A Reservoir of Ionized Gas in the Galactic Halo to Sustain Star Formation in the Milky Way Nicolas Lehner, et al. Science 334, 955 (2011); DOI: 10.1126/science.1209069 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/334/6058/955.full.html Supporting Online Material can be found at: http://www.sciencemag.org/content/suppl/2011/08/24/science.1209069.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/334/6058/955.full.html#related This article has been cited by 1 articles hosted by HighWire Press; see: http://www.sciencemag.org/content/334/6058/955.full.html#related-urlsScience (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright2011 by the American Association for the Advancement of Science; all rights reserved. The title Science is aregistered trademark of AAAS.
REPORTSoutflow studies. In principle, the O VI 1032,1038 Å 8. D. S. Rupke, S. Veilleux, D. B. Sanders, Astrophys. J. 36. C. Tremonti, A. M. Diamond-Stanic, J. Moustakas, indoublet can reveal such gas, but it is unclear Suppl. Ser. 160, 87 (2005). Galaxy Evolution: Emerging Insights and Future 9. C. L. Martin, Astrophys. J. 647, 222 (2006). Challenges, S. Jogee, I. Marinova, L. Hao, G. Blanc, Eds,whether the O VI arises in photoionized 104 K 10. M. Pettini et al., Astrophys. J. 554, 981 (2001). (Astronomical Society of the Pacific Conference Series,gas, hotter material at ~105.5 K, or both (35). The 11. C. A. Tremonti, J. Moustakas, A. M. Diamond-Stanic, San Francisco, 2009), vol. 419, pp. 369–376.Ne VIII doublet avoids this ambiguity, and we Astrophys. J. 663, L77 (2007). 37. A. L. Coil et al., http://arXiv.org/abs/1104.0681have found that this warm-hot matter is a sub- 12. C. C. Steidel et al., Astrophys. J. 717, 289 (2010). (2011). 13. K. H. R. Rubin et al., Astrophys. J. 719, 1503 (2010). 38. J. Tumlinson et al., Science 334, 948 (2011).stantial component in the mass inventory of a 14. F. Hamann, G. Ferland, Annu. Rev. Astron. Astrophys. 37, 39. B. D. Savage, K. R. Sembach, Astrophys. J. 379, 245galactic wind. Moreover, this wind has a large 487 (1999). (1991).spatial extent, and the mass carried away by the 15. J. P. Grimes et al., Astrophys. J. Suppl. Ser. 181, 272 (2009). Acknowledgments: This study has its basis in observationsoutflow will affect the evolution of the galaxy. 16. K. H. R. Rubin, J. X. Prochaska, D. C. Koo, A. C. Phillips, made with the NASA/European Space Agency Hubble B. J. Weiner, Astrophys. J. 712, 574 (2010). Space Telescope (HST); the MMT, a joint facilityWhereas earlier studies of poststarburst outflows 17. M. Moe, N. Arav, M. A. Bautista, K. T. Korista, operated by the Smithsonian Astrophysical Observatoryfocused on Mg II and could not precisely con- Astrophys. J. 706, 525 (2009). and the University of Arizona; and the Large Binocularstrain the metallicity, hydrogen column, and 18. J. P. Dunn et al., Astrophys. J. 709, 611 (2010). Telescope, an international collaboration amongmass, these studies do indicate that post-starburst 19. D. Edmonds et al., Astrophys. J. 739, 7 (2011). institutions in the United States, Italy, and Germany. 20. F. Hamann et al., Mon. Not. R. Astron. Soc. 410, 1957 Support for HST program number 11741 was providedoutflows are common: 22/35 of the post-starbursts (2011). by NASA through a grant from the Space Telescopein (36) showed outflowing Mg II absorption with 21. G. A. Kriss et al., Astron. Astrophys. 534, 41 (2011). Science Institute, which is operated by the Associationmaximum (radial) velocities of 500 to 2400 km s−1, 22. D. M. Capellupo, F. Hamann, J. C. Shields, P. Rodríguez of Universities for Research in Astronomy, Incorporated,similar to the absorption near 177_9 (Fig. 1), and Hidalgo, T. Barlow, Mon. Not. R. Astron. Soc. 413, 908 under NASA contract NAS5-26555. Additional support Downloaded from www.sciencemag.org on November 27, 201177 and 100% of the post-starburst and AGN (2011). was provided by NASA grant NNX08AJ44G. The DEEP2 23. D. A. Verner, D. Tytler, P. D. Barthel, Astrophys. J. 430, survey was supported by NSF grants AST 95-29098,galaxies, respectively, in (37) drive outflows but 186 (1994). 