Probing the Physical Conditions of Atomic Gas at High Redshift
aa r X i v : . [ a s t r o - ph . GA ] N ov Draft version September 8, 2018
Preprint typeset using L A TEX style emulateapj v. 5/2/11
PROBING THE PHYSICAL CONDITIONS OF ATOMIC GAS AT HIGH REDSHIFT
Marcel Neeleman
Department of Physics and Center for Astrophysics and Space Sciences, UCSD, La Jolla, CA 92093, USA
J. Xavier Prochaska
Department of Astronomy & Astrophysics, UCO/Lick Observatory, 1156 High Street,University of California, Santa Cruz, CA 95064, USA andArthur M. Wolfe
Department of Physics and Center for Astrophysics and Space Sciences, UCSD, La Jolla, CA 92093, USA
Draft version September 8, 2018
ABSTRACTA new method is used to measure the physical conditions of the gas in damped Lyman- α systems(DLAs). Using high resolution absorption spectra of a sample of 80 DLAs, we are able to measure theratio of the upper and lower fine-structure levels of the ground state of C + and Si + . These ratios aredetermined solely by the physical conditions of the gas. We explore the allowed physical parameterspace using a Monte Carlo Markov Chain method to constrain simultaneously the temperature,neutral hydrogen density, and electron density of each DLA. The results indicate that at least 5 %of all DLAs have the bulk of their gas in a dense, cold phase with typical densities of ∼
100 cm − and temperatures below 500 K. We further find that the typical pressure of DLAs in our sampleis log( P/k B ) = 3.4 [K cm − ], which is comparable to the pressure of the local interstellar medium(ISM), and that the components containing the bulk of the neutral gas can be quite small withabsorption sizes as small as a few parsec. We show that the majority of the systems are consistentwith having densities significantly higher than expected from a purely canonical WNM, indicatingthat significant quantities of dense gas (i.e. n H > − ) are required to match observations.Finally, we identify 8 systems with positive detections of Si II ∗ . These systems have pressures ( P/k B )in excess of 20000 K cm − , which suggest that these systems tag a highly turbulent ISM in young,star-forming galaxies. Subject headings: galaxies: abundances — galaxies: evolution — galaxies: ISM — ISM: atoms —ISM: evolution — quasars: absorption lines INTRODUCTIONGalaxy formation and evolution is fundamentally de-pendent on the gas that forms the galaxy. From its in-ception as a gravitational bound collection of gas to theformation of stars inside an evolved galaxy, the physi-cal properties of the gas affect the outcome of such pro-cesses. It is therefore of paramount importance to un-derstand the physical properties of the gas inside andaround galaxies.Already early on, Field et al. (1969) noted that whenheating and cooling sources of a neutral gas are in ther-mal equilibrium, the gas naturally segregates into twodistinct phases. A cold, dense phase known as the coldneutral medium (CNM), and a warmer, less dense phase,the warm neutral medium (WNM). This model was im-proved upon by McKee & Ostriker (1977) to include athird phase, namely the hot ionized medium due to ion-izing shock fronts produced by supernova. On the basisof this model, Wolfire et al. (1995, 2003) calculated theequilibrium pressures and temperatures of the neutralgas under a variety of different galactic conditions. Thistheoretical model, in its general form, still holds as theparadigm for the physical conditions of neutral galactic [email protected] gas.Observationally, the validity of this model has beentested for gas in the local universe. The observationalstudies range over a large part of the electromagneticspectrum from X-ray (Snowden et al. 1997) to radio(Heiles & Troland 2003a). The results suggest that in-deed some of the gas has properties of both the CNMand WNM. However, a large fraction of the WNM isactually found to be in the temperature region between500 K and 5000 K (Heiles & Troland 2003b; Roy et al.2013a,b). One possible explanation for the existenceof this gas in what is known as the ‘forbidden region’comes from numerical simulations, which show that tur-bulence could produce the observed gas characteristicswhile still locally satisfying thermodynamic equilibrium(Gazol et al. 2005; Walch et al. 2011).To understand the evolution of galaxies, it would beideal to measure the properties of the gas over a rangeof redshifts and physical conditions. This, however, isdifficult to do because the methods used at low redshiftare not feasible for distant galaxies. In particular, 21 cmline emission has only been detected in galaxies up to z ∼ Table 1
Fine Structure DLA sampleIndex QSO z abs log N HI Metallicity M a log N CII ∗ log N SiII ∗ log N SiII
N13 b ReferencesNumber (cm − ) [M/H] (cm − ) (cm − ) (cm − )1 Q1157+014 1.9437 21.70 ± − . ± > . c . ± . ± ± − . ± < . < .
56 15 . ± −
02 2.0395 21.65 ± − . ± > . c < .
58 16 . ± c Y 4, 74 J2340 − ± − . ± . ± < .
31 15 . ± −
19 2.0762 20.43 ± − . ± < . < .
83 13 . ± References . — (1) Lu et al. (1996);(2) Prochaska & Wolfe (1997);(3) Lu et al. (1998);(4) Prochaska & Wolfe (1999);(5) Petitjean et al.(2000);(6) Prochaska & Wolfe (2000);(7) Prochaska et al. (2001a);(8) Levshakov et al. (2002);(9) Prochaska & Wolfe (2002a);(10) Prochaska et al.(2003a);(11) Dessauges-Zavadsky et al. (2004);(12) Khare et al. (2004);(13) Ledoux et al. (2006);(14) O’Meara et al. (2006);(15) Herbert-Fort et al.(2006);(16) Dessauges-Zavadsky et al. (2007);(17) Prochaska et al. (2007);(18) Wolfe et al. (2008);(19) Jorgenson et al. (2010);(20) Kaplan et al.(2010);(21) Rafelski et al. (2012);(22) Kulkarni et al. (2012);(23) Berg et al. (2013);(24) This Work;(25) Berg et al. (2014)
Note . — (This table is available in its entirety at the end of this manuscript.) a Ion used for metallicity determination b Part of the Neeleman et al. (2013) sample c VPFIT used to determine the column density largest H I gas column densities are known as dampedLyman- α systems (DLAs; for a review see Wolfe et al.2005a). DLAs have neutral hydrogen column densities( N H I ) equal or greater than 2 × cm − , and are likelyassociated with galaxies as is suggested by both observa-tions (e.g Wolfe et al. 2005a) and numerical simulations(e.g. Fumagalli et al. 2011; Cen 2012; Bird et al. 2014).Observational studies of DLAs have focussed mainlyon line-of-sight column density measurements. Althoughsuch studies are able to measure quantities such as themetallicity (e.g. Rafelski et al. 2012) and the velocitystructure of the absorber (e.g Neeleman et al. 2013),these studies are unable to provide detailed informationon the physical conditions of this gas such as the temper-ature and neutral hydrogen density. Several innovativemethods have been devised to measure exactly these pa-rameters for high redshift absorbers. The first methodis to measure 21 cm line absorption in DLAs in front ofradio-loud quasars. The integrated optical depth of the21 cm absorption and the measured H I column den-sity will yield the spin temperature of the associatedgas (see Kanekar et al. 2014, for a detailed descriptionof this method and results). A second method is to mea-sure the fine structure lines of neutral carbon, whose ra-tio is dependent on the physical conditions of the gas(Srianand et al. 2005; Jorgenson et al. 2010).On the theoretical side, Wolfe et al. (2003a) extendedthe work