Strongly Time-Variable Ultra-Violet Metal Line Emission from the Circum-Galactic Medium of High-Redshift Galaxies
N. Sravan, C.-A. Faucher-Giguere, F. van de Voort, D. Keres, A. L. Muratov, P. F. Hopkins, R. Feldmann, E. Quataert, N. Murray
MMon. Not. R. Astron. Soc. , 000–000 (0000) Printed 19 September 2018 (MN L A TEX style file v2.2)
Strongly Time-Variable Ultra-Violet Metal Line Emissionfrom the Circum-Galactic Medium of High-RedshiftGalaxies
Niharika Sravan ∗ , Claude-Andr´e Faucher-Gigu`ere , Freeke van de Voort , ,Duˇsan Kereˇs , Alexander L. Muratov , Philip F. Hopkins , Robert Feldmann ,Eliot Quataert , and Norman Murray , Center for Interdisciplinary Exploration and Research in Astrophysics (CIERA), andDepartment of Physics and Astronomy, Northwestern University, 2145 Sheridan Road, Evanston IL 60208, USA Department of Astronomy and Theoretical Astrophysics Center, University of California, Berkeley, CA 94720-3411, USA Academia Sinica Institute of Astronomy and Astrophysics, PO Box 23-141, Taipei 10617, Taiwan Department of Physics, Center for Astrophysics and Space Science, University of California, San Diego, 9500 Gilman Drive,La Jolla, CA 9209, USA TAPIR, Mailcode 350-17, California Institute of Technology, Pasadena, CA 91125, USA Canadian Institute for Theoretical Astrophysics, 60 St. George Street, University of Toronto, ON M5S 3H8, Canada Canada Research Chair in Astrophysics
19 September 2018
ABSTRACT
We use cosmological simulations from the Feedback In Realistic Environments (FIRE)project, which implement a comprehensive set of stellar feedback processes, to studyultra-violet (UV) metal line emission from the circum-galactic medium of high-redshift( z = 2 −
4) galaxies. Our simulations cover the halo mass range M h ∼ × − . × M (cid:12) at z = 2, representative of Lyman break galaxies. Of the transitions weanalyze, the low-ionization C III (977 ˚A) and Si III (1207 ˚A) emission lines are the mostluminous, with C IV (1548 ˚A) and Si IV (1394 ˚A) also showing interesting spatially-extended structures. The more massive halos are on average more UV-luminous. TheUV metal line emission from galactic halos in our simulations arises primarily fromcollisionally ionized gas and is strongly time variable, with peak-to-trough variationsof up to ∼ Key words: galaxies: formation – galaxies: evolution – galaxies: high-redshift –galaxies: haloes – intergalactic medium – cosmology: theory
In the cosmic web, galaxies form in dark matter over-densities. In order to sustain star formation across cosmictime, they continuously accrete their baryonic fuel from thesurrounding intergalactic medium along with dark matter ∗ [email protected] (IGM; Kereˇs et al. 2005; Prochaska & Wolfe 2009; Bauer-meister et al. 2010; Faucher-Gigu`ere et al. 2011). Some ofthis gas is later returned to the circum-galactic medium(CGM), along with metals produced in stars, via powerfulgalactic winds driven by feedback from stars (e.g. Aguirreet al. 2001; Springel & Hernquist 2003; Oppenheimer & Dav´e2006; Martin et al. 2010; Steidel et al. 2010; Rubin et al.2010) and active galactic nuclei (AGN; e.g., Fabjan et al. c (cid:13) a r X i v : . [ a s t r o - ph . GA ] A ug Sravan et al. Integral field measurements of emission lines are com-plementary to absorption line measurements. In particular,they allow us to directly obtain a 3D picture of the dis-tribution of gas in the CGM by combining the 2D map ofemission on the sky with line-of-sight velocity information.In recent years, we have indeed learned a great deal aboutspatially-resolved kinematics of high-redshift galaxies usingthis technique (e.g., Genzel et al. 2008; F¨orster Schreiber etal. 2009; Wright et al. 2009; Law et al. 2009). Additionally,since gas emissivity scales with the square of density, emis-sion line studies preferentially probe the dense gas closer togalaxies. To date, it has generally not been possible to make3D emission line maps of the CGM of galaxies because halogas is much fainter than galactic gas, thus pushing the capa-bilities of existing astronomical instruments. The situationis however changing thanks to efforts to build sensitive inte-gral field spectrographs (IFS) that are well suited to detectthe low surface brightness features expected from inflowsand outflows around galaxies.Thanks to the abundance of hydrogen, H I Lyman- α (Ly α ) is the most easily detectable line from the CGM.Our first glimpses of the CGM emission have come fromspatially-extended Ly α sources known as “Ly α blobs”(LABs). The blobs have line luminosities up to ∼ ergs − and spatial extents sometimes exceeding 100 proper kpc(Steidel et al. 2000; Matsuda et al. 2004; Yang et al. 2009).The physical nature of LABs is not yet well understood butthere is growing evidence that they are often powered byan energetic source such as an AGN or a starburst galaxy(which can be beamed away from the line of sight and ap-pear obscured; Geach et al. 2009, 2014; Prescott et al. 2015).AGN and star-forming galaxies can induce the CGM to glowthrough several mechanisms. Ly α photons can be producedinside galaxies by the processing of ionizing photons thatare absorbed by the interstellar medium (ISM) and scatter With the advent of 30-m class optical telescopes, it will be-come possible to sample galaxy halos more densely by using or-dinary galaxies as background sources (e.g., Steidel et al. 2010).With sufficiently dense sampling, absorption line measurementscan produce 3D tomographic maps of the foreground gas.
Table 1.
Parameters of the Simulations Analyzed in thisWork
Name M z =2h m b (cid:15) b m dm (cid:15) dm M (cid:12) M (cid:12) pc M (cid:12) pc m12v × ×
10 2.0 × m12q × ×
10 2.8 × m12i × ×
14 2.8 × m13 × ×
20 2.3 × z2h830 × × × z2h650 × × × z2h600 × × × z2h550 × × × z2h506 × × × z2h450 × × × z2h400 × × × z2h350 × × × MFz2 B1 × × × (1) Name: Simulation designation. Simulations mx havea main halo mass ∼ x M (cid:12) at z = 0. (2) M z =2h : Mass of the main halo at z = 2. (3) m b : Initial baryonic (gas and star) particle mass inthe high-resolution region. (4) (cid:15) b : Minimum baryonic force softening (fixed inphysical units past z ∼
10; minimum SPH smoothinglengths are comparable or smaller). Force softeninglengths are adaptive (mass resolution is fixed). in a diffuse halo as they escape the galaxy. Diffuse Ly α halosare now inferred to be generically produced by ordinary star-forming galaxies by this mechanism (Steidel et al. 2011). Ion-izing photons that escape galaxies but are absorbed in theCGM can also produce fluorescent Ly α emission (e.g., Gould& Weinberg 1996; Cantalupo et al. 2005; Kollmeier et al.2010). Alternatively, energy can be injected in the CGM asgalactic winds driven by stellar or AGN feedback encounterhalo gas. This energy can then power Ly α emission fromthe CGM (Taniguchi & Shioya 2000; Taniguchi et al. 2001).We indeed show in this paper that galactic winds can powersignificant metal line emission from circum-galactic gas, andthat this process induces dramatic time variability. Finally,spatially-extended Ly α emission can be powered by gravi-tational potential energy that is dissipated as gas falls intodark matter halos (Haiman et al. 2000; Fardal et al. 2001;Dijkstra & Loeb 2009; Goerdt et al. 2010; Rosdahl & Blaizot2012). Calculations that take into account self-shielding ofdense gas suggest however that gravity-driven Ly α emissionis in general too faint to explain the most luminous LABs(Faucher-Gigu`ere et al. 2010).In spite of a growing body of high-quality observationsof spatially extended Ly α at high-redshift (e.g., Rauch et al.2008; Hennawi et al. 2009; Martin et al. 2014; Cantalupoet al. 2014), these observations have proved difficult to in-terpret because of the strong radiative transfer effects expe-rienced by Ly α photons as they are scattered by interstel-lar and circum-galactic gas (e.g., Dijkstra et al. 2006; Ver-hamme et al. 2006; Faucher-Gigu`ere et al. 2010). In effect,scattering scrambles the photons both spatially and spec-trally, making it challenging to reliably infer the geometryand kinematics of the emitting gas. c (cid:13)000