00-711098, 05-07483, 08-08133, 00-71048,with lower maximum velocities, which may be 24. J. Ding, J. C. Charlton, C. W. Churchill, C. Palma, 05-07428, and 08-07630. Funding for the Sloan Digitaldue to selection of wind-driving galaxies in a Astrophys. J. 590, 746 (2003). Sky Survey has been provided by the Alfred P. Sloanlater evolutionary stage. With existing COS data, 25. See further information in supporting material on Science Foundation, the Participating Institutions, NASA, NSF, the Online. U.S. Department of Energy Office of Science, thethe effects of large-scale outflows on galaxy evo- 26. B. D. Savage, N. Lehner, B. P. Wakker, K. R. Sembach, Japanese Monbukagakusho, and the Max Planck Society.lution can be studied with the techniques pre- T. M. Tripp, Astrophys. J. 626, 776 (2005). We thank C. Churchill for providing the archival Kecksented here but with larger samples (38), with 27. A. Narayanan et al., Astrophys. J. 730, 15 (2011). data and the referees for review comments thatwhich it will be possible to statistically track how 28. Y. Chen, J. D. Lowenthal, M. S. Yun, Astrophys. J. 712, significantly improved this paper. We are also grateful to 1385 (2010). the Hawaiian people for graciously allowing us to conductoutflows affect galaxies. 29. The galaxy appears to be red in Fig. 4 because of its observations from Mauna Kea, a revered place in redshift; in the rest frame of the galaxy, it has a very the culture of Hawaii. The HST data in this paper References and Notes ultraviolet-blue (U-B) color. are available from the Multimission Archive at the 1. P. F. Hopkins et al., Astrophys. J. Suppl. Ser. 163, 30. C. N. A. Willmer et al., Astrophys. J. 647, 853 (2006). Space Telescope Science Institute (MAST) at 1 (2006). 31. For distance calculations, we assume a cold dark matter http://archive.stsci.edu. 2. A. I. Zabludoff et al., Astrophys. J. 466, 104 (1996). cosmology with Hubble constant H0 = 70 km s−1 Mpc−1 3. G. F. Snyder, T. J. Cox, C. C. Hayward, L. Hernquist, and dimensionless density parameters Ωm = 0.30, ΩL = P. Jonsson, Astrophys. J. 741, 77 (2011). 0.70. Supporting Online Material 4. D. Kereš, N. Katz, R. Davé, M. Fardal, D. H. Weinberg, 32. G. J. Ferland et al., Publ. Astron. Soc. Pac. 110, 761 www.sciencemag.org/cgi/content/full/334/6058/952/DC1 (1998). Materials and Methods Mon. Not. R. Astron. Soc. 396, 2332 (2009). 5. B. D. Oppenheimer et al., Mon. Not. R. Astron. Soc. 406, 33. As discussed in the supporting online material, the yellow SOM Text galaxy northwest of the QSO (Fig. 4) does not have a Figs. S1 to S5 2325 (2010). Tables S1 and S2 6. S. Veilleux, G. Cecil, J. Bland-Hawthorn, Annu. Rev. spectroscopic redshift but is likely to have z << 0.927. Astron. Astrophys. 43, 769 (2005). 34. A. C. Fabian et al., Nature 454, 968 (2008). References (40–54) 7. T. M. Heckman, M. D. Lehnert, D. K. Strickland, L. Armus, 35. T. M. Tripp et al., Astrophys. J. Suppl. Ser. 177, 39 15 June 2011; accepted 26 October 2011 Astrophys. J. Suppl. Ser. 129, 493 (2000). (2008). 10.1126/science.1209850 metal-poor gas (metallicity that is less than 10%A Reservoir of Ionized Gas in the of that of the Sun, or Z ≲ 0.1 Z◉) to flow onto galaxies along dense intergalactic filaments (1).Galactic Halo to Sustain Star However, galaxies may also exchange mass with the local intergalactic medium (IGM) through outflows driven by galactic “feedback,” galacticFormation in the Milky Way winds powered by massive stars and their death and from massive black holes. Some of this ma-Nicolas Lehner* and J. Christopher Howk terial may return to the central galaxy as recycled infalling matter—the galactic fountain mechanismWithout a source of new gas, our Galaxy would exhaust its supply of gas through the formation (2, 3). The circumgalactic medium about a gal-of stars. Ionized gas clouds observed at high velocity may be a reservoir of such gas, but