Origin of Weak MgII and Higher Ionization Absorption Lines in an Outflow from an Intermediate-Redshift Dwarf Satellite Galaxy
Akimi Fujita, Toru Misawa, Jane C. Charlton, Avery Meiksin, Mordecai-Mark Mac Low
DDraft version September 22, 2020
Typeset using L A TEX twocolumn style in AASTeX63
Origin of Weak Mg II and Higher Ionization Absorption Lines in an Outflow from anIntermediate-Redshift Dwarf Satellite Galaxy Akimi Fujita, Toru Misawa, Jane C. Charlton, Avery Meiksin, and Mordecai-Mark Mac Low
5, 6 Faculty of Engineering, Shinshu University, 4-17-1 Wakasato, Nagano, Nagano 380-0926, Japan School of General Studies, Shinshu University, 3-1-1, Asahi, Matsumoto City 390-8621, Japan Department of Astronomy & Astrophysics, The Pennsylvania State University, University Park, PA, 16802, USA SUPA a , Institute for Astronomy, University of Edinburgh, Blackford Hill, Edinburgh EH9 3HJ, UK Department of Astrophysics, American Museum of Natural History, New York, NY 10024, USA Center for Computational Astrophysics, Flatiron Institute, New York, NY 10010, USA
Submitted to ApJABSTRACTObservations at intermediate redshifts reveal the presence of numerous, compact, weak Mg ii ab-sorbers with near to super-solar metallicities, often surrounded by more extended regions that produceC iv and/or O vi absorption in the circumgalactic medium at large impact parameters from luminousgalaxies. Their origin and nature remains unclear. We hypothesize that undetected, satellite dwarfgalaxies are responsible for producing some of these weak Mg ii absorbers. We test our hypothesisusing gas dynamical simulations of galactic outflows from a dwarf satellite galaxy with a halo mass of5 × M (cid:12) , which could form in a larger L ∗ halo at z = 2, to study the gas interaction in the halo.We find that thin, filamentary, weak Mg ii absorbers are produced in two stages: 1) when shockedcore collapse supernova (SNII) enriched gas descending in a galactic fountain gets shock compressedby upward flows driven by subsequent SNIIs and cools ( phase 1 ), and later, 2) during an outflowdriven by Type Ia supernovae that shocks and sweeps up pervasive SNII enriched gas, which thencools ( phase 2 ). The width of the filaments and fragments are (cid:46)
100 pc, and the smallest ones cannotbe resolved at 12.8 pc resolution. The Mg ii absorbers in our simulations are continuously generatedfor >
150 Myr by shocks and cooling, though each cloud survives for only ∼
60 Myr. Their metallicityis 10–20% solar metallicity and column density is < cm − . They are also surrounded by larger(0.5–1 kpc) C iv absorbers that seem to survive longer. In addition, larger-scale ( > iv andO vi clouds are produced in both expanding and shocked SNII enriched gas which is photoionized bythe UV metagalactic radiation at intermediate redshift. Our simulation highlights the possibility ofdwarf galactic outflows producing highly enriched multiphase gas. Keywords: galactic outflows — CGM— hydrodynamic simulations — dwarf galaxies INTRODUCTIONGalactic outflows appear to regulate the structure andevolution of galaxies, as they heat, ionize, and chem-ically enrich the surrounding circumgalactic medium(CGM) and even drive unbound winds that can reachthe intergalactic medium (IGM) (see e.g. Somerville &
Corresponding author: Akimi [email protected] a Scottish Universities Physics Alliance
Dav´e 2015; Heckman 2017, for reviews). A robust under-standing of the stellar feedback processes driving theseoutflows, however, remains elusive. The observed prop-erties of the outflows and outflow-CGM interaction atmultiple wavelengths must be used to constrain the-oretical models of the physics governing the outflowsand outflow-CGM interaction. The most prominent ob-served properties are metal absorption lines, seen in thespectra of background quasars, that are believed to arisefrom inhomogeneities in the CGM. Numerical simula-tions are required to predict and interpret the obser-vational signatures of these systems (e.g. Oppenheimer a r X i v : . [ a s t r o - ph . GA ] S e p t al. 2012; Suresh et al. 2015; Keating et al. 2016;Turner et al. 2017; Oppenheimer et al. 2018; Peepleset al. 2019).The derived metallicities of weak, low ionization ab-sorbers are almost always greater than 10% solar andare often as high or even higher than the solar value.Some of them are even iron-enhanced compared with so-lar (Rigby et al. 2002; Charlton et al. 2003; Narayananet al. 2008; Misawa et al. 2008; Lynch & Charlton 2007).In addition, analyses of low-redshift absorbers show thatthere are fewer absorbers at present than in the past,and that most absorbers seem to live in group environ-ments (Muzahid et al. 2017). With all the measuredproperties above, it is plausible to speculate that weakabsorbers are created by transient processes, such asgalactic outflows that carry metals and are less activein the modern Universe. The outflows may originate insatellite dwarf galaxies hosted by a larger halo that aretoo dim to be observed.The covering fraction of the weak absorbers is esti-mated to be (cid:38)
30% in the CGM of galaxies brighter than0.001L ∗ (Narayanan et al. 2008; Muzahid et al. 2017).There would be on the order of a million tiny, weak ab-sorbers per galaxy if a spherical geometry were assumed(Rigby et al. 2002). It has been argued, however, thatweak absorbers reside instead in filamentary and sheet-like structures (Milutinovic et al. 2006).Many of these systems show absorption by multiplehigh ionization species at the same velocity, often withadditional components offset by 5-150 km s − (Miluti-novic et al. 2006). C iv surveys at z ≈ ∼ iv clouds are more diffuse (n HI ∼ − to10 − cm − ) and larger than Mg ii clouds, with sizes be-tween 0.1 kpc and 10 kpc (Simcoe et al. 2004; Schayeet al. 2007; Lehner et al. 2016). Some of C iv cloudsmay have expanded from denser, more compact Mg ii clouds (Schaye et al. 2007). These C iv systems maybe interpreted as being in photoionization equilibriumat T ∼ K , and their metallicities are found to be ∼
1% solar to even solar or more (Simcoe et al. 2004;Schaye et al. 2007; Lehner et al. 2016). There are alsomany O vi absorption systems, which are more likely tohave an origin in photoionized gas (rather than collision-ally ionized gas) at z ∼ vi by Turner et al. (2014,2015), however, suggest the presence of a collisionallyionized gas phase for impact parameters (cid:46)