10; minimum SPH smoothinglengths are comparable or smaller). Force softeninglengths are adaptive (mass resolution is fixed). in a diffuse halo as they escape the galaxy. Diffuse Ly α halosare now inferred to be generically produced by ordinary star-forming galaxies by this mechanism (Steidel et al. 2011). Ion-izing photons that escape galaxies but are absorbed in theCGM can also produce fluorescent Ly α emission (e.g., Gould& Weinberg 1996; Cantalupo et al. 2005; Kollmeier et al.2010). Alternatively, energy can be injected in the CGM asgalactic winds driven by stellar or AGN feedback encounterhalo gas. This energy can then power Ly α emission fromthe CGM (Taniguchi & Shioya 2000; Taniguchi et al. 2001).We indeed show in this paper that galactic winds can powersignificant metal line emission from circum-galactic gas, andthat this process induces dramatic time variability. Finally,spatially-extended Ly α emission can be powered by gravi-tational potential energy that is dissipated as gas falls intodark matter halos (Haiman et al. 2000; Fardal et al. 2001;Dijkstra & Loeb 2009; Goerdt et al. 2010; Rosdahl & Blaizot2012). Calculations that take into account self-shielding ofdense gas suggest however that gravity-driven Ly α emissionis in general too faint to explain the most luminous LABs(Faucher-Gigu`ere et al. 2010).In spite of a growing body of high-quality observationsof spatially extended Ly α at high-redshift (e.g., Rauch et al.2008; Hennawi et al. 2009; Martin et al. 2014; Cantalupoet al. 2014), these observations have proved difficult to in-terpret because of the strong radiative transfer effects expe-rienced by Ly α photons as they are scattered by interstel-lar and circum-galactic gas (e.g., Dijkstra et al. 2006; Ver-hamme et al. 2006; Faucher-Gigu`ere et al. 2010). In effect,scattering scrambles the photons both spatially and spec-trally, making it challenging to reliably infer the geometryand kinematics of the emitting gas. c (cid:13)000 , 000–000 V Emission from the CGM of High-z Galaxies Emission from metal lines in the CGM can be pow-ered by the same physical processes as Ly α . Because met-als are not as abundant, metal line emission is generallyfainter and more difficult to detect than Ly α . Observingmetal lines however provides valuable complementary infor-mation, and in some respects provide a more direct windowinto the physics of circum-galactic gas flows. Most metallines are optically thin and therefore not subject to signif-icant photon scattering, unlike Ly α . Furthermore, differentmetal ions probe different temperature regimes, allowing usto construct a more complete physical picture. In particu-lar, ions with high ionization potential preferentially probemore diffuse, volume-filling gas whereas Ly α emission tendsto peak around cold filaments (e.g., Goerdt et al. 2010;Faucher-Gigu`ere et al. 2010; Rosdahl & Blaizot 2012). Sincemuch of the cooling of the diffuse Universe occurs throughrest-frame ultra-violet (UV) emission (e.g., Bertone et al.2013), rest-UV metal lines offer a powerful way to probegalaxy formation. Rest-frame UV from z ∼ −
4, coveringthe peak of the cosmic star formation history (e.g., Bouwenset al. 2007), also has the benefit of being redshifted into theoptical and thus is accessible using large optical telescopes.A number of optical integral field spectrographs (IFS)with the capacity to detect low surface brightness, redshiftedrest-UV CGM emission have recently been commissioned orare planned for the near future. The Cosmic Web Imager(CWI; Matuszewski et al. 2010) started taking data on theHale 200” telescope at Palomar Observatory in 2009 andthe first science results on luminous spatially-extended Ly α sources at z ∼ − (KCWI, Martin et al. 2010), to be mounted on theKeck II telescope at the W. M. Keck Observatory on MaunaKea, is currently being developed. The Multi Unit Spectro-scopic Explorer (MUSE; Bacon et al. 2010) on the VeryLarge Telescope (VLT) completed its commissioning in Au-gust 2014 and early science results are being reported (e.g.,Fumagalli et al. 2014; Bacon et al. 2015; Richard et al. 2015;Wisotzki et al. 2015). MUSE combines a wide field of view,excellent spatial resolution, and a large spectral range. TheseIFSs not only provide kinematic information not availablewith narrowband imaging but also enable more accuratebackground subtraction to characterize low surface bright-ness features.Cosmological simulations have previously been used topredict UV emission from the IGM and the CGM (e.g.,Furlanetto et al. 2004; Bertone et al. 2010; Frank et al.2012; Bertone & Schaye 2012; van de Voort & Schaye 2013).Our simulations, from the FIRE (Feedback In Realistic En-vironments) project, improve on these previous analysesin several respects. Most of our zoom-in simulations havea gas particle mass of a few times 10 M (cid:12) and a mini-mum (adaptive) gas gravitational softening length of ∼ . × M (cid:12) anda maximum gas gravitational softening length of 700 proper http://muse.univ-lyon1.fr See the FIRE project web site at: http://fire.northwestern.edu. pc (see also Bertone & Schaye 2012). The high resolutionof our zoom-ins allows us to resolve the main structures inthe ISM of galaxies. Our simulations also explicitly treatstellar feedback from supernovae of Types I and II, stellarwinds from young and evolved stars, photoionization, andradiation pressure on dust grains.Our suite of simulations has been shown to successfullyreproduce the observationally-inferred relationship betweenstellar mass and dark matter halo mass (the M (cid:63) − M h rela-tion Hopkins et al. 2014) and the mass-metallicity relations(Ma et al. 2016) of galaxies below ∼ L (cid:63) at all redshifts whereobservational constraints are currently available, as well asthe covering fractions of dense HI in the halos of z = 2 − ad hoc stellarfeedback implementations, our simulations follow hydrody-namical interactions and gas cooling at all times, lendingsome credence to the phase structure of galactic winds pre-dicted in our calculations.Our primary goal in this paper is to investigate the im-plications of the high resolution and explicit stellar feedbackmodels of the FIRE simulations for UV metal line emissionfrom the CGM of high-redshift galaxies and its physical ori-gin. Our analysis is guided by the UV metal lines most likelyto be detectable by MUSE and KCWI, though it is beyondthe scope of this paper to make detailed observational pre-dictions for specific instruments.This paper is organized as follows. We describe our sim-ulations in more detail and our methodology for comput-ing metal line emission in §
2. We present our main resultsin §
3, including a discussion of the strong predicted stellarfeedback-driven UV metal line time variability. We summa-rize our conclusions in §
4. The Appendices contain a con-vergence test and local source test.
Here we summarize the numerical methods used in our sim-ulations and the physics included. For more details on thealgorithms, we refer the reader to Hopkins et al. (2014) andreferences therein.Our simulations were run with the GIZMO code (Hop-kins 2014) in “P-SPH” mode. P-SPH, described in Hopkins(2013), is a Lagrangian-based pressure-entropy formulationof the smooth particle hydrodynamics (SPH) equations thateliminates some of the well-known differences between tra-ditional implementations of SPH and grid-based codes (e.g.,Agertz et al. 2007; Sijacki et al. 2012) while preserving theexcellent conservation properties of SPH. GIZMO derivesfrom Gadget-3 (last described in Springel 2005) and its treeparticle-mesh (TreePM) gravity solver is based on the lat-ter. Our GIZMO runs include improved algorithms and pre-scriptions for artificial viscosity (Cullen & Dehnen 2010),entropy mixing (Price 2008), adaptive time stepping (Durier c (cid:13) , 000–000 Sravan et al.
Figure 1.