of Wolfire et al. (1995) to the physical condi-tions pertinent to DLAs. Under the same assumptionsas before, the gas in DLAs forms a two phase-medium,albeit at somewhat different density and temperatures.Observational measurements of high redshift DLAs showthat indeed some of the gas has properties similar toboth the CNM (Howk et al. 2005; Srianand et al. 2005;Carswell et al. 2010; Jorgenson et al. 2010) and WNM(Lehner et al. 2008; Carswell et al. 2012; Kanekar et al.2014; Cooke et al. 2014). However, disagreement lieswith the percentage of DLAs that contain a significantfraction of CNM. Based on several observational results,Wolfe et al. (2003b, 2004) claim that the star forma-tion rate per unit area is too large for current obser-vational constraints if DLAs occur solely in a WNM(see also Fumagalli et al. 2014). On the other hand the21 cm absorption studies suggest that at least 90 % of DLAs contain a large fraction of WNM (Kanekar 2003;Kanekar et al. 2014).To address this issue and shed additional light on thephysical conditions of gas probed by DLAs, we applyin this paper a third method. This method was firstdescribed by Howk et al. (2005) for DLAs. It relies onthe fact that the ratios of the fine-structure levels of theground states of C + and Si + are solely determined by thephysical parameters of the DLA (see also Silva & Viegas2002). Therefore a measurement of these ratios allows fora determination of the physical parameters of the DLA.This method has several advantages. Unlike C I , bothSi II and C II are the dominant ionization states of theseelements, and therefore they very likely trace the bulk ofthe neutral gas. Furthermore, unlike the 21 cm method,this method does not use a radio source, which couldprobe different gas, as the radio source need not be ascompact as the ultraviolet or optical source (Wolfe et al.2003b; Kanekar et al. 2014).This paper is organized as follows. In Section 2, theselection of the sample used in this paper is explained. InSection 3 we describe the measurements from the obser-vations and literature sample. Section 4 explains in detailthe method used in this paper to measure the physicalparameters of the DLA. The results are tabulated anddescribed in Section 5. Finally we discuss these resultsin Section 6 and summarize them in Section 7. SAMPLE SELECTIONTo apply the method described in this paper, we re-quire accurate measurements of the column density ofthe two fine structure levels of the ground state of bothC + and Si + . We will denote the upper level of the groundstate by an asterisk (e.g. Si II ∗ ), whereas the lower levelwill be represented by the standard notation (e.g. Si II ).To limit saturation issues and to enable individual com-ponent analysis, we restrict ourselves to high resolutiondata. In particular, we limit ourselves to data from thehigh resolution spectrograph (HIRES; Vogt et al. 1994)on the Keck I telescope, which resulted in spectra witha typical resolution of ∼ − . We further requirethat at least one of the transitions of both levels of theC + and Si + are clear of any forest lines or interlopingfeatures. In practice this means selecting those spectrahe Physical Conditions of High- z Atomic Gas 3which have clear spectral regions around C II ∗ λ II ∗ λ II λ II ∗ λ II lines to determine the col-umn densities of the Si + fine structure lines. We do notdirectly measure the C II lower state because the C II λ Q1157+014 −50 0 5001 J1056+1208 −100 −50 0 50 10001 J1142+0701 −40 −20 0 20 4001 J1313+1441 −50 0 50 100 15001 J1417+4132 −50 0 50 100 15001 R e l a t i v e F l u x Q1755+578 J1310+5424 −50 0 50 10001 Velocity (km/s)
Figure 1. (top) Si II ∗ , (middle) C II ∗ and (bottom) Si II transi-tions for the 7 DLAs with detections of Si II ∗ in our sample. Thedark gray (blue) vertical lines indicate the central position of thevelocity components used for the line fit from VPFIT shown by thesolid (red) line. In the top panel, the light gray (orange) verticallines indicate the position of the weaker Si II ∗ λ II ∗ transition is strongly saturated (orblended in the case of Q1755+578). We therefore take these fitsas lower limits (see text). For the analysis discussed in this paper,we only consider the sum of the individual components, exceptfor Section 5.3 in which we discuss the component-by-componentanalysis. ments. These DLAs can in general be divided intotwo categories, those that where selected based onthe strength of the metal lines in their Sloan Digi-tal Sky Survey (SDSS; Abazajian et al. 2009) spectra(Herbert-Fort et al. 2006), and those selected purelyby their H I column density. We assume, as inNeeleman et al. (2013), that the latter subset is a lessbiased sample and represents an accurate subsample ofthe general DLA population. The metal line selectedsample, on the other hand, is likely biased towards highermetallicity systems and may trace the more massive hostgalaxy halos (Neeleman et al. 2013). In Table 4 we havemarked the 48 DLA that are part of the ‘unbiased’ sam-ple described in Neeleman et al. (2013). MEASUREMENTSThis section describes the measurements taken for eachabsorber. The measurements for all of the DLAs in thesample are tabulated in Table 4.3.1. H I Column Density Measurements
The H I column density of the DLAs is measuredby adopting the procedure outlined in Prochaska et al.(2003a). We determine the H I column density, N H I , ofan absorber in a quasar spectrum by simultaneously fit-ting the continuum of the background quasar and fittinga Voigt profile to the Ly- α line of the absorber. Thismethod provides accurate column density measurementsif the continuum can be accurately placed. The mea-surements and their uncertainties are displayed in Table4. 3.2. Metallicity Measurements
For each of the DLAs found, we have measured themetallicity, defined by:[M / H] = log (N M / N H ) − log (N M / N H ) ⊙ (1)The column density of the metals was found us-ing the apparent optical depth method (AODM;Savage & Sembach 1991), where we have used the wave-lengths and oscillator strengths from Morton (2003) andthe solar abundances from Asplund et al. (2009). Weapply the same procedure as outlined in Rafelski et al.(2012), to determine which metal to use as a tracerof the metals in a DLA. In particular, we avoid us-ing Fe as a metal tracer for the DLAs chosen by theirmetal lines, as Fe is more depleted at higher metallicity(Prochaska & Wolfe 2002a; Ledoux et al. 2003; Vladilo2004; Rafelski et al. 2012). In Table 4 we list all of themetallicities of the DLAs and the line used for the deter-mination of the metallicity.3.3. Column Density Measurements
For unsaturated lines and slightly saturated lines, theAODM provides an accurate way of determining the col-umn density of the metals in an absorber. However, whena line is slightly blended or strongly saturated, the resul-tant limits for the column density found using the AODMare very conservative. Stronger constraints to the col-umn density for such metal lines can be found by fittingthe lines with χ Voigt profile fitting routines such as
VPFIT . Since in all cases Si II ∗ is unsaturated; we only Neeleman et al. −5−4−3−2−10
Log r C Log n H = 0(a) Log n H = 2 Log r S i Log n e = −4Log n e = −2Log n e = −1Log n e = 0 −5−4−3−2−10 Log r C Log n e = −3(b) Log n e = −1 Log r S i Log n H = −1Log n H = 0Log n H = 1Log n H = 2 Figure 2.