100 properkpc (pkpc) of large, star-forming galaxies at z ∼ . ∗ galaxy, produce com- pact weak Mg ii absorbers surrounded by larger regionsthat produce C iv and O vi absorption. Using a small-scale hydrodynamical simulation of a dwarf galaxy, wefind such structures are produced by repeated shocksand radiative cooling in the gaseous halo of the galaxy.We will highlight important physical processes at workwhich regulate the production of low and high ioniza-tion clouds, to be explored in larger-scale simulationsin the next paper. We describe our numerical methodin Section 2 and the dynamics of SNII and SNIa drivenoutflows and their interaction with surrounding gas, in-cluding the production of dense clumps and filaments,in Section 3. In Section 4, we study the distributionsof weak Mg ii absorbers and surrounding C iv and O vi absorbers in our simulation, and compare them to theproperties of observed systems, followed by a resolutionstudy (Section 5) and a summary (Section 6). NUMERICAL METHODWe use the adaptive mesh refinement hydrodynamicscode Enzo (Bryan et al. 2014) to simulate repeated su-pernova explosions in the disk of a dwarf galaxy. Wesolve the equations of hydrodynamics using a direct-Eulerian piecewise parabolic method (Colella & Wood-ward 1984; Bryan et al. 2014) and a two-shock approxi-mate Riemann solver with progressive fallback to morediffusive Riemann solvers in the event that higher ordermethods produce negative densities or energies. Oursimulation box has dimensions ( L x , L y , L z ) = (6.5536,6.5536, 32.768) kpc, initially with (32, 32, 160) cells.Only half the galactic disk above its midplane is simu-lated. We refine cells to resolve shocks with a standardminimum pressure jump condition (Colella & Wood-ward 1984) and to resolve cooling at turbulent inter-faces where the sound crossing time exceeds the coolingtime. We use 4 refinement levels resulting in a highestresolution of 12.8 pc ( standard simulation ). We also ranthe same simulation with 3 refinement levels as a com-parative resolution study ( low-res simulation ), and byapplying 6 refinement levels in a region where Mg ii fil-aments form in order to test the effects of resolution onfragmentation ( high-res zoom simulation ). We assumea flat ΛCDM cosmology with the 2018 Planck Collabo-ration measured parameters Ω m = 0 . Λ = 0 . h = 0 . b = 0 . Galaxy Model
We model a dwarf galaxy at redshift z = 2 with ahalo mass M halo = 4 × M (cid:12) , and a virial radius R vir = 17 . M g = 5 . × M (cid:12) . We adopt a Burkert (1995) darkmatter potential with a core radius r = 848 pc andentral density ρ = 1 . × − g cm − , although thispotential profile is a fit to the observed rotation curvesof nearby dwarf galaxies rather than those at z=2. Ourchoice of r and ρ ensures that the resulting poten-tial profile reproduces a Navarro et al. (1997) dark mat-ter potential with c = 12 . r >
400 pc. The gas is described as a softened exponen-tial disk: ρ ( R, z ) = M g πa b g . sech (cid:18) Ra g (cid:19) sech (cid:18) zb g (cid:19) (1)where M g is the total mass of gas in the disk, and a g and b g are the radial and vertical gas disk scale heights(Tonnsen & Bryan 2009). We chose a g = 621 pc basedon the exponential disk approximation of Mo et al.(1998), with λ = 0 .
05, and b g = 160 pc based on thethin disk approximation (Toomre 1963) with an effectivesound speed, c s , eff = 11 . − (Fujita et al. 2009).Given this gas density distribution in the disk, the gastemperature and pressure are calculated to maintainthe disk in hydrostatic equilibrium with the surround-ing halo potential in the z -direction, and the rotationalvelocity of the gas disk is set to balance the radial grav-itational force and the pressure gradient. The disk tem-perature varies between 10 K and a few × K, andthe maximum circular velocity is v max = 48 . − with the escape velocity from the potential v esc =69 . − . Our model galaxy is placed in a static halobackground with ρ bg = 1 . × − g cm − so that thegas mass within the virial radius is M halo (cid:16) Ω b Ω m (cid:17) . Themetallicity of all the gas in the box is initially set at Z = 0 .
001 with mean molecular weight µ = 0 . Cooling
Figure 1 shows the cooling curves used in our simula-tions. We use radiative cooling curves as a function oftemperature above 10 K for gas in collisional ionizationequilibrium (CIE) with various metallicities: [Fe/H]=-3,-2, -1.5, -1, -0.5, 0, +0.5 (Sutherland & Dopita 1993).A radiative cooling rate for gas in a cell with a metal-licity is computed by interpolating between the coolingcurves. Cooling of gas below temperature 10 K is ap-proximated with the cooling curve of Rose & Bregman(1995) computed for solar metallicity. Although, for ex-ample, Maio et al. (2007) shows that the cooling ratestays approximately the same between 10 and 10 Kfor gas with a metallicity below Z = 10 − , we justify thesimplification below 10 K by noting that cooling below10 K has a negligible effect on the formation and frag-mentation of dense clouds as cooling in shocked gas andturbulent mixing layers is limited by numerical resolu-tion rather than by radiative cooling (Fujita et al. 2009; Gronke & Oh 2018, 2020). We justify CIE assumptionbecause past simulations show that the effects of non-equilibrium ionization (NEI) do not much boost highion distributions even in shocked coronal gas (Kwak &Shelton 2010; Armillotto et al. 2016; Cottle et al. 2018).We do not include the effects of a metagalactic UV back-ground radiation in our simulation, but we incorporatethem when we post-process the simulations to computethe ion distributions (see Sect. 4). The modification ofthe ionization fraction by a UV background would affectonly the lower density gas that does not dominate thecooling. Temperature (K) C oo li n g f un c t i o n ( e r g c m s ) SD93:[Fe/H]=0 [Fe/H]=0.5 [Fe/H]=-3RB95:[Fe/H]=0
Figure 1.
Radiative cooling functions used in our simu-lations as a function of temperature T from Sutherland &Dopita (1993) for T ≥ K for different metallicities andfrom Rose & Bregman (1995) for
T < K for solar metal-licity.