their axy is thus a complicated blend of outflowingdistances are key for placing them in the galactic halo and unraveling their role. We have used metal-rich and infalling metal-poor gas. The rela-the Hubble Space Telescope to blindly search for ionized high-velocity clouds (iHVCs) in the tive importance of these processes is poorly con-foreground of galactic stars. We show that iHVCs with 90 ≤ |vLSR| ≲ 170 kilometers per second strained observationally. Here, we demonstrate(where vLSR is the velocity in the local standard of rest frame) are within one galactic radius of the that ionized gas in the local galactic halo providesSun and have enough mass to maintain star formation, whereas iHVCs with |vLSR| ≳ 170 kilometers a major supply of matter for fueling ongoing starper second are at larger distances. These may be the next wave of infalling material. formation. he time scale for gas consumption via star eous fuel in the disks of galaxies for continuedT formation in spiral galaxies is far shorter than a Hubble time (13.8 billion years),requiring an ongoing replenishment of the gas- star formation. Analytical models and hydrody- namical simulations have emphasized the impor- tance of cold-stream accretion as a means for Department of Physics, University of Notre Dame, 225 Nieuwland Science Hall, Notre Dame, IN 46556, USA. *To whom correspondence should be addressed. E-mail: email@example.com www.sciencemag.org SCIENCE VOL 334 18 NOVEMBER 2011 955
REPORTS Inflow and outflow in the Milky Way halo can tocentric distance of <17.7 kpc and subsolar could be situated at 4 ≲ z ≲ 9 kpc. There is no be studied via the so-called high-velocity clouds metallicity; it probably traces gas that is being iHVC absorption at |vLSR| ≳ 170 km s−1 (VHVCs) (HVCs), clouds moving in the local standard of accreted by the Milky Way, possibly associated toward the stars (Fig. 2 and SOM text), even rest (LSR) frame at |vLSR| ≥ 90 km s−1 (4). De- with the well-known large H I complex C (18, 19). though these are observed along the path to termining the distance (d ) of these HVCs is These studies show that some iHVCs probe gas AGNs (Fig. 2 and SOM text). Thus, there is critical for associating HVCs with flows occur- flows in and out of the galactic disk, with some no evidence for VHVCs in the galactic halo at ring near the Milky Way rather than the IGM of originating from the Milky Way and others having |z| ≲ 10 kpc. the Local Group and for quantifying their basic an extragalactic origin. In order to understand the implications of the physical properties because several of these di- Here, we generalize the result to understand iHVC detection rate toward stars, we need to rectly scale with the distance (for example, the these iHVCs in the context of the Milky Way compare it with a distance-independent measure mass M º d2). Major progress has been made in evolution with a survey of the gas in the fore- of the iHVC covering fraction. Earlier studies the past decade for some of the large, predom- ground of 28 distant galactic halo stars with known inantly neutral HVC complexes, placing them distances. These stars were observed with the 4 to 13 kpc from the sun (excluding here the Cosmic Origins Spectrograph (COS) and Space- Magellanic Stream) (5–8). Approximately 37% Telescope Imaging Spectrograph (STIS) on board of the galactic sky is covered by neutral atomic the Hubble Space Telescope (HST), and a large hydrogen (H I) HVCs with column densities majority (23 of 28) of these were obtained through N(H I) ≥ 1017.9 cm−2 (9), but the infall rate of the our HST Cycle 17 program 11592 [supporting Downloaded from www.sciencemag.org on November 27, 2011 largest H I HVC complex within 10 kpc (com- online material (SOM) text]. The main criterion plex C), ∼0.14 M◉ year−1 (where M◉ is the mass for assembling this stellar sample is that these of the Sun) (8), is far too modest for replenish- ultraviolet (UV) bright stars are at height |z| ≳ 3 kpc ing the 0.6 to 1.45 M◉ year−1 