Si and C ratios for a range of different physical parameters. Panel (a) shows how both r C = n C II ∗ n C II and r Si = n Si II ∗ n Si II vary withtemperatures under a variety of different electron densities as the neutral hydrogen density is held constant. In panel (b) the electron densityis held constant while the neutral hydrogen density is varied. These plots show that at low neutral hydrogen density, the fine-structurelevel ratios of C + and Si + are strongly dependent on the electron density, whereas at higher neutral hydrogen densities the ratios onlydependent on the neutral hydrogen density. use VPFIT for our measurements of saturated C II ∗ lines,and for Si II in 8 systems where all the available Si II lines could potentially be saturated.To fit the saturated lines, we select an unsaturated low-ion line such as Si II or Zn II to determine the redshiftsof the components of the low-ion lines. We then tie theDoppler parameters of the C II ∗ and low-ion lines by as-suming that they arise solely from thermal broadening.This assumption is unphysical as turbulent broadeningis likely important for this gas. However, we are after aconservative lower limit to the column density indepen-dent of the physical model used. Since the low-ion usedfor selecting components is heavier than the fitted ion,a thermally linked gas will provide a conservative lowerlimit to the column density. We finally require that therelative number of components is equal across the speciesand allow only the total column density of the saturatedline to change. The lower limits measured using thismethod are marked in Table 4.The above methods are used to determine the col-umn densities of Si II , Si II ∗ , and C II ∗ . However, wecannot directly measure the C II column density be-cause in all cases the resonance lines of C II are toosaturated. We instead use the column density of Sias a proxy for C. Here we assume that both Si andC are not depleted onto dust grains and that C likelytraces Fe (Wolfe et al. 2003a). Under these assumptions,[C/H] gas =[Fe/H] int =[Si/H] gas +[Fe/Si] int , where we take[Fe/Si] int , which is the intrinsic alpha enhancement of theDLA gas, to be − . II and C II are the domi-nant ionization state in the gas, N (C) ≈ N (C II ) and N (Si) ≈ N (Si II ). Hence the column density of C II isassumed to be: log N (C II )=log N (Si II )+0 . II ∗ from a sample of 79 DLAs. None of these detec-tions have been analyzed previously, although two of these detections have been mentioned in the literature(i.e. J1417+4132 (Berg et al. 2013), and Q1755+578(Jorgenson et al. 2010)). The detection of 7 new Si II ∗ measurements is noteworthy, because this state is onlyvery rarely seen in DLAs along quasar sight lines (QSO-DLAs), yet is seen regularly in DLAs detected in gammaray bursts (GRB-DLAs). Here the GRB is likely re-sponsible for optically pumping the excited fine struc-ture state (Prochaska et al. 2006). Together with therecent analysis of Si II ∗ in QSO-DLA J1135 − II ∗ in QSO-DLAs. Forcompleteness, we have therefore included J1135 − II ∗ measurements are shown inFigure 1. We have also plotted a representative low-ionline to show the similarity in velocity structure betweenthe Si II ∗ line and the low-ion lines, which is discussedfurther in Section 5.3. METHODIn this section we detail the method used in this paperto determine the physical parameters of the gas using theSi II ∗ and C II ∗ fine-structure lines. We further describehow this method is applied to all of the DLAs in oursample. 4.1. Si II ∗ and C II ∗ Technique
The technique of using Si II ∗ and C II ∗ to determine thegas temperature of DLAs was first used by Howk et al.(2005). Here we will describe the technique. The P fine-structure states of Si II and C II can be well approxi-mated by a two-level atom for temperatures below 30,000K as electrons are unable to excite the atom through col-lisions to higher energy levels (Silva & Viegas 2002). Assuch we can write the steady state equation for both Si II and C II as: n n = B u ν ( z ) + Γ + Σ k n k γ k A + B ∗ u CMB ( z ) + Γ + Σ k n k γ k (2)he Physical Conditions of High- z Atomic Gas 5Here n refers to the P / state of Si II and C II (i.e.n(Si II ∗ ) and n(C II ∗ )), and n refers to the lower level P / state of these atoms (i.e. n(Si II ) and n(C II )). A , B and B are the Einstein coefficients for thegiven transitions, u CMB is the energy density of the cos-mic microwave background radiation field and Γ andΓ are the fluorescence rates. We assume the fluo-rescence rates are negligible because of the opacity ofthe ground-state transitions (see e.g. Sarazin et al. 1979;Wolfe et al. 2003b), and because of the lack of Fe + ex-cited fine-structure lines in the DLAs (Prochaska et al.2006).Finally, the excitation and de-excitation terms due tocollisions (i.e. the n k γ k terms) are considered. Theseterms are proportional to the number density of thespecies and the collision rate with that species. In thecase of DLAs, we consider collisions with electrons, pro-tons, and atomic hydrogen. The fraction of molecular hy-drogen is assumed to be small for DLAs (Jorgenson et al.2010; Ledoux et al. 2003), such that we can ignore colli-sions with this species. All of the collision rates are takenfrom Silva & Viegas (2002), and references therein.Considering all these processes, the ratios of the upperto lower fine-structure levels of the ground state of C + and Si + become a function of redshift ( z ), temperature(T), neutral hydrogen density (n H ), and electron density(n e ). Since the redshift of the DLA is well-determinedfrom the metal lines; this leaves the three internal pa-rameters that set the two ratios. Figure 2 shows thedependence of the carbon ratio ( r C = n C II ∗ n C II ) and siliconratio ( r Si = n Si II ∗ n Si II ) on temperature for a variety of dif-ferent hydrogen and electron densities. The ratios cor-relates strongly with temperature for temperatures be-low 500 K. Furthermore, at low hydrogen densities (n H .
10 cm − ) collisions with electrons dominate. Hence theratios show a strong correlation with electron density.On the other hand, at large hydrogen densities (n H &
10 cm − ) the collisions with neutral hydrogen dominateand the ratios show a strong correlation with the neutralhydrogen density.Finally, we note that the observable that we measureis the ratio of the column densities, not the ratio of theactual densities. However, the two quantities are relatedby: N N ≈ R n d s R n d s = R n n n d s R n d s (3)The ratio of the two column densities is therefore simplythe metal density weighted average of the density ratioover the path length. Since d N ≈ n ds, the ratio of col-umn densities is also approximately equal to the mean ofthe density ratio weighted by the metal column density ofthe individual components. Under the assumption thatthe amount of neutral gas in the individual components iscorrelated to the column density of the components (i.e.the metallicity of the individual components is similar),then the measured ratio of the column densities will beequal to the mean of the density ratio weighted by theamount of neutral gas in each component.We assume for this paper that the observed ratio ofthe column densities provides a good estimate of the den-sity ratio for the bulk of the gas, as the column densityratio is weighted by the amount of neutral gas. One −2−10120 Log n H −2−10120 Log n H −4−3−2−1010 Log n e −4−3−2−1010 Log n e −2 −1 0 1 2Log n H Log T −2 −1 0 1 2Log n H Log T −4 −3 −2 −1 0 10Log n e −4 −3 −2 −1 0 10Log n e Figure 3.