Starburst
In our study, we set up an instantaneous starburstof total stellar mass 4 × M (cid:12) at the disk center.We use Stellar Yields for Galactic Modeling Applica-tions (SYGMA Ritter et al. 2018) to model the chemi-cal ejecta and feedback from simple stellar populations(SSPs). SYGMA is part of the open-source chemicalevolution NuGrid framework (NuPyCEE ). We com-pute the average mechanical luminosities and the aver-age metal ejection rates for M SSP = 4 × M (cid:12) . Theyare L SNII = 4 × erg s − and ˙ M SNII = 1 . × − M (cid:12) yr − for the initial 40 Myr, which is the lifetime ofthe smallest B star to go core collapse Type II su-pernova (SNII), and L SNIa = 8 × erg s − and˙ M SNIa = 1 . × − M (cid:12) yr − at times ≥
40 Myr poweredby Type Ia supernovae (SNIa). The metals produced bySNII and SNIa are followed and advected separately. o drive a constant-luminosity outflow, during everytime step ∆ t we add mass ( ˙ M in ∆ t ) and energy ( L SNII ∆ t and L SNIa ∆ t ) to a spherical source region with a radiusof 102.4 pc. We choose to increase the amount of massadded from the SYGMA values to ensure that the tem-perature of hot gas in the outflows is 3 × K, whichis far from the peak of the cooling curve at ∼ K,but well below the value implied by only accounting forthe ejecta. This additional mass accounts for the massevaporated off the swept-up shells in the absence ofan implementation of thermal heat conduction. There-fore, we use ˙ M in = 8 . × − M (cid:12) yr − for the SNIIdriven outflow and 1 . × − M (cid:12) yr − for the SNIadriven outflow. The total mass added for 1 Gyr is only3 . × M (cid:12) which is less than 1% of M disk .2.4. Ion Analysis
We use the TRIDENT analysis tool (Hummels et al.2017) to calculate the ionization fractions of the speciesof interest based on the cell-by-cell density, temperature,and metallicity. First, the estimation for the numberdensity of an element X is n X = n H ZZ (cid:12) (cid:18) n X n H (cid:19) (cid:12) , (2)where Z is the metallicity from the simulation, and( n X /n H ) (cid:12) is the solar abundance by number. Ioniza-tion fractions are pre-calculated over a grid of temper-ature, density, and redshift in photoionization equilib-rium (PIE) with the metagalactic UV background radi-ation by Haardt & Madau (2012), using the photoioniza-tion software CLOUDY (Ferland et al. 2013). Thus bylinearly interpolating over the pre-calculated grid, TRI-DENT returns the density of an ion, i , of an element, X as n X i = n X f X I , (3)where f X I is the ionization fraction of the i th ion.To generate an absorption profile along a ray throughthe simulation box, the absorption produced by eachgrid cell is represented by a single Voigt profile at itsinstantaneous velocity v , with a Doppler b parameterspecified by the temperature in the cell. RESULTSFigure 2 shows density, temperature, total velocity,and metallicity slices along the y-z plane at the diskcenter and a neutral hydrogen (HI) column density dis-tribution along x axis in the y - z plane at t=40 Myr. TheHI distribution is calculated with Trident.The swept-up shell driven by repeated SNII explosionscools quickly due to its high density. Because it is ex-panding into a stratified atmosphere, it accelerates and fragments into multiple clumps and shells due to theRayleigh-Taylor (RT) instability. Figure 2 shows thatthe hot, thermalized interior gas expands freely throughthe fragments, forming a supersonic, energy-driven out-flow. Kelvin-Helmholtz instabilities ablate the sides ofthese fragments as the hot gas streams past them. Thisoutflow continues to shock the CGM, and a classic su-perbubble (Weaver et al. 1977) forms in the CGM, asseen in Figure 2: region (a) expanding SNII enrichedgas at v ∼ − , region (b) shocked SNII en-riched gas, region (c) swept-up CGM shell which is verythin and light because there is not much to sweep due toits low density, and region (d) the ambient CGM beyondthe outer shock front at z ∼
17 kpc. Expanding SNII en-riched gas and shocked SNII enriched gas are dividedat the inner shock front at z ∼
12 kpc, and shocked SNIIenriched gas extends out to a contact discontinuity withthe CGM.In our simulations, the high-density, low-temperaturefragments of swept-up ISM material are not resolved af-ter t = 40 Myr with our refinement criteria of strongpressure gradients or the sound crossing time exceedingthe cooling time. They are Lyman Limit Systems (LLSs)and sub-Damped Lyman-alpha Absorbers (DLAs) with N HI (cid:38) − cm − that will likely produce strongMg ii absorbers (see rightmost figure in Figure 2). Thefocus of this study is instead on weak Mg ii absorbersthat are observed to be associated with sub-LLSs with N HI (cid:46) cm − . These unresolved swept-up ISMfragments in the outflow quickly mix with the surround-ing hot, metal enriched gas, but the total amount ofdisk gas mixed in the outflow is only 3–5% of the diskmass initially placed on the grid. We also note that thepowerful SNII driven outflow leaves the box starting at t ∼
20 Myr; by t = 40 to 300 Myr, 38% to 58% of themetal-carrying gas has left the box.After the last SNII goes off at t = 40 Myr, SNIa’sdrive the outflow, but with a mechanical luminosity thatis more than two orders of magnitude smaller. SNIa-enriched gas expands at v ∼
400 km s − through thetunnel created by the previous SNII outflow, but by t ∼
80 Myr the disk gas being pushed aside by theSNII outflow flows back to the central source region,blocking the passage for SNIa-enriched gas. Meanwhile,the shocked SNII-enriched gas (region b) near the innershock front ( z ∼
12 kpc) begins to descend toward thedisk, while the outer shock front (the outer edge of re-gion c) keeps moving at v ∼
400 km s − in the CGMand soon leaves the box. By t ∼
100 Myr, descend-ing shocked SNII enriched gas accumulates at the innershock front and cools to form denser, cool shells thateventually fragment by RT instability. .5 0.0 2.5 y (kpc) z ( k p c ) Density [cm ]t = 40 Myr 1 kpc100 pc 2.5 0.0 2.5 y (kpc) Temperature [K] 1 kpc 2.5 0.0 2.5 y (kpc) region (a)region (b)region (c)region (d)Velocity [km s ] 2.5 0.0 2.5 y (kpc) Metallicity [Z ] 1 kpc 2.5 0.0 2.5 y (kpc)
HI column [cm ] 1 kpc100 pc10 Figure 2.
Sliced density, temperature, velocity magnitude, and metallicity ( from left to right ) distributions and a projectedhydrogen column density distribution along the x -axis ( rightmost ) of the SNII-driven outflow at the box center ( y - z plane) whenthe last SNII goes off at t =40 Myr. The middle figure denotes region a) expanding SNII enriched gas, b) shocked SNII enrichedgas, c) swept-up CGM, and d) the ambient CGM. The sliced density distribution in the y - z plane at x = +1 .