consumed by from the galactic plane. This minimum height Milky Way star formation (10) (we use through- was adopted in view of absence or scarcity of out the following spectroscopic notation: if X is a iHVCs at smaller z (20) and other works on their given atom, then X I refers to the neutral atom X, predominantly neutral counterparts (5). We sys- X II refers to the singly ionized atom X, X IV tematically searched for high-velocity interstel- refers to the triply ionized atom X...). This is lar metal-line absorption in the COS and STIS not entirely surprising, given that recent mod- UV spectra of these stars (SOM text). We noted els show that inflowing gas should be predomi- HVC detections only if high-velocity absorp- nantly ionized in view of the small amounts of tion was seen in multiple ions or transitions (at neutral gas available for inflow at any epoch a minimum two) (Fig. 1). We measured the ap- (11). In ionized gas, N(H I) becomes small rela- parent optical depth-weighted mean velocity 〈v〉 = tive to N(H II), and the H I emission becomes ∫v ta(v)dv /∫ta(v)dv (table S1). Most of the HVCs extremely difficult to impossible to detect. Low seen in absorption against the stars do not have H I content HVCs are, however, routinely found H I 21-cm emission at a level of ≳1018.5 cm−2 in absorption in the spectra of cosmologically according to the Leiden/Argentine/Bonn (LAB) distant objects, such as active galactic nuclei survey (21). The column densities of Si II and O I (AGNs), with a detection rate of ∼60 to 80% imply large ionization fractions in several HVCs (12–15). Their total (neutral and ionized) hydro- of our sample (Fig. 1 and SOM text) (22, 17), gen column density is shown to be quite large [as demonstrating that N(H II) >> N(H I). large as N(H I) in predominantly neutral HVCs]. The iHVC detection rate in our stellar sam- These are therefore ionized HVCs (iHVCs)— ple is 50% (14 of 28). Although a sightline may that is, N(H II) >> N(H I) (12, 16). Given their large have more than one high-velocity absorption covering factor, the iHVCs may represent the component, only one HVC for a given sightline is long-sought supply of gas needed for continued counted for estimating the covering factor. De- Milky Way star formation. However, the iHVCs fining a sample with a more uniform sensitivity have been mostly detected against AGNs: They Wl ≥ 15 mÅ near Si IIl1526 and with no stellar Fig. 1. Example of COS continuum normalized ab- may reside within the Galaxy, the Local Group, contamination (that is, with no observed stellar sorption profiles of various metal-lines and (Top) or the IGM. Thus, as for their larger H I column photospheric absorption at |vLSR| ≥ 90 km; see LAB H I emission line profile toward PG0914+001, density counterparts, direct distance constraints flag Q = 1 in table S1 and SOM text for more a star at d = 16 kpc and z = +8.4 kpc. There are at are required for determining their masses and for details) gives essentially the same detection rate least two iHVCs seen in absorption at +100 and characterizing their role in the evolution of the with 47% (9 of 19). The average distance of the +170 km s−1, as indicated by the dotted lines and Milky Way. stars in the latter sample in which the iHVCs are shaded regions. The ∼0 km s−1 absorption and H I To determine the distances of iHVCs can be detected is 11.5 T 4.1 kpc; the average absolute emission are from the Milky Way disk. We derive for the iHVC component [O I/S II] ≡ log[N(O I)/N(S II)] − directly undertaken by observing the gas in the z-height is 7.3 T 3.0 kpc; for the whole sample, log(AO/AS)◉ ≅ −2.34, where (AO/AS)◉ is the relative foreground of stars at known distances from the these are 〈d〉 = 11.6 T 6.9 kpc and 〈|z|〉 = 6.5 T solar abundance of oxygen to sulfur, implying H II >> sun. Recently, on the basis of observations of 3.2 kpc. Our sky coverage is larger at b > 0° than H I. If the iHVC has a solar abundance, N(H II) > high-velocity interstellar absorption in the