Example of a Monte Carlo Markov Chain run for DLAJ1313+1441. The upper left panel, middle panel and bottom rightpanel display the PDF for the individual parameters. The otherpanels show the parameter space that is covered by the MCMC.The light gray (dark gray) shaded regions are the 3- σ (1- σ ) bound-aries for the complete chain. The gray (orange) and black (red)region are the 3- σ (1- σ ) boundaries for just the solutions that sat-isfy the electron density constraint described in Section 4.3. scenario where this assumption might lead to inaccuratepredictions of the physical conditions of the gas is thecase where the column density of the upper level arisesfrom one phase whereas the bulk of the gas is in an-other phase. This is indeed expected to happen in thetwo-phase model where the CNM will produce the ma-jority of C II ∗ , even though C II could come from eitherphase. However, requiring that the two phases are inpressure equilibrium implies the carbon density ratio, r C ,between the CNM and WNM differ at most by a factorof 25. Therefore even in a 60 % WNM and 40 % CNMmixture, the density obtained using this technique willoverestimate the density for the bulk of the gas (i.e. theWNM component) by less than an order of magnitude.We explore this assumption further and compare the re-sults from individual velocity components to each otherand the system as a whole in Section 5.3.4.2. Applying the Technique
To explore the parameter space of all possible electrondensities, neutral hydrogen densities and temperature,we apply a Monte Carlo Markov Chain (MCMC) methodusing the Hastings-Metropolis algorithm. This allows usto sample the complete parameter space and find theprobability distribution function (PDF) for each of thephysical parameters. The likelihood function used in thealgorithm is discussed in Section 4.2.1. At each step inthe MCMC, this likelihood is then evaluated and multi-plied by the priors. The Hastings-Metropolis algorithmis finally used to accept the step or discard it. To test forconvergence, we run the MCMC five different times withvarying starting points. We run the chains for 10 steps,and discard the first 30 % of the chain as our burn-inperiod. After the run, the PDF of each of the parame-ters is compared for the five different chains to check forconvergence. Figure 3 shows an example of the resultsfor one such run. Neeleman et al. N1 σ N1 detection (a) N1 σ N1 detection (b) σ N1 upper limit (c) σ N1 upper limit (d) N1 lower limit (e) N2 σ N2 detection N2 lower limit N2 σ N2 detection N2 lower limit N2 σ N2 detection N1/N2 σ q (N1+2 σ N1 )/N2 σ N1 /N2 σ N1 /N2 N1/(N2+2 σ N2 ) Figure 4.
The likelihood functions for the different possible ratios. The solid red line is the likelihood function used to describe the ratio.In gray a sample PDF with the given parameters is shown, scaled to the likelihood function. The top panel in each of the subfigures depictsthe numerator, the middle panel the denominator and the bottom panel the ratio of the two. In (a) the likelihood can be approximatedby a Gaussian with mean = N N and σ q = r(cid:0) σ N N (cid:1) + (cid:16) N σ N N (cid:17) . In (c) the likelihood function of the ratio is also Gaussian. For theremaining cases (b), (c), and (e) the likelihood functions are approximated by step functions. These step functions contain > N ). Likelihood Function and Priors of the MCMC
The likelihood function used in the MCMC method isthe product of the likelihood functions of the two indi-vidual ratios: L = Si,C Y k L k (4)Here L k can take on different forms depending on if thecolumn densities measured in the ratios are detections,upper limits, lower limits or some combination of thetwo. In our sample we have 5 different cases which areschematically shown in Figure 4. Note that for all cases r k are linear quantities, not logarithmic.When both the numerator and denominator in the ra-tio are detections (Fig. 4a), we can approximate thePDFs of the individual measurements by Gaussians. Theresultant PDF of the ratio will then also be approxi-mately Gaussian, i.e. L k = e − χ / , where χ = (cid:18) r k , obs − r k , mod σ r k (cid:19) (5)Here r k are the quotient of the upper to lower level fine structure states of C + and Si + ; the subscripts refer toeither the observed or measured values and those fromthe model. σ r k is the uncertainty on the observed ra-tios calculated by standard error propagation. The priorPDF in this case is uniform.For upper limits on the column density measurement,we assume a Gaussian PDF centered around zero wherethe uncertainty is given by the 1- σ upper limit measure-ment. The likelihood of a ratio consisting of an upperlimit and a detection (Fig. 4c) will then also be givenby Equation 5. Of course negative ratios are unphysi-cal, and therefore we assume a prior which is zero forratios smaller than zero and uniform for ratios greaterthan zero.Lower limit measurements of the column density aremore difficult to deal with, as it is hard to estimate an ap-propriate uncertainty on the measurement, because theuncertainty is strongly dependent on the model used tofit the absorption line. If the lower limit measurement isin the denominator of the silicon or carbon ratio (Fig. 4band d), the resultant measured ratio is an upper limit.In this case, ratios derived from the model should have ahigh likelihood when they are smaller than the measuredhe Physical Conditions of High- z Atomic Gas 7 −2−10120
Log n H −4−3−2−10 Log n e Log T Log P / k B Log P / k B
50 60 70 80 Index Number
Figure 5. σ constraints on the neutral hydrogen density, electron density, temperature, and pressure for all of the DLAs in the samplefrom the Si II ∗ and C II ∗ technique. The left panels are the results for the DLAs which are part of the unbiased sample of Neeleman et al.(2013), whereas the right panel shows the result for the metal selected sample. The index number for the DLAs are given in Tables 4 and4. The DLAs marked in black (blue) are those DLAs for which this method provides a well determined constraint on both n H and T. ratio and small when they are bigger than the measuredratio. We therefore adopt the following conservative like-lihood function: L k = (cid:26) r k , mod ≤ r k , obs r k , mod > r k , obs (6)This step function rules out models that produce ratiosgreater than r k , obs , and will give equal likelihood for allratios below r k , obs . Here r k , obs is the 2- σ upper limit ofthe detection or upper limit divided by the lower limit.As Figure 4b and 4c show, this is a conservative ap-proach, as > L k = (cid:26) r k , mod < r k , obs r k , mod ≥ r k , obs (7) This step function rules out all models that produce ra-tios smaller than r k , obs . Here r k , obs is the lower limitdivided by the 2- σ upper limit of the detection. Notethat in these cases we assume a uniform prior.4.3. Electron Density Constraint
When we apply the MCMC chain, we allow the tem-perature, hydrogen density and electron density to varyindependently. This can clearly result in unphysical sit-uations for DLAs where we expect the fractional ioniza-tion to be significantly smaller than 1. To apply this con-straint, we assume that the fractional ionization of hydro-gen, x (H + ), satisfies the following steady state equation(Draine 2011): ζ CR+X (1 + φ s )[1 − x (H + )] = α rr (H + )n [x(H + ) + x(M + )]x(H + )+ α gr (H + )n x(H + ) (8) Neeleman et al. Table 2
Results of Si II ∗ and C II ∗ TechniqueIndex QSO z abs log n H (2- σ ) log n e (2- σ ) log T(2- σ ) log P/k B (2- σ )Number [cm − ] [cm − ] [K] [K cm − ]1 Q1157+014 1.9437 1 . . . . − . − . − . − .
0) 2 . . . .
1) 3 . . . . −
02 2.0395 1 . . . . − . − . − . − .
9) 1 . . . .
2) 3 . . . . − . . . . − . − . − . − .
1) 1 . . . .
2) 2 . . . . . . . . − . − . − . − .
9) 2 . . . .
2) 4 . . . . . . . . − . − . − . − .
9) 1 . . . .
3) 3 . . . . − . . . . − . − . − . − .
1) 1 . . . .
9) 2 . . . . . . . . − . − . − . − .
9) 2 . . . .
1) 4 . . . . . . . . − . − . − . − .
9) 2 . . . .
0) 3 . . . . . . . . − . − . − . − .
9) 2 . . . .
4) 3 . . . . . . . . − . − . − . − .
1) 1 . . . .
8) 2 . . . . . . . . − . − . − . − .
0) 2 . . . .
4) 3 . . . . . . . . − . − . − . − .
2) 1 . . . .
2) 2 . . . . . . . . − . − . − . − .
2) 1 . . . .
3) 2 . . . . − . . . . − . − . − . − .
6) 2 . . . .
4) 4 . . . . . . . . − . − . − . − .
2) 1 . . . .