42 kpc from the disk center at t = 160 Myr( left in Figure 3) shows the formation of such fragmentsin the form of clumps and filaments. They are also vis-ible as clumps and filaments in a projected distributionof neutral hydrogen along the x -axis at t = 160 Myr( left in Figure 4). These clumps and filaments will po-tentially produce weak Mg ii absorbers (we discuss ourion analysis in the next section). We call this process phase 1 formation. They are made of SNII enriched out-flow gas and their metallicity is ∼ . . Z (cid:12) . The sizeof clumps and the thickness of filaments are ∼
100 pc.This size may be limited by our numerical resolution of12.8 pc (Fujita et al. 2009; Gronke & Oh 2018). Wediscuss the effects of resolution further in Sec. 5.Shortly after t = 160 Myr, a superbubble created byrepeated SNIa explosions blows out of the dense ISMand SNIa enriched gas regains a tunnel for expansion,forming a SNIa driven outflow traveling at v ∼ − . In the projected distribution of neutralhydrogen at t = 200 Myr ( middle panel in Fig. 4), frag-ments of swept-up ISM after blowout are visible fram-ing a tunnel for outflow, and hot, low-density SNIaenriched gas in the outflow is seen as a cavity with N HI (cid:46) cm − (we define SNIa enriched gas as regionIa).By t = 220 Myr, this SNIa driven outflow (regionIa) expands into the cooled SNII enriched gas and theclumps and filaments of shocked SNII enriched gas (re- gion b), shocking and sweeping them and forming moreclumps and filaments. Figure 3 shows such a processclearly in a selected region at z >
10 kpc. These arepotential candidates for weak Mg ii absorbers, too: wecall this process phase 2 formation. Their metallicityand size are likewise ∼ . . Z (cid:12) and ∼
100 pc. Hot-ter and lower-density shocked SNII enriched gas carriesmore metals ( Z ∼ . Z (cid:12) ) and lies above z ∼
14 kpc.The SNIa driven outflow continues to shock and sweepgas as well as clumps and filaments to the sides, and by t ∼
300 Myr, all the clumps and filaments as well as 58%of SNII outflow gas and 8% of SNIa outflow gas have leftthe box. Then, there is only very low-density gas withn H < − cm − left above the disk in the box. Themetallicity of SNIa enriched gas is Z (cid:28) . Z (cid:12) as themetal production rate is about two orders of magnitudesmaller than that of SNII, so it is still too early for anysignificant enrichment by SNIa. We stopped computingat t ∼
450 Myr.With a realistic star formation history with multiplestar clusters scattered in time and place, we expect phase1 and phase 2 formation to be repeated in time andplace to produce more clumps and filaments. We willtest this scenario in a larger simulation box in our nextpaper. WEAK MG II ABSORBERS AND C IV /O VI ABSORBERS y (kpc) z ( k p c ) y (kpc) y (kpc) y (kpc) y (kpc) D e n s it y [ c m ] T e m p e r a t u r e [ K ] M e t a lli c it y [ Z ] Figure 3.
Sliced density ( top ), temperature ( middle ), and metallicity ( bottom ) distributions of cool, dense clouds at x = +1 . y - z plane at phase 1 (t=160 and 200 Myr) and phase 2 ( t = 220, 230, and 240 Myr) from leftto right . Phase 1 formation begins when descending shocked SNII enriched gas ( region b ) collides with the expanding SNIIenriched gas ( region a ) at the inner shock front, and phase 2 formation begins when SNIa driven outflow ( region Ia ) rams intothe rest of the SNII enriched gas and the clouds made at phase 1 . The arrows in bottom figures show the direction of gas flowwith v max = 429 km s − . y (kpc) z ( k p c ) t = 160 Myr 3 2 1 0 1 2 3 y (kpc) t = 200 Myrregion (Ia)region (a)region (b) 3 2 1 0 1 2 3 y (kpc) t = 240 Myr 10 H I c o l u m n d e n s it y [ c m ] Figure 4.
Projected neutral hydrogen distributions at t =160 ( left ), 200 ( middle ), and 240 Myr ( right ), along the x -axis in the y - z plane. SNIa driven outflow is visible as acavity ( region Ia ) We calculate the fractions of H i , H ii , Mg ii , C iv , andO vi ions using Trident as described in Section 2.4 byassuming all the gas in our simulation to be in photoion-ization equilibrium (PIE) with the UVB radiation at agiven redshift (Haardt & Madau 2012). Our simulationdoes not include the effects of UVB radiation, so for ex-ample, expanding SNII enriched gas tends to overcool tolower temperature, (cid:46) K. However, this overcooled,low-density ( ≤ − cm − ) gas contributes very little tothe total ion budgets, and denser clouds that produceMg ii absorbers are self-shielded to the surrounding UVBradiation as long as n H ∼ × − cm − at z=2 (Rah-mati et al. 2013). Thus overcooling will not significantlyaffect our analysis (see Appendix).4.1. Overview
Figure 5 shows projected density distributions of Mg ii ,C vi , and O vi ions along the x-axis in the y-z plane att=160, 200, and 240 Myr, and Figure 6 shows sliceddensity, temperature, metallicity, Mg ii , C vi , and O vi ion density distributions at x=+1.92 kpc from the diskcenter in the y-z plane at t=200 Myr. This sight line was z ( k p c ) t = 160 Myr t = 200 Myr t = 240 Myr4681012141618 z ( k p c ) region (a)region (b) region (Ia)2 0 2 y (kpc) z ( k p c ) y (kpc) y (kpc) M g II c o l u m n d e n s it y [ c m ] C I V c o l u m n d e n s it y [ c m ] OV I c o l u m n d e n s it y [ c m ] Figure 5.
Projected Mg ii ( top ), C iv ( middle ), and O vi density ( bottom ) distributions at t = 160 ( left ), 200 ( middle ), and240 Myr ( right ), along the x -axis in the y - z plane. Density [cm ]region (a)region (b)region (Ia)t = 200 Myr Temperature [K] Metallicity [Z ]2 0 2 y (kpc) z ( k p c ) MgII [cm ] 2 0 2 y (kpc) CIV [cm ] 2 0 2 y (kpc) OVI [cm ]10 Figure 6.