ultra- at b < 0° (Fig. 2); the difference of detection rates 1019.6 cm−2 (based on S II), and N(H I) ≅ 1017.30 cm−2 violet spectra of two galactic stars, two of these between the northern (60%, 12 of 20) and south- (based on O I). These column densities would in- iHVCs were found within 8 to 15 kpc from the ern (29%, 2 of 7) sky may be due to statistical crease if the iHVC metallicity is subsolar. We find sun (17, 18). One of them was found toward fluctuations in the smaller southern sample. very subsolar [Fe II/S II] < −1.5 and [Si II/S II] ≅ the inner Galaxy and has supersolar metallicity, With the better sampled northern galactic sky, −1.4 ratios, indicating strong ionization and dust probing gas that has been ejected from and is there is some evidence for an increase in the de- depletion effects; the latter could suggest a galactic raining back onto the Milky Way disk (17). An- tection rate to 73% (11 of 15) for sightlines with origin. Examples of negative-velocity iHVCs toward other was observed at l ∼ 103° with a galac- z ≳ 4 kpc, implying that most of the iHVCs stars can be found in (17, 18).956 18 NOVEMBER 2011 VOL 334 SCIENCE www.sciencemag.org
REPORTSdetermined the covering factors of these iHVCs slow through their interaction with the galactic cm−2. These assumptions imply a total masstoward AGNs (12–14, 16), but they concentrated gas as they approach the plane, as predicted by MiHVC ≈ 1.1 × 108 (d/12 kpc)2 ( fc/0.5)(Z/0.2Z◉)−1on a single species (O VI or Si III), which differs some models of HVCs (23–26). In the latter M◉. We estimate infall times of 80 to 130 mil-from our search method. We combined observa- scenario, some of the VHVCs could be at large lion years, assuming infall velocity of about 90tions from (14) and (15) to assemble an AGN distances from the Milky Way or other galaxies to 150 km s−1, which implies a mass infall ratesample using the same search criteria adopted for (such as Andromeda) and be a component of the for the ionized gas of ∼0.8 to 1.4 M◉ year−1 (d =our stellar survey (using the same metal ions) multiphase local intergalactic medium, which may 12 kpc, fc = 0.5, and Z = 0.2Z◉). Although this(SOM text). Our AGN sample is summarized in permeate the Local Group of galaxies (27, 28). value may be reduced somewhat because theretable S2, and its sky distribution is shown on A majority of the so-called O VI HVCs have must be a mixture of outflows and inflows, largeFig. 2. It has a similar sensitivity and size as the accompanying C III and H I HVC absorption at H I complexes have subsolar abundances (5),stellar sample. The detection rate is 67% (16 of similar high velocities, demonstrating their mul- suggesting a substantial fraction of the circum-24) for the iHVCs and 42% (11 of 26) for the tiphase nature (13). The velocity sky-distribution galactic neutral and ionized gas is infalling. SomeVHVCs (excluding the four Magellanic Stream of the O VI, Si III, and H I HVCs are also alike of the outflowing gas must also be recycled viasightlines). For the HVC sample, six sightlines (12–14), and the sky-distribution of the iHVCs a galactic fountain because the iHVCs do notwith H I LAB emission at 90 ≤ |vLSR| ≤ 170 km s−1 seen toward the AGNs and stars is moreover reach the Milky Way escape velocity. We didwere excluded because of strong selection effects remarkably similar considering our better sam- not bracket the distances of the iHVCs, butin the AGN sample: Many of these AGNs were pled north galactic sky (Fig. 2). These and the their distances are unlikely to be much smallerindeed initially targeted to study known H I HVCs fact that the H I HVCs and iHVCs are now than 3 kpc because otherwise they would al- Downloaded from www.sciencemag.org on November 27, 2011[including these sightlines yields a covering factor known to be at similar distances strongly suggest ready have been detected serendipitously in the( fc) = 73%]. After removing those, the AGN that they are all related, probing separate phases UV