7) 2 . . . . − . . . . − . − . − . − .
1) 1 . . . .
5) 2 . . . . − . . − . . − . − . − . − .
3) 1 . . . .
8) 1 . . . . Note . — (This table is available in its entirety at the end of this manuscript.)
Here, ζ CR+X is the primary ionization rate of both cosmicrays and strong X-rays, φ s are the secondary ionizationrates, α rr is the rate coefficient for radiative recombi-nation of H + which is a function of temperature, and α gr is the effective rate coefficient for grain-assisted re-combination of H + , which is a function of temperature,electron density and the UV radiation field. The ion-ization of metals is assumed to be the same as it is forlocal ISM scaled to the metallicity of the DLA. Usingthe estimates for these parameters in Draine (2011); andreferences therein, we can make an estimate for the elec-tron density as a function of the neutral hydrogen den-sity, temperature, primary ionization rate and the UVradiation field.This constraint is imposed upon the Monte CarloMarkov Chain after the run. All the values that donot satisfy the above equation are rejected. Since thevalue of ζ CR+X at these redshifts is uncertain (see e.g.Dutta et al. 2014), and the UV radiation field can likelytake on a wide range of values depending on the sepa-ration between the DLA and any potential star form-ing region, we allow for a wide range of acceptablevalues. Specifically, the UV radiation field may rangebetween 0.1 and 100 G (Habing’s constant; G =1 . × − ergs cm − s − ) and ζ CR+X between 10 − and10 − s − .As a result of this constraint, the electron densitiesnever exceed densities of about 0.1 cm − , because suchelectron densities would require the gas to be significantlyionized. Similarly, electron densities below ∼ − cm − are ruled out because of the intrinsic electron density dueto the singly-ionized metals such as carbon. One exampleof the application of this constraint to the MCMC chainare shown by the orange and red contours in Figure 3. RESULTSThis section describes the results from the Si II ∗ andC II ∗ technique. The output of the technique is a PDFon each of the three physical parameters (e.g. n H , n e andT). We have plotted the 1- σ ranges for each of these pa- rameters in Figure 5; we also have included the pressureconstraints of these systems in this figure (see Section5.1). The results are tabulated in Table 4. Not all DLAshave well-determined ranges on all internal parameters,either due to low S/N spectra or because the resultantratios are not strongly correlated with a specific inter-nal parameter (see Section 4.1). Those DLAs that havewell-determined ranges in both neutral hydrogen densityand temperature are plotted in blue in Figure 5 and areshown in the abbreviated version of Table 4.5.1. Physical Parameters of DLAs
The top panel of Figure 5 shows the distribution ofneutral hydrogen density in our sample. The range ofallowed neutral hydrogen column densities varies signifi-cantly between DLAs. Several DLAs such as J0927+1543are very dense, with 1- σ lower limits on the density of100 cm − . These high values are driven by high r C , andthe non-detection of Si II ∗ . On the other hand, severalother DLAs have 1- σ upper limits of 10 cm − indicat-ing that the gas in DLAs exhibits a wide range of neu-tral hydrogen density. It is important to note that theSi II ∗ and C II ∗ method cannot precisely measure neu-tral hydrogen densities below 1 cm − , because for thesedensities the interactions with neutral hydrogen becomessubdominant to collisions with electrons. As such, onlyDLAs with neutral hydrogen densities above this valuehave well-determined constraints on their neutral hydro-gen density.The second panel of Figure 5 shows the distributionof electron densities. One interesting feature of this dis-tribution is that the variation in the range of electrondensities is significantly less compared to the range ofneutral hydrogen densities. This is a direct consequenceof applying Equation 8, which makes the electron densityonly weakly dependent on the temperature and densityof the gas. Specifically, Equation 8 gives an electron den-sity of n e = 0.01 cm − for both a canonical CNM (n H =30 cm − and T = 50 K) and WNM (n H = 0.5 cm − andT = 5000 K) for local ISM conditions. Note that thishe Physical Conditions of High- z Atomic Gas 9is slightly higher than the median value of the completeDLA sample (n e = 0 . ± . − ). The differenceis likely due to the lower metallicity of the DLA sam-ple and different UV radiation fields and ionization ratesfrom cosmic rays and X-rays (Wolfire et al. 1995). As aresult, we are unable to differentiate between the electrondensities of DLAs, but a typical DLA will have an elec-tron density of about 4 × − cm − , which is consistentwith the values found by Srianand et al. (2005).The third panel of Figure 5 shows the temperaturerange for each of the DLAs in our sample. Several DLAshave temperature ranges that are consistent with thetemperatures expected from a CNM. To be specific, nineof the DLAs have 1- σ upper limits on the temperatureof 500 K. On the other hand, there are several DLAsthat have ranges that are at significantly higher temper-ature. We again would like to stress that the Si II ∗ andC II ∗ method is unable to measure the precise tempera-ture above 500 K as r C and r Si become weak functionsof temperature. Therefore only the coldest DLAs havewell determined ranges on their temperature.Finally, in the bottom panel of Figure 5 we have plottedthe pressure range for each of the DLAs. The pressurewas calculated from the MCMC chains by taking theproduct of the neutral hydrogen density and the tem-perature, since P/k B = n H T . As pressure shows thestrongest correlation with r Si and r C , it is the best con-strained parameter. The median pressure for the com-plete sample is log P/k B = 3 . − ]. This is lowercompared to the pressure of the ISM measured locally us-ing the C I method (Jenkins & Tripp 2011). We discussthis further in Section 6.2.In Figure 6 we have plotted the temperature versusthe density for the complete sample of DLAs. We havealso indicated the typical ranges for a canonical CNM,WNM and the classically forbidden region defined bythe two-phase model (see e.g. Heiles & Haverkorn 2012).Nine DLAs with well-determined ranges on the temper-ature and neutral hydrogen density have physical condi-tions that are consistent with those expected from gasin a CNM. Two of these DLAs are from the 48 DLAswhich are part of the unbiased sample of Neeleman et al.(2013). We therefore conclude that at least 5% of arandom sample of DLAs contain significant fractions ofCNM. This percentage is a lower limit because manyDLAs with less well-determined ranges are consistentwith gas in a CNM as is shown by the gray and darkgray contours.The remaining DLAs are spread over a wide range oftemperature and neutral hydrogen densities, all of themconsistent with both canonical CNM and WNM condi-tions. In Figure 6 the 68 % and 99 % confidence con-tours for the temperature and neutral hydrogen densityof the remaining DLAs are shown. These contours ruleout the parameter space of high neutral hydrogen densityand high temperature, as such physical conditions wouldproduce too large r Si and r C (see Figure 2).5.2. Correlations with Global DLA Properties
In this section we explore possible correlations betweenthe physical parameters measured with the Si II ∗ andC II ∗ method and the global properties of the DLAs.We consider all of the global DLA parameters discussedin Neeleman et al. (2013). Figure 7 shows a selection of −1 0 1 2 3log n H [cm −3 ]1234 l og T [ K ] −1 0 1 2 3log n H [cm −3 ]1234 l og T [ K ] CNMForbidden RegionWNM
Figure 6.