Theoretical surface brightness maps for the UV metal lines studied in this work in z2h506 in a region with side length equalto a virial radius at z = 2 (top), 3 (middle), and 4 (bottom). The halo mass in log M h (M (cid:12) ) is 12.06 ( z = 2), 11.36 ( z = 3), and 11.11( z = 4). From left to right, the vertical panels at each redshift show the emission for the C III, C IV, N V, O VI, Si III, and Si IV lineslisted in Table 2. The dashed lines show the virial radius in each panel. The solid white contours enclose regions with SB > − ergs − cm − arcsec − , a proxy for the SB detectable in (deep but non-stacked) MUSE and KCWI observations. All the emission linesshown would redshift into either the MUSE or KCWI spectral ranges, except C III and O VI at z = 2 (indicated by dashed red bordersaround the panels). The observed surface brightness for the CIII, SiIII, and OVI lines is subject to IGM attenuation (10-20%, 30-50%,and factor 2-5 effects at z = 2 , , and 4, respectively; § & Dalla Vecchia 2012), the smoothing kernel (Dehnen & Aly2012), and adaptive gravitational softening (Price & Mon-aghan 2007; Barnes 2012). Star formation in the FIRE sim-ulations only occurs in molecular, self-gravitating gas with n H (cid:38) −
50 cm − . The energy, momentum, mass, andmetal yields from photo-ionization, radiation, stellar windsand supernovae are calculated using STARBURST99 (Lei-therer et al. 1999). The abundances of nine metallic species(C, N, O, Ne, Mg, Si, S, Ca, and Fe) are also trackedusing this method and used to evaluate cooling rates us-ing a method similar to Wiersma et al. (2009). Ionizationbalance of these elements are computed assuming the cos-mic ionizing background of Faucher-Gigu`ere et al. (2009)(FG09). Self-shielding of hydrogen is accounted for with alocal Jeans-length approximation (integrating the local den-sity at a given particle out to a Jeans length to determine asurface density Σ), then attenuating the background seen atthat point by e − κ Σ (where κ is the opacity). Confirmationof the accuracy of this approximation in radiative transferexperiments can be found in Faucher-Gigu`ere et al. (2010).The simulations analyzed in this work are listed in Ta-ble 1. The initial conditions for m13 , m12q , and m12i were Publicly available at http://galaxies.northwestern.edu/uvb. chosen to match ones from the AGORA project (Kim et al.2014). The initial conditions for m12v are the same as the‘B1’ run studied by Kereˇs & Hernquist (2009) and Faucher-Gigu`ere & Kereˇs (2011). The z2hxxx series of simulations,first described in Faucher-Gigu`ere et al. (2015), consists ofhalos in the mass range M h = 1 . × M (cid:12) − . × M (cid:12) at z = 2. These halos are representative of ones hostingLBGs at that redshift, though somewhat on the low-massend of the mass distribution (LBGs at z ∼ M h ∼ M (cid:12) ; Adelbergeret al. 2005; Trainor & Steidel 2012). MFz2 B1 is part ofthe MassiveFIRE simulation suite (Feldmann et al. in prep).Our simulations do not include feedback from AGN. Whilesuch feedback may be potentially important, especially inthe more massive halos, the proper modeling of AGN feed-back is still a major theoretical challenge (e.g., Somerville &Dav´e 2014) and we have decided to postpone its treatmentto future work. We focus our analysis on main halos (i.e.,we do not center on satellite halos). The surface brightnessprofiles of UV line emission may, however, include emissioncontributed by satellite galaxies. Halos are identified usingAmiga’s Halo Finder (AHF; Knollmann & Knebe 2009) andwe adopt the virial overdensity definition of Bryan & Nor-man (1998). c (cid:13)000
50 cm − . The energy, momentum, mass, andmetal yields from photo-ionization, radiation, stellar windsand supernovae are calculated using STARBURST99 (Lei-therer et al. 1999). The abundances of nine metallic species(C, N, O, Ne, Mg, Si, S, Ca, and Fe) are also trackedusing this method and used to evaluate cooling rates us-ing a method similar to Wiersma et al. (2009). Ionizationbalance of these elements are computed assuming the cos-mic ionizing background of Faucher-Gigu`ere et al. (2009)(FG09). Self-shielding of hydrogen is accounted for with alocal Jeans-length approximation (integrating the local den-sity at a given particle out to a Jeans length to determine asurface density Σ), then attenuating the background seen atthat point by e − κ Σ (where κ is the opacity). Confirmationof the accuracy of this approximation in radiative transferexperiments can be found in Faucher-Gigu`ere et al. (2010).The simulations analyzed in this work are listed in Ta-ble 1. The initial conditions for m13 , m12q , and m12i were Publicly available at http://galaxies.northwestern.edu/uvb. chosen to match ones from the AGORA project (Kim et al.2014). The initial conditions for m12v are the same as the‘B1’ run studied by Kereˇs & Hernquist (2009) and Faucher-Gigu`ere & Kereˇs (2011). The z2hxxx series of simulations,first described in Faucher-Gigu`ere et al. (2015), consists ofhalos in the mass range M h = 1 . × M (cid:12) − . × M (cid:12) at z = 2. These halos are representative of ones hostingLBGs at that redshift, though somewhat on the low-massend of the mass distribution (LBGs at z ∼ M h ∼ M (cid:12) ; Adelbergeret al. 2005; Trainor & Steidel 2012). MFz2 B1 is part ofthe MassiveFIRE simulation suite (Feldmann et al. in prep).Our simulations do not include feedback from AGN. Whilesuch feedback may be potentially important, especially inthe more massive halos, the proper modeling of AGN feed-back is still a major theoretical challenge (e.g., Somerville &Dav´e 2014) and we have decided to postpone its treatmentto future work. We focus our analysis on main halos (i.e.,we do not center on satellite halos). The surface brightnessprofiles of UV line emission may, however, include emissioncontributed by satellite galaxies. Halos are identified usingAmiga’s Halo Finder (AHF; Knollmann & Knebe 2009) andwe adopt the virial overdensity definition of Bryan & Nor-man (1998). c (cid:13)000 , 000–000 V Emission from the CGM of High-z Galaxies Figure 2.
Same as in Figure 1 but for the more massive halo
MFz2 B1 . The halo mass in log M h (M (cid:12) ) is 12.93 ( z = 2), 12.52( z = 3), and 12.21 ( z = 4). All our simulations assume a “standard” flat ΛCDMcosmology with h ≈ .
7, Ω m = 1 − Ω Λ ≈ .
27 andΩ b ≈ . Our method for calculating gas emissivities is similar to theone used by van de Voort & Schaye (2013).Following previous work that identified the most im-portant contributions to rest-UV emission from the high-redshift CGM (Bertone & Schaye 2012; van de Voort &Schaye 2013), we compute emissivities for UV doublets C IV (1548 ˚A, 1551 ˚A), Si IV (1394 ˚A, 1403 ˚A), N V (1239 ˚A, 1243˚A), and O VI (1032 ˚A, 1038 ˚A), and the singlets Si III (1207˚A) and C
III (977 ˚A). We do not show predictions for theNe
VIII (770 ˚A, 780 ˚A) doublet because its intrinsic CMGluminosity is ∼ ≈
2. There-fore, the total intensity for the doublet (e.g., as would bemeasured by an instrument that does not spectrally resolve
Table 2.
UV Lines Analyzed in this Work
MUSE KCWIIon Name λ λ z min z max z min z max ˚A ˚AC IV 1548 1551 2.00 5.01 1.26 5.78Si IV 1394 1403 2.34 5.67 1.51 6.53N V 1239 1243 2.75 6.51 1.82 7.47O VI 1032 1038 3.51 8.01 2.39 9.17Si III 1207 - 2.85 6.71 1.90 7.70C III 977 - 3.76 8.52 2.58 9.75 (1) λ : Rest-frame wavelength of the emission line orthe stronger line for doublets. (2) λ : Rest-frame wavelength of the weaker line fordoublets. (3) z min : The minimum redshift from which thestronger transition line lies within the wavelengthcoverage of MUSE and KCWI. (4) z max : The maximum redshift from which thestronger transition line lies within the wavelengthcoverage of MUSE and KCWI.Note: The wavelength coverage of MUSE and KCWIare 4650˚A – 9300˚A and 3500˚A – 10,500˚A, respectively. the two transitions) can be obtained by multiplying the in-tensities predicted by a factor of ≈ .
5. Table 2 lists theemission lines we study and the minimum and maximumredshifts at which they lie within the wavelength coverageof MUSE and KCWI. We first pre-compute grids of lineemissivities as a function of gas temperature and hydrogen c (cid:13) , 000–000 Sravan et al. number density over the redshift interval z = 2 − CLOUDY (version 13.03 of the code last described by Fer-land et al. 2013). The grids sample temperatures in the range10 < T < . K in intervals of ∆ log T = 0 .
05 and hydro-gen number densities in the range 10 − < n H < cm − inintervals of ∆ log n H = 0.2. The grids are computed at red-shift intervals of ∆ z = 0 .