spectra of more nearby stars (20). If allsample may still be biased and overestimate the in which the H I HVCs may be the densest of the iHVCs were at d = 5 kpc, then thetrue covering factors because a given AGN could regions of the more diffuse ionized complexes. infall rate would still be significant with ∼0.4have been specifically targeted for studying an Having demonstrated that the iHVCs are to 0.6 M◉ year−1.HVC or could have been favored over other in the local galactic halo, we can reliably esti- Although the infall rate depends on severalAGNs because of previously known HVCs near mate their mass and assess their importance for parameters that are not well constrained, ourthe line of sight. future star formation in the Milky Way. The distance estimates allow us to unambiguously The covering factors of the iHVCs are mass of these iHVCs can be estimated MiHVC ≅ place the iHVCs at 90 ≤ |vLSR| ≤ 170 km s−1 intherefore fc = 50% (14 of 28) and ≤67% (16 of 1.3mH (4pd2) fc NH II ≈ 9.6 × 10−12 (d/12 kpc)2 the Milky Way’s halo. Assuming an average met-24) for the stellar and AGN samples, respective- ( fc/0.5)NHII M◉ (where mH is the mass of hydro- allicity of 0.2Z◉, the estimated infall rate is a fac-ly. Hence, a majority (if not all) of the iHVCs gen and the factor 1.3 accounts for the addi- tor of 6 to 10 larger than that of the large H Iseen toward AGNs are within 〈d 〉 ≅ 12 T 4 kpc tional mass of helium). The ionized gas probed complex C (6, 8). (Even with an unrealisticallyand 〈|z|〉 ≅ 6 T 3 kpc, implying that the iHVCs by Si II, Si III, and Si IV has 〈NH II〉 ≈ 6 × 1018 solar metallicity value for all the iHVCs, theirat 90 ≲ |vLSR| ≤ 170 km s−1 mostly trace flows (Z/0.2Z◉)−1 cm−2 (16), whereas the O VI phase has infall rate would still be important.) The present-of ionized gas in the Milky Way halo. On the 〈NH II 〉 ≈ 4.5 × 1018 (Z/0.2Z◉)−1 cm−2 (12, 13, 16), day total star formation rate of the Milky Way isother hand, VHVCs must then lie beyond d ≳ where the metallicity of 0.2Z◉ is representa- 0.6 to 1.45 M◉ year−1 (10), and chemical evo-10 to 20 kpc (|z| ≳ 6 to 10 kpc) because they tive of complex C and other large H I com- lution models require a present-day infall rate ofare not detected toward any star. The VHVCs plexes (5, 19). Because O VI and Si III are not only 0.45 M◉ year−1 (29)—a value that could becould be associated with the outer reaches of expected to exist in the same gas phase, the total somewhat reduced according to stellar mass lossthe Milky Way or gas in the Local Group. In H II column density is a sum of the phases traced models (30). The iHVCs are thus sufficient tothe former scenario, this would imply that iHVCs by these ions, NH II ≈ 1.1 × 1019 (Z/0.2Z◉)−1 sustain star formation in the Milky Way.Fig. 2. Aitoff projection galactic(longitude, latitude) map of thesurvey directions for the stellar sam-ple (circles) and extragalactic sample(square). A solid circle/square indi-cates an HVC in the foreground ofthe star/AGN, whereas an open circle/square implies no HVC along thestellar/AGN sightline. The velocityvalue is color-coded following thehorizontal color bar. The positiveand negative numbers indicate thez-height (in kpc) of the stars. Thesquares with an “X” are HVCs inwhich H I 21-cm LAB emission ispresent at similar velocities seenin absorption. An “S” near a squareindicates a sightline that passesthrough the Magellanic Stream. www.sciencemag.org SCIENCE VOL 334 18 NOVEMBER 2011 957