Sliced density, temperature, metallicity ( top from left to right ), and Mg ii , C iv , and O vi ( bottom from left to right )density distributions at x=+1.92 kpc from the disk center in the y-z plane, at t=200 Myr. A line of sight from [x,y,z]=[+1.92 kpc,-3.28 kpc, +2.45 kpc] to [+1.92 kpc, +3.28 kpc, +14.4 kpc] is shown by a green line . The arrows in the bottom right figureshow the direction of the gas flow with v max =353 km s − . elected as an example with a large pathlength throughlow ionization gas.The clumps and filaments have hydrogen numberdensities, n H = 10 − to 10 − cm − , and theirsizes/thickness, ∼
100 pc, which is the smallest scale oursimulation can resolve, as discussed in Section 3. In-dividual weak Mg ii absorbers seem to survive for ∼ phase 1 to phase 2 forma-tion for over 150 Myr from a single instantaneous star-burst source. Weak Mg ii absorption with N MgII > cm − is also found in a blob of gas that carries a swept-up ISM shell fragment in the expanding SNII enrichedgas seen at e.g. [y,z]=[+2 kpc, 10 kpc] (see top left figurein Figure 5) and in fragmented shells of ISM swept-upby the SNIa driven outflow at e.g. z=2–4 kpc (see topmiddle figure in Figure 5). The blob has cooled slowlywithout fragmentation, and its size is (cid:38) kpc. It is ex-panding into the phase 1 shells in region (b) above, butSNIa driven outflow will shock and sweep up expand-ing SNII enriched gas including the blob in region (a)and the phase 1 shells in region (b) to produce phase 2 shells and fragments (see top right figure in Figure 5).Higher ion absorbers are found in region (a) where ex-panding SNII enriched gas cools and in region (b) whereshocked SNII enriched gas cools in phase 1 and phase2 . In both cases, the hydrogen number density of theabsorbers is n H ∼ a few × − cm − , but the absorbersin region (a) extend over 1-4 kpc while the absorbersin region (b) are smaller, 500 pc–1 kpc. The sizes ofhigh ion absorbers agree with the observed estimatesfor C iv absorbers by Misawa et al. (2008) and Schayeet al. (2007). They are ∼
100 pc - 5 kpc in a sub-LLS(10 . 38 km s − ),which is visible in the absorption profile as a slight asym-metry(Figure 8).The same shells produce C iv absorption, but no O vi absopbtion. O vi absorbers in region (b) are in a dif-ferent, coronal phase. C iv absorbers in region (a) are > few kpc in size: one is at z ∼ ∼ 10 km s − ), one is at z ∼ ∼ -5 km s − ), one is at z ∼ ∼ 30 km s − ) and the other is at z ∼ ∼ 40 km s − ), both below the cooling shell(v ∼ 38 km s − ). The first two absorbers produce thedouble absorption profiles in Figure 8, and the last twoabsorbers produce the saturated absorption profile atv=20–45 km s − , together with C iv absorbers in region(b).O vi absorbers in region (a) arise from the same coldclouds, producing two sets of double absorption profiles,but the sightline is also going through a turbulent mixinglayervof swept-up shells by the SNIa driven outflow atz=5–7 kpc. Its temperature is (cid:38) K. The signal isburied in the double absorption profiles at v ∼ 10 km s − .The O vi absorber in region (b) is coronal and turbulentwith v ∼ -10–40 km s − , but is weak compared with theother O vi absorbers.We note that some SNII outflow gas in region a coolsto temperature below 10 K by t (cid:38) 200 Myr, however,this overcooled, low-density ( ≤ − cm − ) gas onlymakes a little contribution to C iv and O vi column col-umn densities (see Appendix).Figure 8 also shows that our mock spectra reproducequalitative features of the observed profiles of observed l o g ( n H ) ( c m ) v ( k m s ) l o g ( T ) ( K ) 0 4 8 12a line of sight (kpc)-18-16-14-12-10 l o g ( n M g II ) ( c m ) 0 4 8 12a line of sight (kpc)-14 -12 -10 -8 l o g ( n C I V ) ( c m ) 0 4 8 12a line of sight (kpc)-12-11-10-9 l o g ( n O V I ) ( c m ) Figure 7. Hydrogen density ( top left ), sightline velocity ( top middle ), temperature ( rop right ), Mg ii density ( bottom left ), C iv density ( bottom middle ), and O vi density ( bottom right ) distributions along the line of sight from [x,y,z]=[+1.92 kpc, 0 kpc,2.45 kpc] to [+1.92 kpc, 6.55 kpc, 14.4 kpc] ( green line in Figure 6) at t=200 Myr. Lya 1216 Lyb 1026 SiII 1193 SiII 1260 MgII 2796 MgII 2804 SiIII 1270 SiIV 1403 N o r m a li z e d f l u x CIV 1548 CIV 1551 CII 1335 CII 1036 100 0 1000.00.51.0 OVI 1032 50 0 50 100 Relative velocity (km/s) OVI 1038 50 0 50 100 FeII 2383 50 0 50 100 FeII 2600 Figure 8. Mock spectra along the line of sight ( green line in Figure 6) at t=200 Myr, compared to the observed profiles ofsystem 3 at z=1.75570 ( blue dashed line , Misawa et al. 2008). igure 9. Mg ii ( top row ), C iv ( middle row ), and O vi ( bottom row ) versus H i column densities in sightlines parallel to each of the threecardinal axes at t=160 ( left column ), 200 ( middle column ), and 240 Myr ( right column ) with different colors indicating Mg ii , C iv , and O vi density-weighted metallicities, to be compared to the observed Mg ii /C iv clouds by Misawa et al. 2008 ( circle ) and the observed C iv /O vi observations by Schaye et al. 2007 ( square ) and D’Odorico et al. 2016 ( star, but gray star for detection of only one member of the doublet ) .Note O vi densities from Schaye et al. 2007 ( open square ) and C iv and O vi densities from D’Odorico et al. 2016 ( open star ) are upper limits( open square ). Grey points indicate ion versus HI column density distributions expected when all the gas in our simulation is assumed tohave solar metallicity. weak Mg ii system 3 at z=1.75570 published in Misawaet al. (2008).4.2. Comparison to observations Column densities and metallicities In Figure 9, we show the relations between ion columndensities and HI column densities in our simulation att=160, 200, and 240 Myr in sightlines parallel to eachof the three cardinal axes, and compare them with theobserved relations. The colors indicate Mg ii , C iv , andO vi density weighted metallicities respectively. Effec-tive lower limits to the Mg ii , C iv , and O vi column den-sities are 3 . × , 7 . × , and 4 . × cm − inour simulations. Mg ii , C iv , and O vi absorbers in oursimulation are enriched to Z=0.1-0.2 Z (cid:12) by SNII from an instantaneous starburst, as the SNIa contribution isnegligible at this point. Top figures in Figure 9 show that sightlines withhigher metallicities have higher Mg ii column densitiesat given HI column densities, and they are comparedto the Mg ii -HI observations from three Mg ii absorbersat z ∼ . ii ab-sorbers at lower redshift (z=0.65-0.91) from Charltonet al. (2003) and Ding et al. (2005). The Mg ii columndensities in our simulation are up to 1 order of magni-tude smaller than the observed values at the given HIcolumn densities, N HI > cm − (i.e. sub-LLS). Thisis mainly because our simulation can only be run up to ∼ 300 Myr before most gas leaves the grid: metal mixingand cooling should be computed for a much longer dura-tion, also including