1. For the grids, the gas is assumedto be of solar metallicity, optically thin and in photoionisa-tion equilibrium with the cosmic ionizing background (thecontribution of collisional ionization to equilibrium balanceis automatically included). In computing emissivities fromour simulations, we bi-linearly interpolate the pre-computedgrids in logarithmic space and scale linearly with the actualmetallicity of the gas. For most of our calculations, we usethe cosmic ionizing background model of Faucher-Gigu`ereet al. (2009); we show in § In order to make full use of the spatially-adaptive resolu-tion of our Lagrangian SPH simulations, we evaluate lineluminosities from the particle data directly. For a given gasparticle, the luminosity of each line listed in Table 2 is cal-culated as L part = (cid:15) (cid:12) ( z, n H , T ) (cid:18) m gas ρ gas (cid:19) (cid:18) Z gas Z (cid:12) (cid:19) , (1)where m gas , ρ gas , and Z gas are the mass, density and metal-licity of the gas particle, respectively, and (cid:15) (cid:12) ( z, n H , T ) isthe emissivity interpolated from the pre-computed solar-metallicity grid. This approach for evaluating emissivitiesis based on total gas metallicity and assumes that the lineemitting gas has solar abundance ratios. Our SPH simula-tions do not include a model for metal diffusion by unre-solved turbulence (e.g., Shen et al. 2010). While this couldpotentially cause the metals in the simulations to be tooclumped and artificially enhance our predicted luminosities,our simulations do capture the mixing of metals due to re-solved gas flows. In Appendix A, we present a convergencetest showing that our predicted luminosities do not increasesignificantly with increasing resolution (and hence increasedmetal mixing).To produce surface brightness profiles, we assign parti-cle luminosities to 3D Cartesian grids. The flux from eachgrid cell is given by F cell = L cell πd , (2)where d L is the luminosity distance and L cell = (cid:80) L part isthe sum of the particle luminosities assigned to the cell. Thecell side length of the grids is 1 proper kpc for all simula-tions. This introduces smoothing on that spatial scale (cor-responding to 0.1 arcsec at z = 2) but does not bias the sur-face brightness calculations since luminosities are computed from the un-degraded particle-carried information. Finally,the surface brightness (SB) projected on the sky is calcu-lated by dividing the flux from each grid cell by the solidangle Ω cell it subtends and summing along the line of sight(los): SB = (cid:88) los F cell Ω cell . (3)In the small angle limit and for cubical grid cells, Ω cell =( a cell /d A ) , where a cell is the proper side length of a gridcell and d A is the angular diameter distance. Absorption by intervening intergalactic gas affects the de-tectability of the emission lines we predict (e.g., Madau1995). The emission lines we analyze in this paper all haverest wavelength λ >
912 ˚A and so are not affected by Ly-man continuum opacity. The CIII, SiIII, and OVI emis-sion lines that we consider however have rest wavelengths λ < e − τ for theCIII 977 ˚A, SiIII 1207 ˚A, and OVI 1032 ˚A lines. For thesethree lines, we find transmission factors (0 . , . , .
9) at z = 2, (0 . , . , .
7) at z = 3, and (0 . , . , .
4) at z = 4,respectively. Thus, IGM attenuation has a small effect on ourpredictions at z = 2. Even out to z = 4, IGM attenuationis only a factor ∼ § z = 4, the IGM may suppress observable SiIII emission bya factor of ∼
5, which will make this line more challengingto detect at such redshifts. The CIV, NV, and SIV lines weanalyze do not suffer from Lyman-series absorption and soare not significantly affected by hydrogen absorption fromthe IGM. To allow readers to easily correct our results forother IGM attenuation models (proximity effects could beimportant for SiIII 1207 ˚A, which has a rest wavelengthclose to Ly α Figures 1 and 2 show surface brightness maps for the UVmetal lines in Table 2 for z2h506 and
MFz2 B1 , respec-tively. These are our two most massive halos at z = 2. Inthese maps, solid white contours enclose regions with SB > − erg s − cm − arcsec − , a proxy for the SB de-tectable in (deep but non-stacked) MUSE and KCWI obser-vations. High-redshift Ly α CGM emission has been detectedin narrow-band imaging observations down to ∼ − ergs − cm − arcsec − in individual objects (Steidel et al. 2000;Matsuda et al. 2004) and down to ∼ − erg s − cm − arcsec − in stacks (Steidel et al. 2011). Current and upcom-ing IFSs such as MUSE and KCWI are expected to reach anorder of magnitude deeper. For reference, SB = 10 − erg c (cid:13)000
MFz2 B1 , respec-tively. These are our two most massive halos at z = 2. Inthese maps, solid white contours enclose regions with SB > − erg s − cm − arcsec − , a proxy for the SB de-tectable in (deep but non-stacked) MUSE and KCWI obser-vations. High-redshift Ly α CGM emission has been detectedin narrow-band imaging observations down to ∼ − ergs − cm − arcsec − in individual objects (Steidel et al. 2000;Matsuda et al. 2004) and down to ∼ − erg s − cm − arcsec − in stacks (Steidel et al. 2011). Current and upcom-ing IFSs such as MUSE and KCWI are expected to reach anorder of magnitude deeper. For reference, SB = 10 − erg c (cid:13)000 , 000–000 V Emission from the CGM of High-z Galaxies Figure 3.
Median (solid lines) and mean (dotted lines) UV metal line theoretical radial surface brightness profiles for the sample ofLyman break galaxy simulations described in § MFz2 B1 ) at z = 2 (left), 3 (middle), and 4 (right), assuming the FG09UV/X-ray ionizing background model (top), the galaxy+quasar HM05 UV/X-ray background scaled by a factor of 0.447 (to match thehydrogen photoionization rate of the FG09 model at z = 3; middle), and the FG09 background multiplied by a factor of 10 (bottom).The mean and median halo masses in log M h (M (cid:12) ) are 11.79 and 11.82 ( z = 2), 11.44 and 11.40 ( z = 3), and 11.11 and 11.10 ( z = 4),respectively. The vertical dotted, dot-dashed, dashed and solid grey lines indicate radii of 5, 10, 20, and 50 proper kpc for median-masshalos at each redshift. Surface brightness profiles shown with thinner curves correspond to emission lines that do not redshift into theMUSE or KCWI bands (indicated by the dashed red border panels in Figs. 1 and 2). The observed surface brightness for the CIII, SiIII,and OVI lines is subject to IGM attenuation (10-20%, 30-50%, and factor 2-5 effects at z = 2 , , and 4, respectively; § s − cm − arcsec − is the 1 σ detection threshold for the az-imuthally averaged Ly α radial profiles in the recent ultra-deep exposure of the Hubble Deep Field South obtainedwith MUSE (Wisotzki et al. 2015). Comparing Figures 1( M h = 1 . × M (cid:12) at z = 2) and 2 ( M h = 8 . × M (cid:12) at z = 2) suggests a significant halo mass dependence for thedetectability of metal UV lines, with the more massive haloshowing more spatially extended and more luminous emis-sion (note the different spatial scales). In § on av-erage .For MFz2 B1 , the C III and Si III lines are predicted to produce emission above 10 − erg s − cm − arcsec − onspatial scales ∼
100 proper kpc (in the elongated direction)at z = 2, and C IV and Si IV should be detectable at thissurface brightness on scales ∼
50 proper kpc. For z2h506 ,we predict similar structures but with spatial extent smallerby approximately the ratio of the virial radius in z2h506 relative to that of
MFz2 B1 (a factor of ≈ . z = 2).These transitions preferentially probe relatively cool, T ∼ . − K gas. All these lines would redshift into eitherthe MUSE or KCWI spectral bands if observed from z =2 −
4, except C III at z ∼
2. In Figure 3, we show themedian and mean radial surface brightness profiles for all To calculate the average mean and median radial profiles, wefirst compute the cylindrically averaged profiles for individual ha-c (cid:13) , 000–000
Sravan et al.
Figure 4.