REPORTS We have implicitly assumed the HVCs 3. B. Oppenheimer et al., Mon. Not. R. Astron. Soc. 406, 23. R. A. Benjamin, L. Danly, Astrophys. J. 481, 764 survive their fall onto the galactic disk. Hydro- 2325 (2010). (1997). 4. B. P. Wakker, H. van Woerden, Annu. Rev. Astron. 24. J. E. G. Peek, M. E. Putman, C. F. McKee, C. Heiles, dynamical simulations of HVCs support this Astrophys. 35, 217 (1997). S. Stanimirović, Astrophys. J. 656, 907 (2007). idea (25, 26). In these models, the HVCs lose 5. B. P. Wakker, Astrophys. J. Suppl. Ser. 136, 463 25. F. Heitsch, M. E. Putman, Astrophys. J. 698, 1485 most of their H I content within 10 kpc of the (2001). (2009). disk and continue to fall toward the disk as warm 6. B. P. Wakker et al., Astrophys. J. 670, L113 (2007). 26. F. Marinacci et al., Mon. Not. R. Astron. Soc. 415, 1534 7. B. P. Wakker et al., Astrophys. J. 672, 298 (2008). (2011). ionized matter. This is consistent with the iHVCs 8. C. Thom et al., Astrophys. J. 684, 364 (2008). 27. L. Blitz, D. N. Spergel, P. J. Teuben, D. Hartmann, covering more galactic sky and tracing a larger 9. E. M. Murphy, F. J. Lockman, B. D. Savage, Astrophys. J. W. B. Burton, Astrophys. J. 514, 818 (1999). mass reservoir than do the predominantly neutral 447, 642 (1995). 28. F. Nicastro et al., Nature 421, 719 (2003). HVCs. Because the iHVCs are still overdense 10. T. P. Robitaille, B. A. Whitney, Astrophys. J. 710, L11 29. C. Chiappini, F. Matteucci, D. Romano, Astrophys. J. 554, (2010). 1044 (2001). relative to the halo medium, they can continue to 11. A. Bauermeister, L. Blitz, C.-P. Ma, Astrophys. J. 717, 323 30. S. N. Leitner, A. V. Kravtsov, Astrophys. J. 734, 48 sink toward the galactic plane, where they de- (2010). (2011). celerate and feed the warm ionized medium 12. K. R. Sembach et al., Astrophys. J. Suppl. Ser. 146, 165 Acknowledgments: This work was based on observations (Reynolds layer) 1 to 2 kpc from the galactic (2003). made with the NASA/European Space Agency Hubble 13. A. J. Fox, B. D. Savage, B. P. Wakker, Astrophys. J. Space Telescope, obtained at the Space Telescope Science disk. In the Reynolds layer, the clouds have low Suppl. Ser. 165, 229 (2006). Institute (STScI), which is operated by the Association of velocities and cannot be identified as HVCs 14. J. A. Collins, M. Shull, M. L. Giroux, Astrophys. J. 705, Universities for Research in Astronomy, under NASA anymore, but as low- or intermediate-velocity 962 (2009). contract NAS5-26555. We greatly appreciate funding clouds, which is consistent with the observed 15. P. Richter, J. C. Charlton, A. P. M. Fangano, N. B. Bekhti, support from NASA grant HST-GO-11592.01-A from Downloaded from www.sciencemag.org on November 27, 2011 absence of iHVCs at low z-height. In this sce- J. R. Masiero, Astrophys. J. 695, 1631 (2009). STScI. We are grateful to P. Chayer for computing several 16. J. M. Shull, J. R. Jones, C. W. Danforth, J. A. Collins, stellar spectra for us. The data reported in this paper are nario, VHVCs are also not expected to be seen Astrophys. J. 699, 754 (2009). tabulated in the SOM and archived at the Multimission near the galactic disk because the HVC velocities 17. W. F. Zech, N. Lehner, J. C. Howk, W. V. D. Dixon, Archive at STScI (MAST, http://archive.stsci.edu/). decrease with decreasing z-height due to drag. T. M. Brown, Astrophys. J. 679, 460 (2008). This is consistent with the lack of VHVCs at 18. N. Lehner, J. C. Howk, Astrophys. J. 709, L138 (2010). Supporting Online Material 19. T. M. Tripp, L. Song, Astrophys. J., http://arxiv.org/abs/ |z| ≲ 10 to 20 kpc and the observed velocity 1101.1107. www.sciencemag.org/cgi/content/full/science.1209069/DC1 SOM Text sky distribution (23). 20. J. Zsargó, K. R. Sembach, J. C. Howk, B. D. Savage, Tables S1 to S3 Astrophys. J. 586, 1019 (2003). References (31–41) References and Notes 21. P. M. W. Kalberla et al., Astron. Astrophys. 440, 775 1. D. Keres, L. Hernquist, Astrophys. J. 700, L1 (2009). (2005). 31 May 2011; accepted 17 August 2011 2. F. Fraternali, J. J. Binney, Mon. Not. R. Astron. Soc. 386, 22. N. Lehner, F. P. Keenan, K. R. Sembach, Mon. Not. R. Published online 25 August 2011; 935 (2008). Astron. Soc. 323, 904 (2001). 