the SNIa metal contribution. More-ver, we set the initial metallicity of our dwarf disk andhalo gas to be Z=10 − Z (cid:12) , to study the effects of metalcontribution by our simulated starburst alone. Thus, weare likely underestimating the metallicities of Mg ii ab-sorbers. If we assume that all the gas in our simulationbox has a solar metallicity, the boosted Mg ii columndensities ( grey points in Figure 9) agree more with theobserved values.At lower N HI < cm − (i.e. sub-LLS to Ly α for-est), there is no dense cloud formation in our simulationthus no Mg ii clouds with N MgII > cm − . Thereare two Mg ii absorbers with N HI < . cm − at z ∼ ii column densitiesare larger than predicted by our simulations for sight-lines with this N HI by two orders of magnitude. Thismight also be due to lower metal enrichment in our simu-lation. The estimated metallicities for the two absorbersare very high, Z=0.63-0.79 and even super solar, Z > . (cid:12) respectively. This discrepancy could also be relatedto the limited simulation resolution. We hope to studythe possible formation of super solar, weak Mg ii cloudswith our future global simulations.Simulated C iv column density distributions appear toagree better with the observed column densities of C iv absorbers that are found in the same sightlines with theMg ii absorbers studied by Misawa et al. (2008). TheseC iv absorbers are in sub-LLS environments, and havesimilar metallicities, Z=0.1-0.3 Z (cid:12) to our simulation val-ues, except for one absorber with Z=0.8 Z (cid:12) : this metalrich C iv absorber is in a structure related to the supersolar, weak Mg ii absorber with Z > . (cid:12) .On the other hand, our simulated C iv column densi-ties are smaller than those of the C iv absorbers studiedby Schaye et al. (2007): the disagreement is by an orderof magnitude. This is probably because these absorbersare selected for the high metallicities, Z ∼ Z (cid:12) . Theyare found in Ly α forest environments and are smaller insize ( ∼ 500 pc – 1 kpc). In our simulation, smaller C iv clouds are found in region (b) and arise from the sameclouds that currently host or used to host even smaller,weak Mg ii absorbers in sub-LLS to Ly α forest environ-ments. Our metallicity boosted values better agree withthe observations (Figure 9). The upper limits for O vi column densities associated with the observed C iv ab-sorbers (Schaye et al. 2007) are also above what our sim-ulation predicts, and lie in the metallicity boosted greyarea, just like most of the observed weak Mg ii and C iv absorbers. There is no other information about theirphysical properties available.The observed C iv column densities by D’Odorico et al.(2016) appear to agree with our simulated values at N HI > . cm − , however, they are much lower than our simulated values, by up to one order of magnitude,at N HI < . cm − . These C iv absorbers are ob-served at a higher redshift, z ∼ . 8, and the majority ofthem have their estimated metallicities between 10 − . Z (cid:12) and 10 − Z (cid:12) , much lower than our simulated val-ues. There is no information about the physical proper-ties available for the C iv and O vi absorbers observed byD’Odorico et al. (2016). The data for O vi column den-sities are mostly upper limits except three detectionsof which one shows a very weak C iv line and anothershows none. Out of 15 O vi possible detections with sin-gle lines, six of them do not show an associated C iv line.Despite the estimated low metal contents, the observedO vi column densities and their upper limits appear toagree better with our simulated values at all HI environ-ments. The sizes and thermal properties are unknownfor these C iv and O vi absorbers.We note that the observed estimates and upper limitsfor C iv and O vi column densities at given HI columndensities vary over 4 orders of magnitudes. This maybe due to the presence of HI dominated gas in observedsightlines which originate in regions that are not cov-ered by our simulations. However, for Mg ii absorbesand associated C iv absorbers, a major reason for thediscrepancy seems to be a lack of metal enrichment aswell as the low initial metallicity of disk and halo gasin our simulation. We speculate that galactic outflowsfrom repeated bursts of star formation for a longer du-ration ( ∼ Covering fractions Figure 10 shows fractions of sight lines that occupyour simulation box above the galactic disk and withinthe virial radius as functions of Mg ii , C iv , and O vi col-umn densities along x, y, an z axes at three differenttimes. The Grey region in Figure 10 depicts predictedfractions of sight lines as a function of column densitiesof the observed weak Mg ii absorbers at various redshiftsby Rigby et al. (2002); Charlton et al. (2003); Ding et al.(2005); Misawa et al. (2008); Narayanan et al. (2008),based on an assumption that they cover ∼ − 30% ofa halo. The total covering fraction of weak Mg ii ab-sorbers in L ∗ galactic haloes is estimated to be ∼ ii absorbers, if undetected, satellite dwarf galaxies are re-sponsible for producing weak Mg ii absorbers in a L ∗ halo. .20.40.60.81.0 t=160Myr xyzx (Z )y (Z )z (Z ) F r a c t i o n s o f s i g h t li n e s w i t h > N t=200Myr log N(MgII) t=240Myr Weak MgII absorbers 12 13 14 15 log N(CIV) 12 13 14 15 log N(OVI) Figure 10. The Mg ii ( left ), C iv ( middle ), and O vi ( right ) covering fractions as functions of column densities along each of thethree cardinal axes at t=160 ( top ), 200 ( middle ) and 240 Myr ( right ). All sightlines between z=2.5 kpc (the disk edge) and 17.5kpc (virial radius) are included. The dasremote.net.ed.ac.ukhed lines show the covering fractions when all the gas is assumed tohave solar metallicities. The grey region indicates estimated fractions of sight lines as a function of Mg ii column densities whenwe assume that the observed weak Mg ii clouds at various redshifts (Rigby et al. 2002; Charlton et al. 2003; Ding et al. 2005;Misawa et al. 2008; Narayanan et al. 2008) cover 5–30% of a halo. The observed Mg ii column densities are ≥ cm − . Weak Mg ii absorbers with column densities greaterthan the observed minimum, ∼ cm − occupy aboutonly f MgII ∼ 5% of the dwarf halo in our simulation.However, this is a lower limit for the covering fractionbecause 38 − 58% of SNII outflow gas leaves the boxby t = 40 − 300 Myr. Boosting the metallicities of allthe gas to 1 Z (cid:12) (see dashed lines in Figure 10) raisesthe fractions of sight lines with N MgII (cid:38) cm − to f MgII ∼ ii clouds with higher column densities, (cid:38) cm − . Mostobserved weak Mg ii absorbers have column densities (cid:38) cm − . As we argued in the previous section,repeated bursts of star formation will likely create moreclumps and filaments, like the brightest structures inFigure 5, through cycles of phase 1 and phase 2 forma-tion. Then, a larger fraction of the dwarf galaxy halomay be covered with moderately dense Mg ii absorbers.However, the formation of denser, high column density,weak Mg ii clouds may require other mechanisms thatinvolve more gas and more metals with stronger shocks,as the shell density scales like the square of the Machnumber in the isothermal shocks expected, so more pow-erful outflows may be responsible for the higher col-umn density Mg ii absorbers. In addition, interactionof outflows with cosmological infall will likely producestronger shocks, so possibly denser clouds. Note we have a static background in our simulations. However, cool-ing is strongly limited by numerical resolution, so gasbehind isothermal shocks will not cool to the theoreticaldensity (n M ) even with 0.2 pc resolution Fujita et al.