Cylindrically averaged UV metal line radial theoretical surface brightness profiles for halos
MFz2 B1 (top), z2h506 (middle),and z2h450 (bottom) at z = 2 (left), 3 (middle), and 4 (right). Values of log M h ( M (cid:12) ) and R vir (proper kpc) for each halo are indicatedat the top right of each panel. The vertical dotted, dot-dashed, dashed and solid grey lines indicate radii of 5, 10, 20, and 50 proper kpc,respectively. Surface brightness profiles shown with thinner curves correspond to emission lines that do not redshift into the MUSE orKCWI bands (indicated by the dashed red border panels in Figs. 1 and 2). The observed surface brightness for the CIII, SiIII, and OVIlines is subject to IGM attenuation (10-20%, 30-50%, and factor 2-5 effects at z = 2 , , and 4, respectively; § the halos in our sample but excluding excluding MFz2 B1 (which is of substantially higher mass than the rest). O VI(which preferentially probes warmer gas, T ∼ . K gas)is less luminous than C IV and Si IV by ∼ z = 2. OVI emission may be detectable at higher redshifts in verydeep integrations of massive halos but could prove beyondthe reach of the current generation of instruments. As weshow in § los, then take the mean and median over halos in our sample. Tocompute the cylindrical averages, the radial profiles are radiallybinned in bins of width 3 proper kpc. generally easier to detect at z ∼ z ∼ −
4. Two ef-fects contribute to this. First, our halos grow with time andtheir star formation rates also on average increase with timeover that redshift interval (e.g., Hopkins et al. 2014; Muratovet al. 2015), thus increasing the amount of energy injectedin the CGM and radiated by the UV lines. Second, the sur-face brightness is subject to cosmological surface brightnessdimming, SB ∝ (1 + z ) − . Note that the virial radii of thesimulated halos in Figure 3 increase with decreasing redshiftso that the total luminosities of the z = 2 halos are system-atically higher even though the surface brightness at a fixedfraction of R vir appears to follow a different redshift trend.The different rows in Figure 3 correspond to differentassumptions for the ionizing flux illuminating CGM gas. Thetop row assumes our fiducial FG09 UV/X-ray background.To test the sensitivity to the shape of the ionizing spec-trum, we have repeated our CLOUDY emissivity calcula-tions using the Haardt & Madau 2005 background, which c (cid:13)000
4. Two ef-fects contribute to this. First, our halos grow with time andtheir star formation rates also on average increase with timeover that redshift interval (e.g., Hopkins et al. 2014; Muratovet al. 2015), thus increasing the amount of energy injectedin the CGM and radiated by the UV lines. Second, the sur-face brightness is subject to cosmological surface brightnessdimming, SB ∝ (1 + z ) − . Note that the virial radii of thesimulated halos in Figure 3 increase with decreasing redshiftso that the total luminosities of the z = 2 halos are system-atically higher even though the surface brightness at a fixedfraction of R vir appears to follow a different redshift trend.The different rows in Figure 3 correspond to differentassumptions for the ionizing flux illuminating CGM gas. Thetop row assumes our fiducial FG09 UV/X-ray background.To test the sensitivity to the shape of the ionizing spec-trum, we have repeated our CLOUDY emissivity calcula-tions using the Haardt & Madau 2005 background, which c (cid:13)000 , 000–000 V Emission from the CGM of High-z Galaxies we rescaled by a factor of 0.447 to match the hydrogen pho-toionization rate Γ HI = 5 . × − s − of the FG09 modelat z = 3 (middle row). We also tested the sensitivity ofour results to the magnitude of the ionizing flux by mul-tiplying the FG09 background by a factor of 10 (bottomrow). The mean and average surface brightness profiles arealmost identical between the different rows, at z = 2 , T ∼ . − K, which is generally collisionally heated. We have verifiedthat our simulations predict emissivity-weighted tempera-tures broadly consistent with Figure 6 of van de Voort &Schaye (2013). This implies that self-shielding effects do notsignificantly affect our UV metal line emission, unlike forLy α emission for which self-shielding effects are critical (e.g.,Faucher-Gigu`ere et al. 2010). This experiment also indicatesthat ionizing photons emitted by local galaxies (which wehave neglected) likely would not substantially alter our pre-dictions. Measurements of the fraction of ionizing photonsthat escape the ISM of LBGs at z ∼ f esc ∼
5% (e.g., Shapley et al. 2006). If this fractioncorresponds to the fraction of directions along which ion-izing photons escape relatively unimpeded, and the ISM isopaque to ionizing photons along other directions, most ofthe CGM around star-forming galaxies is not strongly illu-minated by the central galaxy. In Appendix B, we presentanother test explicitly including a local source spectrum (in-cluding X-rays from supernova remnants) that also showsthat our main predictions are not significantly affected bylocal sources (though some details of the line emission atlow surface brightnesses are).To give a sense of how our predictions vary from haloto halo, Figure 4 shows the predicted cylindrically averagedradial surface brightness profiles for a representative sampleof individual halos (
MFz2 B1 , z2h506 , and z2h450 ). Thespikes in the radial profiles (for example, at R/R vir ∼ MFz2 B1 at z = 2) are due to satellite galaxies (see themaps in Figs. 1 and 2).Our predicted average SB profiles shown in the top pan-els of Figure 3 agree with those obtained by van de Voort &Schaye (2013) from the OWLS simulations to within ∼ III in z2h506 and z2h450 at z = 4 differ by ∼ M h ≈ . M (cid:12) at that redshift.We address this in the next section. One of the key predictions of the FIRE simulations withresolved ISM is that the star formation histories of galaxieshave a strong stochastic component and are much more timevariable (e.g., Hopkins et al. 2014; Sparre et al. 2015) than in lower-resolution simulations in which the ISM is modeledwith a sub-grid effective equation of state. Such sub-gridequations of state are standard in current large-volume cos-mological simulations (e.g., Dav´e et al. 2013; Vogelsbergeret al. 2014; Schaye et al. 2015). In particular, a sub-gridequation of state was used in the OWLS simulations ana-lyzed by van de Voort & Schaye (2013) for UV line emissionfrom the CGM. We have previously shown that the timevariability of star formation in the FIRE simulations resultsin time variable galactic winds (Muratov et al. 2015), anincreased cool gas content of galaxy halos (Faucher-Gigu`ereet al. 2015), and can transform dark matter halo cusps intocores in dwarf galaxies (Onorbe et al. 2015; Chan et al.2015). Similar time variability has also been found in otherzoom-in simulations implementing different stellar feedbackmodels (Governato et al. 2010; Guedes et al. 2011; Stin-son et al. 2013; Agertz & Kravtsov 2014), indicating that itis a generic consequence of resolving the ISM of galaxies, inwhich local dynamical time scales can be very short. We nowshow that time-variable star formation also has importantimplications for the observability of CGM gas in emission,and explains why two halos of the same mass and redshiftcan vary in metal UV line luminosity by orders of magni-tude.In Figure 5 we plot the luminosities of UV metal lineswithin 1 R vir (a proxy for emission from central galaxies), for z2h506 and z2h450 , as a function of redshift. We also plotthe star formation rates within 1 R vir and gas mass outflowrates at 0 . R vir in these halos as a function of redshift. Asthe Figure shows, the UV metal line luminosities, star forma-tion rates, and mass outflow rates all exhibit strong and cor-related time variability. In particular, epochs of peak metalline emission coincide closely with massive outflow eventsin the halos (though some outflow events do not result instrongly enhanced UV line emission), with peak-to-troughvariations of up to ∼ ≈
60 Myr, corresponding to the travel timefrom the galaxy to 0 . R vir (Muratov et al. 2015). This in-dicates that while metal UV line emission time variabilityis ultimately driven by star formation time variability, it isthe star formation-driven galactic winds specifically that in-ject energy into the CGM, which is ultimately radiated awayby UV metal lines (among other channels). This finding isconsistent with our conclusion in the previous section thatthe UV metal line emission arises primarily in collisionallyionized gas, with little sensitivity to the ionizing flux. Thekinetic energy carried by an outflow with mass outflow rate˙ M out and velocity v w is˙ E w ≡
12 ˙ M out v (4) ≈ × erg s − (cid:18) ˙ M out
100 M (cid:12) yr − (cid:19) (cid:16) v w
300 km s − (cid:17) / , (5)which is sufficient to power the predicted UV metal lineemission (see Muratov et al. 2015 for measurements of out-flow velocities in our simulations). Figure 5 explains whyhalo z2h506 is much more luminous than halo z2h450 at c (cid:13) , 000–000 Sravan et al.
Figure 5.
Colored lines show UV metal line luminosities within 1 R vir but excluding the inner 10 proper kpc (a proxy for centralgalaxies) in simulated halos z2h506 (top) and z2h450 (bottom) as a function of redshift. Star formation rates within 1 R vir and gasmass outflow rates at 0 . R vir as a function of redshift are plotted as grey and black lines, respectively. The UV metal line luminosities,star formation, and mass outflow rates are all strongly time variable and correlated. Peaks in CGM luminosity correspond more closelywith peaks in mass outflow rates, which follow peaks of star formation with a time delay, indicating that energy injected in the halos bygalactic winds powers the UV metal line emission. z = 4 in spite of being nearly identical in mass: at z ∼ z2h506 experiences a very massive and energetic mass out-flow.Figure 6 shows more quantitatively the correlations be-tween CGM UV metal line luminosities (for the CIII andSiIII lines, which are generally most luminous), star forma-tion rate, and mass outflow rate. Overall, CGM UV lineluminosities correlate positively with mass outflow rate atfixed redshift but no significant correlation with instanta-neous star formation rate is apparent. The relationship withinstantaneous star formation rate is much weaker becauseof the time delay between star formation bursts and out-flows through 0 . R vir . This again indicates that energyinjection from galactic winds is the primary driver of theemission. Despite the correlation, the relationship betweenCGM luminosities and outflow rate is not one-to-one (in some redshift intervals, the relationship even appears in-verted) because the luminosity is not simply a function ofmass outflow rate but also of how the wind energy is dissi-pated as the wind interacts with other CGM gas. At highredshift, the FIRE simulated halos are very dynamic, whichintroduces significant fluctuations in how the wind energy isdissipated. The FIRE simulations predict that star forma-tion rates and outflows both become more time steady atlow redshift ( z (cid:46)
1; Muratov et al. 2015), so a tighter rela-tionship between CGM luminosities and star formation ratemay result at low redshift. Since the metal line emissivityscales with gas metallicity (eq. 1), another potential sourceof time variability is fluctuations in gas metallicity. However,the average wind metallicity in our simulations fluctuates bya factor of only ∼ c (cid:13)000
1; Muratov et al. 2015), so a tighter rela-tionship between CGM luminosities and star formation ratemay result at low redshift. Since the metal line emissivityscales with gas metallicity (eq. 1), another potential sourceof time variability is fluctuations in gas metallicity. However,the average wind metallicity in our simulations fluctuates bya factor of only ∼ c (cid:13)000 , 000–000 V Emission from the CGM of High-z Galaxies Figure 6.