Temperature and neutral hydrogen density for the com-plete sample. The data points are those DLAs with well-definedranges on n H and T (see Figure 5. The dark gray (light gray) linesindicate the 68 % (99 %) confidence interval of the neutral hy-drogen density and temperature for the remaining DLAs. We havealso plotted the typical ranges for a canonical cold neutral medium,warm neutral medium, and the classically forbidden region. these correlations. As panel (a) of this figure shows, thereis a clear correlation between the neutral hydrogen den-sity and the 158 µ m cooling rate of the DLAs ( ℓ c ; see e.g.Wolfe et al. 2003a). This result is due in part becausethe cooling rate is proportional to the C II ∗ column den-sity and larger C II ∗ column densities result in larger r C ,which in turn yield higher neutral hydrogen column den-sities (see Figure 2). We consider this result evidence forthe two-phase model described by Wolfe et al. (2003a),where higher star formation rates, and therefore highercooling rates, result in higher stable equilibrium densitiesfor hydrogen. Since ℓ c is correlated to metallicity, red-shift and the kinematical parameters (Wolfe et al. 2008;Neeleman et al. 2013), the neutral hydrogen density alsoshows a correlation (albeit weaker) with these parame-ters.Panel (b) is a plot of neutral hydrogen density vs H I column density. The ratio of these components gives acrude estimate of the absorption length of the DLA. Tobe specific, the Si II ∗ and C II ∗ technique provides an es-timate of the density of the bulk of the neutral gas (Sec-tion 4.1). Therefore the ratio of the neutral hydrogendensity and H I column density will give an upper limitto the size of the component which contains the bulk ofthe neutral gas. The results show that for the majorityof DLAs the absorption lengths for these components isless than 1 kpc. Indeed some absorption lengths are assmall as a few tens of parsec. These very small absorp-tion lengths correspond in general to those DLAs withwell determined cold temperatures (T ≤
500 K). Thissuggests that for these DLAs the bulk of the gas is lo-cated in relatively small cold components.Unlike the hydrogen density, the electron density showsno significant correlation with any of the global DLAparameters. Similarly, the temperature shows no sig-nificant correlation either. However, we would like todiscuss two interesting features of the temperature mea-0 Neeleman et al. −2−10123 −2−10123
Log n H (a) (b) p c p c p c p c −28 −27 −26 Log c
1 2 3 4
Log T (c) 20 21 22Log N(HI) (d) Figure 7.
Selection of possible correlations seen in the data. The black (green) data points are those points with strong constraints ontheir physical parameter from the MCMC analysis. Panel (a) shows the correlation between cooling rate and the hydrogen density. Thiscorrelation is expected as larger cooling rates indicate larger C II ∗ ratios which result in higher neutral hydrogen densities. Panel (b) plotshydrogen density vs neutral hydrogen column density. This plot indicates a maximum cloud size of DLAs of less than 1 kpc. The last twopanels show that temperature is not strongly correlated with any of the external parameters. surements. The first feature is that the DLAs with thehighest cooling rate have on average a lower tempera-ture range (Figure 7c). This likely is related to thecorrelation between cooling rate and neutral hydrogendensity, as higher neutral hydrogen densities likely cor-respond to colder environments (Figure 6). The secondfeature is that the two highest column density systemshave the highest limits on the temperature (Figure 7d).A possible explanation for this result is that low tem-perature gas will form molecular gas, limiting the max-imum allowed column density of neutral atomic hydro-gen (Schaye 2001). The molecular fraction is believedto be anti-correlated with temperature (Schaye 2001;Richings et al. 2014b) and therefore larger atomic neu-tral hydrogen column densities are possible for highertemperature DLAs.5.3. Component Analysis
In all of the analysis we have assumed that the ratio ofcolumn densities is approximately equal to the ratio ofthe densities of the bulk of the neutral gas. To explorethis assumption, we have repeated the analysis on each ofthe individual velocity components of those DLAs withmeasurable Si II ∗ (Figure 1). The results are displayed inFigure 8. As can be seen from the individual panels, themajority of the velocity components have physical pa-rameters that are within 1- σ equal to the measurementsfrom treating the system as a whole. This result is dueto the similarity in the column density ratio between theupper and lower levels of the fine structure states of Si II .Note that in Figure 1 the Si II ∗ line traces the low-ionline quite well for the majority of DLAs. There are sev-eral exceptions such as the component at +50 km s − for DLA J1313+1441. This component has an r Si fivetimes greater than the mean value of the DLA, resultingin a temperature range inconsistent with that found forhe Physical Conditions of High- z Atomic Gas 11the total DLA. However, such components are uncom-mon; the mean deviation in r Si and r C from componentto component is less than 50 % of the mean value, whichresults in similar ranges for the physical parameters.The similarity between r Si and r C for the different ve-locity components strengthens the assumption to takethe ratio of the total column densities to be equal to theratio of the densities, since a per component analysis willproduce similar results. One possible explanation for thesimilarity between the individual velocity components isthat the distinct components are physically close to eachother and experience similar exterior physical conditions,or a second explanation could be that external conditionsare similar over a large portion of the absorbing galaxy.There are three caveats to this results. The first caveatis that this result does not exclude the existence of any clumps of gas with strongly varying physical parametersalong the quasar line of sight. It does, however, suggestthat these clumps can only contribute a very small frac-tion of the total metal column density, and therefore arenot likely to describe the bulk of the neutral gas. Thesecond caveat is that the DLAs used in the individualcomponent analysis all have measurable levels of Si II ∗ and therefore might not be representative of the DLAsin general. We have tested this caveat by considering theratio of C II ∗ to Si II in a sample of DLAs for which both −2 0 2 Log n H J1056+1208−50 0 50 234
Log T J1313+1441−50 0 50 −2 0 2
Log n H J1310+5424−50 0 50 234
Log T J1142+0701−50 0 50 −2 0 2
Log n H Q1157+014−50 0 50 234
Log T J1417+41320 100 −2 0 2
Log n H Q1755+5780 100 200 300 234
Log T Velocity (km s −1 ) Figure 8.
Temperature and neutral hydrogen density for the indi-vidual components for those DLAs with detectable levels of Si II ∗ .The gray (orange) values are the measurements for the hydrogendensity and temperature for the individual components, and areplaced at the velocity center of the component. The black (blue)data point is the result from the MCMC analysis by consideringthe system as a whole. transitions are detected, and we find that this ratio is alsonot strongly varying between the individual components(see also Wolfe et al. 2003a,b). Therefore we believe thatthis is a general result holding for the majority of DLAs.Finally, the third caveat is that it could be that the in-dividual velocity components are in actuality composedof a collection of smaller components. In this case, eachindividual component is averaged in a similar manner asthe whole DLA, and therefore weighted most strongly bythe component with the largest metal column density. Asa result the Si II ∗ and C II ∗ technique will still recoverthe physical conditions of the bulk of the gas. DISCUSSIONThe discussion section is organized as follows. In Sec-tion 6.1 we discuss the empirical results. We comparethese results to previously measured data for both DLAsand the local ISM in Section 6.2. In Section 6.3 we dis-cuss how the results of this paper fit into models describ-ing the ISM, in particular the two-phase model. Finally,in Section 6.4 we will comment on what these resultssuggest for the ISM of high redshift galaxies.6.1.