(2009).We can estimate the number density of weakMg ii absorbers per unit comoving path length to be dN MgII /dX ≈ f MgII ∼ 5% for N MgII ≥ cm − when metallicity is boosted to Z=Z (cid:12) , 1.13Mpc − for halo comoving number density with M halo ≥ × M (cid:12) at z=2 (Murray et al. 2013), and π (17 . − . ) kpc for halo proper cross section. This yields avalue comparable to dN MgII /dX = 0 . 33 at 1 . < z < . dN MgII /dX =0 . 41 at < z > =2.34 by Codoreanu et al. (2018). Like-wise, the number density of high ionization clouds (C iv and O vi ) per unit comoving path length is estimated tobe dN CIV /dX ≈ dN OV I /dX ≈ f CIV = f OV I ∼ dN CIV /dX ≈ ∼ iv and O vi ions are measured to be0.3-0.8 at impact parameters (cid:46) ∼ ii mass densityeems to increase nearly a factor of 10 from < z > =2.34 to < z > =4.77 (Codoreanu et al. 2018) with a large numberof weak Mg ii absorbers even up to z ∼ ii absorbers suggeststhat they are associated with dwarf galaxies, includingsmaller, numerous galaxies during the epoch of reion-ization, and the presence of the abundant weak Mg ii absorbers must be explained without more powerful out-flows from larger galaxies.We assess this as follows: 1) a SNII driven outflow islaunched from a star cluster every 100 Myr, the time bywhich gas flows back to the central source region in oursimulation, and it takes 50 Myr for a SNII driven outflowwith v=200-400 km s − to reach the shocked enrichedgas from previous outflows (region b). 2) SNIa drive asuperbubble and an outflow after SNII stop in 50 Myr(we choose 50 instead of 40 Myr for simplicity), and ittakes 100 Myr for a SNIa driven outflow to reach region(b) based on our simulation result. 3) repeated burstslast for 1 Gyr. 4) interaction from a newly launchedoutflow produces weak Mg ii absorbers that cover 3-6%of our dwarf halo and those weak Mg ii absorbers sur-vive for at least 150 Myr based on our simulation result.Then, we estimate that the covering fraction of dwarfhalos by weak Mg ii absorbers will be 12-24%. This num-ber should go up once the CGM is more metal enriched,because the covering fraction of 3-6% is computed whenmetallicities of absorbers are Z=0.1-0.2 Z (cid:12) . We hope totest this hypothesis with our future global simulation ina larger box with repeated bursts in time and place. RESOLUTION STUDYOur standard simulation employs a highest resolu-tion of 12.8 pc with four refinement levels, thus resolves ∼ 100 pc structures for our purposes. We base theestimate of roughly eight cells being required to mini-mally resolve structures on two arguments. First, thenumerical dissipation range for supersonic turbulencecomputed with Enzo extends over almost an order ofmagnitude (e.g. Kritsuk et al. 2007, Figure 5), similarto most other grid codes (Kitsionas et al. 2009). Second,modeling of a cloud in a supersonic flow shows that aradius of six zones using a second-order method is in-sufficient to capture fragmentation by instabilities (MacLow & Zahnle 1994, Figure 4).To study the extent to which the production of clumpsand filaments as well as their sub-structures and frag-mentation are dependent on numerical resolution, weran the same simulation with 3 refinement levels ( low-res simulation), and by applying 5 refinement levels in a re-gion where the largest filaments form at [∆x,∆y,∆z]=[(-0.5 kpc, 3.28 kpc), (-0.5 kpc, 3.28 kpc), (10 kpc, 15 kpc)] ( high-res zoom simulation). We only ran the high-reszoom simulation up to t=200 Myr.Figure 11 shows phase 1 formation of filaments andclumps computed with the three different resolutions.Figure 11 compares the degrees of fragmentation in high-res zoom and our standard simulations. In the high-reszoom simulation, gas fragments into thinner filamentsand smaller clouds compared with our standard simu-lation. The smallest structures are resolved across ∼ ∼ 50 pc in the high-res zoom simula-tion compared with ∼ 100 pc in our standard simulation.These filaments and clumps will further fragment intosmaller pieces with higher resolution. Gas structuresseem drastically different in the low-res simulation, withmuch larger clouds compared with the higher resolutionruns.Despite the differences in fragmentation seen in simu-lations with different resolutions, there is no significantdifference in projected Mg ii distributions ( bottom fig-ures). We see no change in the fraction of weak Mg ii absorbers with high column densities, and the coveringfractions of weak Mg ii absorbers as well as C iv and O vi absorbers remain practically the same. We will add ionversus H i column density distributions with metallicities(Figure 13) and ion covering fractions as functions of col-umn densities (Figure 14) for the low-res simulation inthe appendix.We conclude that resolution has a small effect on theprojected distribution of dense clumps and filaments,and this in turn warns us to be very careful when weinterpret observations (Peeples et al. 2019). There is apossibility that, at much higher resolution filaments andfragments will further ”shatter” into ∼ pc sized cloudletsthat agree with some of the observed Mg ii absorbers(Gronke & Oh 2018; McCourt et al. 2018; Gronke &Oh 2020; Begelman 1990), but to test this possibilityrequires 1/10th pc resolution in a galactic-scale simu-lation which is not feasible at the moment. Opposingfragmentation is possible cloud coalescence, which maybe numerically challenging to reproduce in 3D simula-tions (Waters & Proga 2019). SUMMARYIn this paper, we use hydrodynamical simulations ofgalactic outflows to explore the production of weak Mg ii absorbers and C iv and O vi absorbers in the CGM of adwarf satellite galaxy with a halo mass of 5 × M (cid:12) at z = 2, expected to be hosted by a larger L ∗ halo.With our standard numerical resolution of 12.8 pc, wemodel the formation of superbubbles and outflows froma galactic disk assuming a single instantaneous starburstin a simulation box with dimensions (6.5536, 6.5536, z ( k p c ) t=200 Myrx=6.4 pc x=12.8 pc x=25.6 pc10111213141516 z ( k p c ) y (kpc) z ( k p c ) y (kpc) y (kpc) D e n s it y [ c m ] M g II d e n s it y [ c m ] M g II c o l u m n d e n s it y [ c m ] Figure 11. Sliced density ( top ) and Mg ii density ( middle ) distributions at x=+2.4 kpc