UV metal line luminosities within 1 R vir but excluding the inner 10 proper kpc of halo centers (a proxy for central galaxies)in simulated halos z2h506 (top) and z2h450 (bottom). The luminosities are plotted versus the star formation rate within R vir (left)and versus the mass outflow rate at 0 . R vir . Circle symbols show the CIII luminosity and square symbols show SiIII luminosity. Colorsindicate redshift bins (100 time slices between z = 4 and z = 2 are shown). Overall, CGM UV line luminosities correlate positively withmass outflow rate at fixed redshift but no significant correlation with instantaneous star formation rate is apparent. prep.) so this must be a subdominant effect on the totalorder-of-magnitude time variability.There is some support from observations that starformation-driven outflows inject energy into galaxy halosthat can collisionally excite metal lines. Turner et al. (2015)inferred based on ionization modeling of O VI absorptionlines coincident with low HI columns the presence of a sub-stantial mass of metal-enriched, T > K collisionally ion-ized gas around z ∼ . Motivated by current and upcoming integral field spectro-graphs on 8-10 m-class telescopes, such as MUSE on theVLT and KCWI on Keck, we used cosmological zoom-insimulations from the FIRE project to make predictions forUV metal line emission from the CGM of z = 2 − M h ∼ × − . × M (cid:12) at z = 2, repre-sentative of Lyman break galaxies. For each simulation, wepredicted the emission from the C III (977 ˚A), C IV (1548˚A), Si III (1207 ˚A), Si IV (1394 ˚A), O VI (1032 ˚A), and N V (1239 ˚A) metal lines.Our results can be summarized as follows:(i) Of the transitions we analyze, the low-ionization C III(977 ˚A) and Si III (1207 ˚A) emission lines are the most lumi-nous, with C IV (1548 ˚A) and Si IV (1394 ˚A) also showing c (cid:13) , 000–000 Sravan et al. interesting spatially-extended structures that should be de-tectable by MUSE and KCWI. At z (cid:54)
3, the CIII and SiIIIlines are attenuated by Lyman-series opacity from the in-tergalactic medium by factors (cid:46)
2, but at z = 2 the SiIIIline is expected to be attenuated by a factor ∼
5. The moremassive halos are on average more UV-luminous.(ii) The UV metal line emission from galactic halos inour simulations arises primarily from collisionally ionized gasand is thus weakly sensitive to the ionizing flux, includingionization from local galaxies.(iii) The UV metal line emission from galactic halos in oursimulations is strongly time variable, with peak-to-troughvariations of up to ∼ R vir . Our simulations thus in-dicate that stellar feedback powers UV metal line emissionthrough energy injected in halos by galactic winds.The prediction that UV metal line emission arises ingas collisionally excited in outflows could be tested by com-paring the emission line kinematics with outflow velocitiespredicted by the simulations. In the FIRE simulations, themedian outflow velocity at 0 . R vir scales with the halo cir-cular velocity v c (Muratov et al. 2015). On its own, this testmay prove ambiguous since ∼ v c is also the velocity expectedof gas falling into halos. Furthermore, kinematic measure-ments transverse to galaxies cannot in general distinguishinflows and outflows since it is not known whether the emit-ting gas is in front or behind the galaxy. On the other hand,“down-the-barrel” absorption spectroscopy of galaxies (e.g.,Shapley et al. 2003; Steidel et al. 2010; Martin et al. 2012;Rubin et al. 2014) can unambiguously probe outflows be-cause the absorbing material is known to lie in front of thestars. Thus, correlating outflow kinematics probed by down-the-barrel absorption spectroscopy with circum-galactic UVmetal line emission could help identify the UV metal lineemission with outflows (with the caveat that a significanttime delay may separate peaks in CGM emission and ab-sorption, as well the possible collimation of outflows).In terms of overall luminosity, our average predictionsfor UV metal lines from the CGM of high-redshift galaxiesare broadly consistent with those of van de Voort & Schaye(2013), which were based on the large-volume OWLS sim-ulations. However, the resolved ISM and stellar feedbackphysics implemented in the FIRE simulations is critical tocapture the strong time variability of star formation and theresulting time variability in UV metal line emission. Thestrong predicted time variability of UV metal line emissionhas important implications for upcoming observations: evensome relatively low-mass halos may be detectable in deep ob-servations with current generation instruments. Conversely,flux-limited samples will be biased toward halos whose cen-tral galaxy has recently experienced a strong burst of starformation.Our finding that galactic winds power UV metal lineemission from CGM gas is reminiscent of the “super wind”model for Ly α blobs (Taniguchi & Shioya 2000; Taniguchiet al. 2001) and highlights the potential of detecting halogas in emission for constraining stellar feedback. Our cur-rent calculations do not include the effects of active galactic nuclei (AGN), whose radiative and mechanical interactionswith halo gas also likely contribute substantially to metalUV line emission. It is clearly worthwhile to include the ef-fects of AGN in future calculations. ACKNOWLEDGMENTS
We are grateful to Alex Richings for assistance with thelocal source test. The simulations analyzed in this paperwere run on XSEDE computational resources (allocationsTG-AST120025, TG-AST130039, and TG-AST140023), onthe NASA Pleiades cluster (allocation SMD-14-5189), on theCaltech compute cluster “Zwicky” (NSF MRI award
REFERENCES