Discussion of the Empirical Results
We found in Section 5 that the temperature distribu-tion of the full sample cannot precisely be determinedbecause the Si + and C + ratios are insensitive to tem-perature changes when T exceeds 500 K. However, wecan measure the minimum fraction of DLAs that havegas temperatures consistent with a CNM (i.e. T < r C . Thiscorrelation is shown in Figure 9. The tracks are theo-retical temperature paths for the indicated neutral hy-drogen density assuming the measured median electrondensity of 0.0044 cm − . The black data points are de-tections or lower limits to r C . This figure illustrates twothings. Firstly, the majority of detections and lower lim-its in r C cannot be produced in neutral gas with n H . − , which means that a canonical WNM is rarelyable to produce detectable levels of C II ∗ , and because r C values can be measured in approximately 40 % ofDLAs (see Neeleman et al. 2013) this indicates that theseDLAs must contain some fraction of gas not in a canon-ical WNM. Secondly, the DLAs with the highest C + ra-2 Neeleman et al.tios have pressures and neutral hydrogen densities signif-icantly higher than the median, and are in general thoseDLAs with well-defined limits on temperature and den-sity.Indeed, if we consider just the systems with well-defined ranges on temperature and density, we find thatthese systems show significantly higher velocity widthswith a median velocity with of 131 km s − which is al-most double the median value of a random DLA sam-ple (Neeleman et al. 2013). They also show an increasedmetallicity and cooling rate, all suggesting that these sys-tems are part of the most massive dark matter haloswhich give rise to DLAs (see further Section 6.4.2).Finally a detailed look at the individual velocity com-ponents of DLAs shows that there exists little differencesbetween the measured ratios between the individual ve-locity components. As was suggested in Section 5.3 thiscould be due to close proximity of the individual com-ponents. In particular, the observed r C =C II ∗ /C II isproportional to: r C ∝ ℓ c [M / H] (9)Hence, if we assume that the metallicity of the indi-vidual components does not vary significantly, then theconstancy in r C indicates that the cooling rates are thesame across the individual components. This is in-deed expected and assumed in the two phase model ofWolfe et al. (2003a), as they found that heating rate (andbecause of the assumed thermal equilibrium thereforealso cooling rate) is a global property of the DLA andnot a local property as it is in the local ISM. The con-stancy of r C across the different components is thereforeexpected in the two-phase model, as it is a direct conse-quence of the global nature of the heating rate. −3.5 −3.0 −2.5 −2.0 −1.5 −1.0r C Log P / k B ( K c m − ) −3.5 −3.0 −2.5 −2.0 −1.5 −1.0r C Log P / k B ( K c m − ) log T [K] −3 n H = 1 cm −3
10 cm −3
100 cm −3 Figure 9.
Correlation between pressure and r C . The black datapoints are lower limits and measurements of r C , whereas the graydata points are 2- σ lower limits. The grayscale (colored) tracks arethe theoretical solutions to r C and pressure for the given neutralhydrogen density and temperature assuming an electron density ofn e = 0.0044 cm − . Most r C detections, which account for about 40% of a random DLA sample, are inconsistent with the conditionsfound in a canonical WNM. Table 3
Comparison between Temperature MeasurementsQSO T (K) (1 − σ constraint)C II ∗ /Si II ∗ Other MethodQ0336 −
01 (28 - 5200) > −
02 (50 - 316) (465 - 655) 21 cm absorptionQ1157+014 (260 - 5500) (760 - 1270) 21 cm absorptionJ0812+3208 (25 - 178) (32 - 88) C I J2100 − I Q2206 − >
25 (9200 - 15200) line-fittingJ2340 − I Comparison with Previous Observations
In Table 4 we have listed the results of our study on theneutral hydrogen density, electron density, temperature,and pressure of DLAs using the Si II ∗ and C II ∗ method.This is not the first study of these parameters as sev-eral other methods provide estimates. A comparison forthose 7 DLAs which have measured temperatures frommultiple methods is shown in Table 3. These DLAs sug-gest that there is a reasonable agreement between thephysical parameters derived using the Si II ∗ and C II ∗ method and previous methods.6.2.1.
21 cm Absorption
As was discussed in the introduction, one method ofmeasuring the spin temperature of DLAs is by measuring21 cm absorption in DLAs located in front of radio-loudquasars. A comprehensive paper describing this methodwas recently published by Kanekar et al. (2014). Theyfound that the median temperature of the gas insideDLAs responsible for 21 cm absorption in their samplewas greater than 900 K. Furthermore, they found thatonly 2 out of the 23 DLAs above a redshift of 1.7 wereconsistent with having a significant fraction of CNM. Wenote that we found in our sample that this fraction mustbe at least 5 %. These results are consistent within theuncertainty of the measurements due to the small samplesizes of both methods. Because the former measurementis an upper limit and the latter a lower limit, the twomethods suggest that roughly between 5 and 10 % of allDLAs have the bulk of their gas in a CNM phase.This fraction is somewhat in conflict with the resultsfrom Wolfe et al. (2003b, 2004), who argued that the ma-jority of all C II ∗ detections in DLAs must come from gasin a CNM, and about 40 % of all DLAs have detectablelevels of C II ∗ (e.g. Neeleman et al. 2013). Kanekar et al.(2014) resolves this conflict by assuming that only a smallfraction of the gas (10 - 20 %) in the DLAs with C II ∗ detections is in actuality CNM, with the bulk of the gasin a WNM phase. There are two problems with thisscenario. First, it is unclear why in this scenario, theC II ∗ /Si II ratio would be relatively constant among theindividual velocity components, as it is in observations.Secondly, using the technique described in this paper, wecan calculate the amount of C II needed in the CNM toproduce the required amount of C II ∗ observed. In thetwo cases mentioned in Kanekar et al. (2014) (i.e. DLAsQ1157+014 and Q0458 − II columndensity needed in a canonical (T = 100 K) CNM to pro-duce the observed C II ∗ column density is larger than theobserved total C II column density. Hence, at least forhe Physical Conditions of High- z Atomic Gas 13these two DLAs, we can rule out a scenario where only10 - 20 % of the gas is in a CNM.A more plausible explanation for the conflicting resultsis that we cannot assume that a simple two-phase modelconsisting of a canonical CNM of T = 100 K and WNM ofT = 8000 K is capable of reproducing the results for thelarge range of physical conditions applicable for all DLAs.Indeed considering the wide variety of ranges in metal-licity, dust-to-gas ratios, and UV radiation fields, theresults from both Wolfire et al. (1995) and Wolfe et al.(2003a) suggest that CNM temperatures can range from10 K to 500 K, with higher temperatures more likely forlower metallicities, higher dust-to-gas ratios and higherUV radiation fields. From the Si II ∗ and C II ∗ tech-nique we can conclude that DLA Q0458 −
02 likely con-tains the bulk of the gas at a temperature of ∼
300 K,still well within the range of a CNM phase as defined byWolfe et al. (2003a), and fully consistent with the resultfound in Kanekar et al. (2014). DLA Q1157+014 is anexception as the results from this paper suggest it has thebulk of its gas at a temperature of ∼ II ∗ (See Section 6.2.4). Furthermore, we wouldexpect to find some DLAs with gas temperatures incon-sistent with either the CNM or WNM as such gas is seenoften in the local ISM (Roy et al. 2013b).6.2.2. C I Fine-Structure Study
A second method used to measure the physical param-eters of DLAs is by considering the fine structure lines ofC I . This was done for several DLAs by Srianand et al.(2005) and Jorgenson et al. (2010). Jorgenson et al.(2010) found that the densities and temperatures derived −1 0 1 2 3log n H [cm −3 ]1234 l og T [ K ] −1 0 1 2 3log n H [cm −3 ]1234 l og T [ K ] K c m − K c m − K c m − K c m − This WorkJorgenson+10Cooke+14
Figure 10.