10.1126/science.1209069 strated high-functionality active piezoelectric de- Giant Piezoelectricity on Si for vices incorporating PZT on Si (3–6). Relaxor ferroelectrics with engineered do- Hyperactive MEMS main states have dramatically enhanced piezo- response over PZT. Pb(Mg1/3Nb2/3)O3-PbTiO3 S. H. Baek,1 J. Park,2 D. M. Kim,1 V. A. Aksyuk,3 R. R. Das,1 S. D. Bu,1 D. A. Felker,4 J. Lettieri,5 (PMN-PT), one of the lead-based relaxor ferro- V. Vaithyanathan,5 S. S. N. Bharadwaja,5 N. Bassiri-Gharb,5 Y. B. Chen,6 H. P. Sun,6 electrics, exhibits strain levels and piezoelectric C. M. Folkman,1 H. W. Jang,1 D. J. Kreft,2 S. K. Streiffer,7 R. Ramesh,8 X. Q. Pan,6 coefficients that can be 5 to 10 times those of S. Trolier-McKinstry,5 D. G. Schlom,5,9,10 M. S. Rzchowski,3 R. H. Blick,2 C. B. Eom1* bulk PZT ceramics and has a large electrome- chanical coupling coefficient of k33 ~ 0.9 (7, 8). Microelectromechanical systems (MEMS) incorporating active piezoelectric layers offer integrated Performance of piezoelectric MEMS could be actuation, sensing, and transduction. The broad implementation of such active MEMS has long enhanced dramatically by incorporating these been constrained by the inability to integrate materials with giant piezoelectric response, such as materials. This requires the integration of PMN- Pb(Mg1/3Nb2/3)O3-PbTiO3 (PMN-PT). We synthesized high-quality PMN-PT epitaxial thin films on PT films with Si, while preserving their enhanced vicinal (001) Si wafers with the use of an epitaxial (001) SrTiO3 template layer with superior piezoelectric properties. The piezoelectric proper- piezoelectric coefficients (e31,f = –27 T 3 coulombs per square meter) and figures of merit for ties of ferroelectrics depend on both intrinsic (stoi- piezoelectric energy-harvesting systems. We have incorporated these heterostructures into microcantilevers that are actuated with extremely low drive voltage due to thin-film piezoelectric 1 Department of Materials Science and Engineering, University properties that rival bulk PMN-PT single crystals. These epitaxial heterostructures exhibit very large of Wisconsin, Madison, WI 53706, USA. 2Department of Elec- electromechanical coupling for ultrasound medical imaging, microfluidic control, mechanical trical and Computer Engineering, University of Wisconsin, Madison, WI 53706, USA. 3Center for Nanoscale Science and sensing, and energy harvesting. Technology, National Institute of Standards and Technology, Gaithersburg, MD 20899, USA. 4Department of Physics, Uni- ilicon is the gold standard for microelectronic transform electrical energy into mechanical en- versity of Wisconsin, Madison, WI 53706, USA. 5Department of S devices as well as for micro-electromechanical systems (MEMS), which are electrically driven mechanical devices ranging in size from a ergy and vice versa through their linear electro- mechanical coupling effect. The highest-performing piezoelectric MEMS heterostructures, including Materials Science and Engineering, Penn State University, University Park, PA 16802, USA. 6Department of Materials Science and Engineering, University of Michigan, Ann Arbor, MI 48109, USA. 7Center for Nanoscale Materials, Argonne Na- micrometer to a few millimeters. However, the energy-harvesting devices and actuator struc- tional Laboratory, Argonne, IL 60439, USA. 8Department of main drawback in the world of MEMS is that tures, have been fabricated with PZT piezoelec- Materials Science and Engineering, University of California, Si is a passive material; that is, it takes metallic tric layers. A. K. Sharma et al. demonstrated Berkeley, CA 94720, USA. 9Department of Materials Science and Engineering, Cornell University, Ithaca, NY 14853–1501, electrodes to capacitively displace MEMS. On the integration of epitaxial PZT thin films on Si USA. 10Kavli Institute at Cornell for Nanoscale Science, Ithaca, the other hand, active piezoelectric materials such substrates with the use of a SrTiO3/MgO/TiN buf- NY 14853, USA. as lead zirconate titanate (PZT) enable mechan- fer layer (4). Subsequent work further enhanced *To whom correspondence should be addressed. E-mail: ical displacement (1–3). Piezoelectric materials PZT thin-film piezoelectric properties and demon- firstname.lastname@example.org 18 NOVEMBER 2011 VOL 334 SCIENCE www.sciencemag.org