Adelberger K. L., Shapley A. E., Steidel C. C., Pettini M.,Erb D. K., Reddy N. A., 2005, ApJ, 629, 636Adelberger K. L., Steidel C. C., Pettini M., Shapley A. E.,Reddy N. A., Erb D. K., 2005, ApJ, 619, 697Agertz O., Kravtsov A. V., 2014, ArXiv e-printsAgertz O., Moore B., Stadel J., Potter D., Miniati F., ReadJ., Mayer L., Gawryszczak A., Kravtsov A., Nordlund˚A., Pearce F., Quilis V., Rudd D., Springel V., Stone J.,Tasker E., Teyssier R., Wadsley J., Walder R., 2007, MN-RAS, 380, 963Aguirre A., Hernquist L., Schaye J., Weinberg D. H., KatzN., Gardner J., 2001, ApJ, 560, 599Bacon et al. 2010, in Society of Photo-Optical Instrumenta-tion Engineers (SPIE) Conference Series Vol. 7735 of So-ciety of Photo-Optical Instrumentation Engineers (SPIE)Conference Series, The MUSE second-generation VLT in-strumentBacon et al. 2015, A&A, 575, A75Barnes J. E., 2012, MNRAS, 425, 1104Bauermeister A., Blitz L., Ma C.-P., 2010, ApJ, 717, 323Bertone S., Aguirre A., Schaye J., 2013, MNRAS, 430, 3292Bertone S., Schaye J., 2012, MNRAS, 419, 780Bertone et al. 2010, MNRAS, 408, 1120Bouwens R. J., Illingworth G. D., Franx M., Ford H., 2007,ApJ, 670, 928Bryan G. L., Norman M. L., 1998, ApJ, 495, 80 c (cid:13) , 000–000 V Emission from the CGM of High-z Galaxies Cantalupo S., 2010, MNRAS, 403, L16Cantalupo S., Arrigoni-Battaia F., Prochaska J. X., Hen-nawi J. F., Madau P., 2014,
Nature , 506, 63Cantalupo S., Porciani C., Lilly S. J., Miniati F., 2005,ApJ, 628, 61Cervi˜no M., Mas-Hesse J. M., Kunth D., 2002, A&A, 392,19Chan T. K., Kereˇs D., O˜norbe J., Hopkins P. F., MuratovA. L., Faucher-Gigu`ere C.-A., Quataert E., 2015, ArXive-printsCullen L., Dehnen W., 2010, MNRAS, 408, 669Dav´e R., Katz N., Oppenheimer B. D., Kollmeier J. A.,Weinberg D. H., 2013, MNRAS, 434, 2645Dehnen W., Aly H., 2012, MNRAS, 425, 1068Dijkstra M., Haiman Z., Spaans M., 2006, ApJ, 649, 14Dijkstra M., Loeb A., 2009, MNRAS, 400, 1109Durier F., Dalla Vecchia C., 2012, MNRAS, 419, 465Fabjan D., Borgani S., Tornatore L., Saro A., Murante G.,Dolag K., 2010, MNRAS, 401, 1670Fardal M. A., Katz N., Gardner J. P., Hernquist L., Wein-berg D. H., Dav´e R., 2001, ApJ, 562, 605Faucher-Gigu`ere C.-A., Feldmann R., Quataert E., KeresD., Hopkins P. F., Murray N., 2016, arXiv:1601.07188Faucher-Gigu`ere C.-A., Hopkins P. F., Kereˇs D., MuratovA. L., Quataert E., Murray N., 2015, MNRAS, 449, 987Faucher-Gigu`ere C.-A., Kereˇs D., 2011, MNRAS, 412, L118Faucher-Gigu`ere C.-A., Kereˇs D., Dijkstra M., HernquistL., Zaldarriaga M., 2010, ApJ, 725, 633Faucher-Gigu`ere C.-A., Kereˇs D., Ma C.-P., 2011, MNRAS,417, 2982Faucher-Gigu`ere C.-A., Lidz A., Zaldarriaga M., HernquistL., 2009, ApJ, 703, 1416Ferland G. J., Porter R. L., van Hoof P. A. M., WilliamsR. J. R., Abel N. P., Lykins M. L., Shaw G., Henney W. J.,Stancil P. C., 2013, Revista Mexicana de Astronomia yAstrofisica, 49, 137F¨orster Schreiber et al. 2009, ApJ, 706, 1364Frank S., Rasera Y., Vibert D., Milliard B., PoppingA., Blaizot J., Courty S., Deharveng J.-M., P´eroux C.,Teyssier R., Martin C. D., 2012, MNRAS, 420, 1731Fumagalli M., Fossati M., Hau G. K. T., Gavazzi G., BowerR., Sun M., Boselli A., 2014, MNRAS, 445, 4335Furlanetto S. R., Schaye J., Springel V., Hernquist L., 2004,ApJ, 606, 221Geach et al. 2009, ApJ, 700, 1Geach et al. 2014, ApJ, 793, 22Genzel et al. 2008, ApJ, 687, 59Gnat O., Sternberg A., 2007, ApJS, 168, 213Goerdt T., Dekel A., Sternberg A., Ceverino D., TeyssierR., Primack J. R., 2010, MNRAS, 407, 613Gould A., Weinberg D. H., 1996, ApJ, 468, 462Governato F., Brook C., Mayer L., Brooks A., Rhee G.,Wadsley J., Jonsson P., Willman B., Stinson G., QuinnT., Madau P., 2010,
Nature , 463, 203Guedes J., Callegari S., Madau P., Mayer L., 2011, ApJ,742, 76Haiman Z., Spaans M., Quataert E., 2000, ApJL, 537, L5Hennawi J. F., Prochaska J. X., Kollmeier J., Zheng Z.,2009, ApJL, 693, L49Hopkins P. F., 2013, MNRAS, 428, 2840Hopkins P. F., 2014, ArXiv e-printsHopkins P. F., Kereˇs D., O˜norbe J., Faucher-Gigu`ere C.- A., Quataert E., Murray N., Bullock J. S., 2014, MNRAS,445, 581Inoue A. K., Shimizu I., Iwata I., Tanaka M., 2014, MN-RAS, 442, 1805Kereˇs D., Hernquist L., 2009, ApJL, 700, L1Kereˇs D., Katz N., Weinberg D. H., Dav´e R., 2005, MN-RAS, 363, 2Kim et al. 2014, ApJS, 210, 14Knollmann S. R., Knebe A., 2009, ApJS, 182, 608Kollmeier J. A., Zheng Z., Dav´e R., Gould A., Katz N.,Miralda-Escud´e J., Weinberg D. H., 2010, ApJ, 708, 1048Law D. R., Steidel C. C., Erb D. K., Larkin J. E., PettiniM., Shapley A. E., Wright S. A., 2009, ApJ, 697, 2057Leitherer C., Schaerer D., Goldader J. D., DelgadoR. M. G., Robert C., Kune D. F., de Mello D. F., De-vost D., Heckman T. M., 1999, ApJS, 123, 3Ma X., Hopkins P. F., Faucher-Gigu`ere C.-A., Zolman N.,Muratov A. L., Kereˇs D., Quataert E., 2016, MNRAS,456, 2140Madau P., 1995, ApJ, 441, 18Martin C., Moore A., Morrissey P., Matuszewski M., Rah-man S., Adkins S., Epps H., 2010, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Se-ries Vol. 7735 of Society of Photo-Optical InstrumentationEngineers (SPIE) Conference Series, The Keck CosmicWeb ImagerMartin C. L., Scannapieco E., Ellison S. L., Hennawi J. F.,Djorgovski S. G., Fournier A. P., 2010, ApJ, 721, 174Martin C. L., Shapley A. E., Coil A. L., Kornei K. A.,Bundy K., Weiner B. J., Noeske K. G., Schiminovich D.,2012, ApJ, 760, 127Martin D. C., Chang D., Matuszewski M., Morrissey P.,Rahman S., Moore A., Steidel C. C., 2014, ApJ, 786, 106Martin et al. 2014a, ApJ, 786, 106Martin et al. 2014b, ApJ, 786, 107Mas-Hesse J. M., Ot´ı-Floranes H., Cervi˜no M., 2008, A&A,483, 71Matsuda Y., Yamada T., Hayashino T., Tamura H., Ya-mauchi R., Ajiki M., Fujita S. S., Murayama T., NagaoT., Ohta K., Okamura S., Ouchi M., Shimasaku K., Sh-ioya Y., Taniguchi Y., 2004, AJ, 128, 569Matuszewski M., Chang D., Crabill R. M., Martin D. C.,Moore A. M., Morrissey P., Rahman S., 2010, in Societyof Photo-Optical Instrumentation Engineers (SPIE) Con-ference Series Vol. 7735 of Society of Photo-Optical In-strumentation Engineers (SPIE) Conference Series, TheCosmic Web Imager: an integral field spectrograph forthe Hale Telescope at Palomar Observatory: instrumentdesign and first resultsMuratov A. L., Kereˇs D., Faucher-Gigu`ere C.-A., HopkinsP. F., Quataert E., Murray N., 2015, MNRAS, 454, 2691O˜norbe J., Boylan-Kolchin M., Bullock J. S., Hopkins P. F.,Kerˇes D., Faucher-Gigu`ere C.-A., Quataert E., Murray N.,2015, ArXiv e-printsOppenheimer B. D., Dav´e R., 2006, MNRAS, 373, 1265Oppenheimer B. D., Schaye J., 2013, MNRAS, 434, 1043Prescott M. K. M., Momcheva I., Brammer G. B., FynboJ. P. U., Møller P., 2015, ArXiv e-printsPrice D. J., 2008, Journal of Computational Physics, 227,10040Price D. J., Monaghan J. J., 2007, MNRAS, 374, 1347Prochaska J. X., Hennawi J. F., 2009, ApJ, 690, 1558 c (cid:13) , 000–000 Sravan et al.
