however. Even relatively secluded galaxies
used the Cosmic Origins Spectrograph (COS)
on the Hubble Space Telescope (HST) to obtain
high-resolution spectra of the quasi-stellar object
galaxies trigger star-formation bursts by driving accrete matter from the intergalactic medium (QSO) PG1206+459 (at redshift zQSO = 1.1625).
matter into galaxy centers, but theory predicts that (IGM), form stars, and drive matter outflows into By exploiting absorption lines imprinted on the
such starbursts are short-lived: The central gas is their halos or out of the galaxies entirely (4, 5). QSO spectrum by foreground gaseous material,
rapidly driven away by escaping galactic winds In either case, the competing processes of gas we can detect the low-density outer gaseous en-
powered by massive stars and supernova explo- inflows and outflows are expected to regulate velopes of galaxies, regions inaccessible to other
sions or by a central supermassive black hole galaxy evolution. techniques. We focus on far-ultraviolet (FUV) ab-
(1). Such feedback mechanisms could trans- Outflows are evident in some nearby objects sorption lines at rest wavelengths lrest 912 Å.
form gas-rich spiral galaxies into post-starburst (6–9) and are ubiquitous in some types of gal- This FUV wavelength range is rich in diagnostic
axies (10–15); their speeds can exceed the escape transitions (23), including the Ne VIII 770.409,
1 velocity. Nevertheless, their broader impact on 780.324 Å doublet, a robust probe of warm-hot
Department of Astronomy, University of Massachusetts, Am-
herst, MA 01003, USA. 2University of California Observatories/ galaxy evolution is poorly understood. First, their gas, as well as banks of adjacent ionization stages.
Lick Observatory, University of California, Santa Cruz, CA 95064, full spatial extent is unknown. Previous studies The sight line to PG1206+459 pierces an absorp-
USA. 3Steward Observatory, University of Arizona, Tucson, AZ (6, 9, 16–22) have revealed flows with spatial tion system, at redshift zabs = 0.927, that provides
85721, USA. 4Department of Physics, University of Notre Dame, extents ranging from a few parsecs up to ~20 kilo- insights about galactic outflows. This absorber
Notre Dame, IN 46556, USA. 5Princeton University Obser-
vatory, Princeton, NJ 08544, USA. 6Space Telescope Science parsecs (kpc). However, because of their low has been studied before (24), but previous obser-
Institute, Baltimore, MD 21218, USA. densities, outer regions of outflows may not have vations did not cover Ne VIII and could not pro-
*To whom correspondence should be addressed. E-mail: been detected with previously used techniques, vide accurate constraints on H I in the individual
tripp@astro.umass.edu and thus the flows could be much larger. Second, absorption components.
952 18 NOVEMBER 2011 VOL 334 SCIENCE www.sciencemag.org