from the disk center and projected Mg ii distributions ( bottom ) along the x -axis, all in the y-z plane, at phase 1 (t=200 Myr), resolved with highest resolutions of 6 . white rectangle ), 12.8 (our standard simulation), and25.6 pc ( from left to right ). Regions enclosed in cyan rectangles are shown in Figure 12. y (kpc) z ( k p c ) t=200 Myrx=6.4 pc 1.0 1.5 2.0 2.5 3.0 y (kpc) x=12.8 pc 10 D e n s it y [ c m ] Figure 12. The same as top figures in Figure 11, but showing only regions enclosed in cyan rectangles for 6.4 ( left ) and 12.8( right ) resolutions. ∼ 300 Myr, as mostmetal enriched gas leaves the simulation box by thistime, our results highlight the possibility of dwarf galac-tic outflows producing transient, but continuously gen-erated Mg ii clouds as well as larger C iv and O vi cloudsin sub-LLS and Ly α forest environments.Our main findings are: • Thin, filamentary, weak Mg ii absorbers are pro-duced in two stages: – phase 1 shocked SNII enriched gas loses en-ergy and descends toward expanding SNII en-riched gas and is shocked, cools, and frag-ments. – phase 2 SNIa driven outflow gas shocks theSNII enriched gas as well as phase 1 shells,which then cool and fragment.The width of the filaments and fragments are (cid:46) 100 pc with our standard numerical resolu-tion. A single Mg ii cloud survives for ∼ 60 Myr,but we suggest Mg ii absorbers will continuouslybe produced through cycles of phase 1 and phase2 formation for > 150 Myr by repeated bursts ofstar formation. • C iv absorbers are produced in expanding SNII en-riched gas (region a) and shocked SNII enrichedgas (region b). C iv absobers in region (a) extendover 1–4 kpc and C iv absorbers in region (b) aresmaller, 500 pc–1 kpc, but they are both cool andphotoionized. The smaller C iv absorbers originatefrom the same clouds that produce weak Mg ii ab-sorbers, and they surround the dense Mg ii clouds.As the clouds get destroyed and mixed with thesurrounding gas, Mg ii absorbers disappear first,but C iv absorbers survive for another 20–30 Myr. • O vi absobers are also produced in expanding SNIIenriched gas in region (a) and shocked SNII en-riched gas in region (b). O vi absorbers in region(a) originate from the same cool clouds that pro-duce C iv absorbers, but O vi absorbers in region(b) are not coincident with Mg ii absorbers or C iv absorbers. Their sizes are over (cid:38) • C iv absorbers and most O vi absorbers are cool,photoionized clouds while O vi absorbers arisingin turbulent mixing layers in region (b) are hot-ter and collisionally ionized. Photoinization dom-inates in sub-LLS and Ly α environments at inter-mediate redshft . • The metallicities of Mg ii , C iv , and O vi absorbersare Z=0.1–0.2 Z (cid:12) by t= ∼ phase 1 and phase 2 formation in adwarf disk and halo with a low initial metallicity,Z=0.001 Z (cid:12) . We speculate that the clouds form-ing in shocked outflow gas (region b) will be pro-gressively enriched with more metals when burstsof star formation are repeated. • A lower limit for the covering fraction of weak Mg ii absorbers in our dwarf halo is 3–6% because wecompute only one cycle of phase 1 and phase 2 formation and more than half the metal enrichedgas leaves the simulation box early. To reproducethe observed estimate for the covering fraction ina L ∗ halo (30%) with outflows from such galaxiesalone, sightlines must go through haloes of mul-tiple dwarf satellite galaxies. We also speculatethat the covering fraction in a single dwarf halowill be boosted with repeated bursts with manycycles of phase 1 and phase 2 formation in a largesimulation box that covers the entire halo.There are two major problems in our current simula-tions: 1) a deficiency of weak Mg ii absorbers with highcolumn density, (cid:38) cm − and 2) low metallicity ofweak Mg ii absorbers.1) The formation of denser, high column, weak Mg ii absorbers may require stronger shocks with more pow-erful outflows and/or outflows interacting with dynamicinfall. Repeated outflows shocking the clumps and fil-aments formed by previous outflows may also producedenser Mg ii clouds. However, cooling behind shocks islimited by numerical resolution. Our resolution studyshows that the sizes of filaments and fragments decreaseby a factor of two with a resolution twice as high, how-ever, the projected properties are insensitive to changesin resolution that is (cid:29) pc. We may need less than a pcscale resolution to address this problem.2) The metallicity, less than solar, of our Mg ii ab-sorbers is the result of our assumption of a single in-stantaneous starburst and the limited duration of oursimulations ( ∼ 300 Myr) neglecting the SNIa metal con-tribution. Starting with a higher initial metallicity forour dwarf disk and halo gas will also alleviate the prob-lem.This paper nonetheless highlights the possibility ofgalactic outflows from invisible, dwarf satellite galaxiesto produce highly enriched, multiphase gas. We hopeto address the remaining problems with our next, moreglobal simulations.CKNOWLEDGMENTSThis work was supported by the Grants-in-Aid forBasic Research by the Ministry of Education, Scienceand Culture of Japan, Grant Number 19K03911. Weacknowledge use of the Cray XC50 at the Center forComputational Astrophysics (CfCA) of the National As-tronomical Observatory of Japan (NAOJ). J.C.C. ac-knowledges support by the National Science Founda-tion under grant No. AST-1517816. AM acknowledgessupport from UK Science and Technology FacilitiesCouncil, Consolidated Grant ST/R000972/1. M-MMLwas partly supported by US NSF grant AST18-15461.Computations described in this work were performedusing the publicly available Enzo code (http://enzo-project.org), which is the product of a collaborative ef-fort of many independent scientists from numerous in-stitutions around the world. Their commitment to openscience has helped make this work possible. Facilities: CfCA(NAOJ) Software: Enzo (Bryan et al. 2014), yt (Turk et al.2011) Trident (Hummels et al. 2017), SYGMA (Ritteret al. 2018)PPENDIX Figure 13. Same as Figure 9 for low-res simulation (25.6 pc) t=160Myr x (12.8 pc)yzx (25.6pc)yz F r a c t i o n s o f s i g h t li n e s w i t h > N t=200Myr log N(MgII) t=240Myr 12 13 14 15 log N(CIV) 12 13 14 15 log N(OVI) Figure 14. Same as Figure 10 but comparing the results in in our standard simulations (12.8 pc solid ) and low-res simulation (25.6 pc dashed ). igure 15. Same as Figure 9 but without overcooled gas (n H ≤ − cm − , T < K) in our standard simulation. The overcooled,low-density gas is metal-enriched outflow gas in region a. With or without it, there is very little change for Mg ii and C iv distributions,while there is a marginal difference in the distribution of higher metallicity O vi systems. t=160Myr xyzx (Z )y (Z )z (Z ) F r a c t i o n s o f s i g h t li n e s w i t h > N t=200Myr log N(MgII) t=240Myr Weak MgII absorbers 12 13 14 15 log N(CIV) 12 13 14 15 log N(OVI) Figure 16.