Prochaska J. X., Wolfe A. M., 2009, ApJ, 696, 1543Rauch M., Haehnelt M., Bunker A., Becker G., MarleauF., Graham J., Cristiani S., Jarvis M., Lacey C., MorrisS., Peroux C., R¨ottgering H., Theuns T., 2008, ApJ, 681,856Richard J., Patricio V., Martinez J., Bacon R., Cl´ementB., Weilbacher P., Soto K., Wisotzki L., Vernet J., PelloR., Schaye J., Turner M., Martinsson T., 2015, MNRAS,446, L16Rosdahl J., Blaizot J., 2012, MNRAS, 423, 344Rubin K. H. R., Prochaska J. X., Koo D. C., Phillips A. C.,Martin C. L., Winstrom L. O., 2014, ApJ, 794, 156Rubin K. H. R., Weiner B. J., Koo D. C., Martin C. L.,Prochaska J. X., Coil A. L., Newman J. A., 2010, ApJ,719, 1503Rudie G. C., Steidel C. C., Trainor R. F., Rakic O., Bo-gosavljevi´c M., Pettini M., Reddy N., Shapley A. E., ErbD. K., Law D. R., 2012, ApJ, 750, 67Schaye et al. 2015, MNRAS, 446, 521Shapley A. E., Steidel C. C., Pettini M., Adelberger K. L.,2003, ApJ, 588, 65Shapley A. E., Steidel C. C., Pettini M., Adelberger K. L.,Erb D. K., 2006, ApJ, 651, 688Shen S., Wadsley J., Stinson G., 2010, MNRAS, 407, 1581Sijacki D., Vogelsberger M., Kereˇs D., Springel V., Hern-quist L., 2012, MNRAS, 424, 2999Somerville R. S., Dav´e R., 2014, ArXiv e-printsSparre M., Hayward C. C., Feldmann R., Faucher-Gigu`ereC.-A., Muratov A. L., Kereˇs D., Hopkins P. F., 2015,ArXiv e-printsSpringel V., 2005, MNRAS, 364, 1105Springel V., Hernquist L., 2003, MNRAS, 339, 289Steidel C. C., Adelberger K. L., Shapley A. E., Pettini M.,Dickinson M., Giavalisco M., 2000, ApJ, 532, 170Steidel C. C., Bogosavljevi´c M., Shapley A. E., KollmeierJ. A., Reddy N. A., Erb D. K., Pettini M., 2011, ApJ,736, 160Steidel C. C., Erb D. K., Shapley A. E., Pettini M., ReddyN., Bogosavljevi´c M., Rudie G. C., Rakic O., 2010, ApJ,717, 289Stinson G. S., Brook C., Macci`o A. V., Wadsley J., QuinnT. R., Couchman H. M. P., 2013, MNRAS, 428, 129Taniguchi Y., Shioya Y., 2000, ApJL, 532, L13Taniguchi Y., Shioya Y., Kakazu Y., 2001, ApJL, 562, L15Trainor R. F., Steidel C. C., 2012, ApJ, 752, 39Tumlinson J., Thom C., Werk J. K., Prochaska J. X., TrippT. M., Katz N., Dav´e R., Oppenheimer B. D., MeiringJ. D., Ford A. B., O’Meara J. M., Peeples M. S., SembachK. R., Weinberg D. H., 2013, ApJ, 777, 59Turner M. L., Schaye J., Steidel C. C., Rudie G. C., StromA. L., 2015, MNRAS, 450, 2067van de Voort F., Schaye J., 2013, MNRAS, 430, 2688Verhamme A., Schaerer D., Maselli A., 2006, A&A, 460,397Vogelsberger M., Genel S., Springel V., Torrey P., SijackiD., Xu D., Snyder G., Nelson D., Hernquist L., 2014, MN-RAS, 444, 1518Wiersma R. P. C., Schaye J., Smith B. D., 2009, MNRAS,393, 99Wiersma R. P. C., Schaye J., Theuns T., 2011, MNRAS,415, 353Wisotzki et al. 2015, ArXiv e-prints Wright S. A., Larkin J. E., Law D. R., Steidel C. C., Shap-ley A. E., Erb D. K., 2009, ApJ, 699, 421Yang Y., Zabludoff A., Tremonti C., Eisenstein D., Dav´eR., 2009, ApJ, 693, 1579
APPENDIX A: CONVERGENCE TEST
We test the convergence of our predictions for the intensityof metal emission lines with respect to the resolution of theSPH calculation. To do this we use three versions of the m12i simulation: one having a ‘fiducial resolution’ (the oneanalyzed in the main text and listed in Table 1), one havinga ‘low resolution’ (with a gas particle mass 8 times largerand a minimum gas softening length 5 times larger), and onehaving a ‘high resolution’ (with a gas particle mass 8 timessmaller but same minimum softening). The ‘high resolution’version was not evolved all the way to z = 2 due to its highcomputational cost; the figure shows this simulation onlyto z = 2 .
6. We also use another version of this halo whoseparameters are similar to the ‘fiducial resolution’ versionbut in which the normalization of the artificial conductivity(entropy mixing parameter) is multiplied by a factor of 4.In Figure A1 we plot the C
III luminosity within a virialradius as a function of redshift in all four realizations ofthe m12i halo described above. In this calculation we haveexcluded a region of 10 proper kpc centered on the halosas a proxy for removing emission from the galaxy. Consis-tent with the results found in § III luminosity does not depend systematically onresolution or on the normalization of the artificial conductiv-ity, and thus that our predictions are reasonably convergedfor our fiducial resolution simulations.
APPENDIX B: LOCAL SOURCE TEST
To further test the sensitivity of our CGM emission predic-tions to local sources of ionizing radiation, we repeated ourcalculations explicitly including a local source spectrum inaddition to the FG09 cosmic background. Specifically, weadded a local source spectrum from Cervi˜no et al. (2002)that includes X-ray emission from supernova remnants fol-lowing Cantalupo (2010). To emphasize the effects of localsources, we chose a relatively high normalization for the lo-cal source spectrum corresponding to a distance of 10 properkpc from a solar-metallicity galaxy with a star formationrate of 100 M (cid:12) yr − . We assume that a fraction 5% of themechanical energy from the starburst is converted into ISMheating and X-ray emission (Mas-Hesse et al. 2008). We as-sume that all the X-rays escape the galaxy but an escapefraction of 5% for the softer ionizing photons between 1 and4 Ry. Figure B1 shows the results of this test and comparesthem with calculations including only the FG09 background(left) and only the FG09 background but with normaliza-tion multiplied by a factor of ten (right). For surface bright-nesses > − erg s − cm − arcsec − potentially observablein deep and stacked observations with MUSE and KCWI,there is no significant difference between the different ioniza-tion models. At lower surface brightnesses, some differences c (cid:13)000
To further test the sensitivity of our CGM emission predic-tions to local sources of ionizing radiation, we repeated ourcalculations explicitly including a local source spectrum inaddition to the FG09 cosmic background. Specifically, weadded a local source spectrum from Cervi˜no et al. (2002)that includes X-ray emission from supernova remnants fol-lowing Cantalupo (2010). To emphasize the effects of localsources, we chose a relatively high normalization for the lo-cal source spectrum corresponding to a distance of 10 properkpc from a solar-metallicity galaxy with a star formationrate of 100 M (cid:12) yr − . We assume that a fraction 5% of themechanical energy from the starburst is converted into ISMheating and X-ray emission (Mas-Hesse et al. 2008). We as-sume that all the X-rays escape the galaxy but an escapefraction of 5% for the softer ionizing photons between 1 and4 Ry. Figure B1 shows the results of this test and comparesthem with calculations including only the FG09 background(left) and only the FG09 background but with normaliza-tion multiplied by a factor of ten (right). For surface bright-nesses > − erg s − cm − arcsec − potentially observablein deep and stacked observations with MUSE and KCWI,there is no significant difference between the different ioniza-tion models. At lower surface brightnesses, some differences c (cid:13)000 , 000–000 V Emission from the CGM of High-z Galaxies Figure A1.
Convergence test for the predicted C III luminosity for m12i with respect to the simulation resolution and the artificialconductivity normalization. We plot the C III luminosity within a virial radius as a function of redshift for the fiducial resolution (analyzedin the main text), for a lower resolution run, for a higher resolution run (stopped at an earlier redshift), and for fiducial resolution runwith the normalization of artificial conductivity multiplied by a factor of 4 (see text for more details). Accounting for the stochasticityof the simulations, we conclude that the predicted C III luminosity does not depend systematically on resolution or on the normalizationof the artificial conductivity, and thus that our predictions are reasonably converged for our fiducial resolution simulations. are noticeable with the inclusion of a local ionizing source.In particular, the C III line is suppressed at large radii whilethe C IV line is enhanced. However, at those very low sur-face brightnesses, our test overestimates the magnitude ofthe effect since we do not model the ∝ R − drop off of thelocal flux with radius. The effects are seen also in the calcula-tion in which we simply boosted the cosmic UV background,indicating that it is primarily due to the enhancement of rel-atively soft ionizing photons rather than X-rays from localsupernova remnants. c (cid:13) , 000–000 Sravan et al.
Figure B1.
Test of the effects of a local source including X-rays on theoretical surface brightness profiles. We compare the median(solid) and mean (dashed) surface brightness profiles for five halos ( z2h506 , z2h450 , z2h400 , z2h830 , and m12q ) at z = 3. Left:
Profiles including only the FG09 cosmic ionizing background (default in main text).
Middle:
Profiles including the sum of the FG09background and a local source spectrum from Cervi˜no et al. (2002), which includes the contribution of supernova remnants to X-rays.The normalization of the X-ray contribution is set to the value expected at 10 proper kpc from a galaxy with a star formation rate of100 M (cid:12) yr − . Right:
Profiles including only the FG09 background but with normalization boosted by a factor of ten. The mean andmedian halo masses in log M h (M (cid:12) ) are 11.46 and 11.44, respectively. The vertical dotted, dot-dashed, dashed and solid grey linesindicate radii of 5, 10, 20, and 50 proper kpc for the median mass. c (cid:13)000