Allowed temperature and neutral hydrogen densityparameter space for DLAs. The larger (outlined in red) squaredata points are those with detectable levels of Si II ∗ . Overplottedon this figure are lines of constant pressure. The contours markthe 68 % (dark gray) and 99 % (light gray) confidence levels of theunbiased sample of Neeleman et al. (2013). The dotted (red) lineis the pressure of the local ISM as measured by Jenkins & Tripp(2011). from this method could only result from very dense (n H ≥
30 cm − ) and cold gas (T ≤
150 K). They therefore sur-mise that C I traces very dense pockets of very cold gas atslightly higher pressures. This is indeed seen in Figure 10where the results from Jorgenson et al. (2010) trace thecoldest and densest measurements from our sample. Fur-thermore, the mean pressure from the Jorgenson et al.(2010) sample is higher than the median pressure forthe complete sample in this paper (log( P/k B ) = 3.0 [Kcm − ]).A comparison between the three DLAs that have beenanalyzed using both the C I and the C II ∗ and Si II ∗ anal-ysis (Table 3) shows that both methods give remarkablysimilar temperatures and densities. This is somewhatat odds with the scenario put forth in Jorgenson et al.(2010). They suggest that C I traces small dense clumpsof cold neutral gas in a larger less dense medium of coldgas. However, we find that for these three DLAs the C I method gives values in agreement with the measurementsof the bulk of the gas from the Si II ∗ and C II ∗ technique,removing the need for this scenario in these DLAs.One possible explanation for this result is that the C I analysis can only be performed when multiple C I fine-structure states can be measured. Such measurementsare easiest for those DLAs with large column densities ofthe C I fine-structure states, which results in preferen-tially selecting DLAs which contain the bulk of their gasin a cold and dense phase. This assessment is corrobo-rated by the fact that 5 out of the 9 DLAs with 1- σ tem-perature measurements below 500 K show C I absorp-tion. For the unbiased DLA sample of Neeleman et al.(2013), the fraction of DLAs showing detectable levels ofC I is more than 10 times smaller; only 4 out of the 80DLAs show C I absorption. A positive detection of C I is therefore a strong indicator that the DLA contains asignificant fraction of cold, dense gas.6.2.3. Other Studies
The third and final method discussed here for measur-ing the temperature of DLAs is the use of fitting routinesto measure the Doppler parameter of individual compo-nents. By measuring a wide range of ion species, one isable to untangle the thermal broadening of the Dopplerparameter from the turbulent or bulk motion of the gas.The thermal broadening gives an estimate of the temper-ature of the gas. This method has been used to find thetemperature of individual components in DLAs, resultingin detection of both cold and warm gas (Carswell et al.2010, 2012). However, the multiple velocity componentsof a typical DLA, make this method daunting.Recently this method has been used for a selection ofvery metal poor DLAs, which have simpler velocity struc-ture (Cooke et al. 2014; Dutta et al. 2014). The resultfrom these studies indicate that these DLAs have highertemperature and lower densities than the DLAs in oursample with conditions similar to those expected from aWNM. This suggests that the gas being traced by thesevery metal-poor DLAs is less likely to host star forma-tion, which is corroborated by the lower metallicity ofthe gas.Finally, we can compare our results to those found forthe local ISM. Using C I , Jenkins & Tripp (2011) findthat the CNM in the local ISM has an average pressure4 Neeleman et al.of log( P/k B ) = 3.58 ± − ]. The median pres-sure for our sample is log( P/k B ) = 3.0 [K cm − ]. How-ever, if we include only those DLAs with well-determinedranges on their pressure, the median pressure becomeslog( P/k B ) = 3.4 [K cm − ]. We believe that this pres-sure is more representative of the complete DLA sample,as the large number of lower limits will artificially lowerthe median pressure. This pressure is very similar to thepressure found locally, although our sample has a largerrange of allowed pressures. This extended range in pres-sures is easily explained by the fact that DLAs probe avariety of different galaxies, with a wider range of physi-cal conditions compared to those seen in our own Galaxy.6.2.4. Summary
Table 3 lists the results for the 7 DLAs which werepreviously examined using either 21 cm absorption, C I absorption or line profile fitting. These 7 DLAs showthat for the limited sample of DLAs with temperaturemeasurements from two different methods, the Si II ∗ andC II ∗ method measures temperatures that are in generalagreement with the results from other techniques. Twodiscrepancies exist. The temperature measurement forQ0458 −
02 from the 21 cm absorption study is likely highbecause the optical and radio line of sight encounter dif-ferent column densities of gas (see Kanekar et al. 2014).The only other measurement that is inconsistent within1- σ is that of DLA Q0336 −
01; this discrepancy is alsodiscussed in Kanekar et al. (2014). The remarkableagreement between the methods suggest that at least forthe subset of DLAs with large fractions of cold gas, theSi II ∗ and C II ∗ method is able to accurately determinethe temperature and density of the gas.The results from this section are summarized in Fig-ure (Figure 10). The gray contours are the 68 % and99 % confidence intervals of the unbiased sample ofNeeleman et al. (2013). The data points for our sam-ple are those DLAs with well-defined ranges. Of theseDLAs, the ones marked with larger squares (outlined inred) are those with measurable Si II ∗ . The DLAs withdetectable levels of Si II ∗ fall outside the 68 % contour in-tervals, indicating that the conditions conducive to Si II ∗ detections are not common in a random sample of DLAs.Indeed one DLA, J1135 − I method byJorgenson et al. (2010) are consistent with measurementfor the coldest and densest DLAs in our sample. Thisis not unexpected as C I likely traces the coldest gasin DLAs. On the other hand the metal poor sampleof Cooke et al. (2014) have temperatures and densitiesconsistent with a WNM. These DLAs fall outside the 68% contour of the unbiased sample, suggesting that thelow densities for this sample are not common in a typ-ical DLA and could be due to the very low metallicityof these DLAs. Finally, the dotted line in Figure 10 isthe average pressure of the local ISM (Jenkins & Tripp2011), which is consistent with the pressures found inDLAs.6.3. Comparison with the Two-Phase Model
As discussed in the introduction, Wolfe et al. (2003a)adopted the two-phase medium model from Wolfire et al.(1995) to describe the physical conditions of the gasaround DLA galaxies. The results from this paper areable to test the validity of the two-phase model, sincethe Si II ∗ and C II ∗ method provides independent mea-surements of the temperature and density of the DLAgas.The first such test is to check that the two-phase modelis able to reproduce the range of allowed pressures. Wefind allowable pressures ranges of log( P/k B ) between 1[K cm − ] and 6 [K cm − ]. This large range of pressuresis allowed within the two-phase model (see Fig 5a and5c of Wolfe et al. 2003a), since the lower metallicity ofDLAs and varying star formation rate density can giverise to a large range of pressures that are able to maintaina stable two-phase structure.A second test of the two-phase model is provided bycomparing the star formation per unit area (Σ SFR ) pre-dicted from the two-phase model with that measuredfrom emission lines of the DLA galaxy. Detecting DLAgalaxies in emission is rare (see e.g. Krogager et al. 2012);hence only 1 DLA (J1135 − SFR from emission studies. Weconvert our pressure estimate and density measurementof this DLA into a star formation rate per unit area,in a similar way as was done in Figure 5 of Wolfe et al.(2003a). Using this method we find a star formation rateof 0.3 M ⊙ yr − kpc − . This compares well with the ob-served rate predicted from emission lines, which is ∼ ⊙ yr − kpc − (Noterdaeme et al. 2012).Finally, we can compare the temperatures from theSi II ∗ and C II ∗ method, to see if we find any evidencefor two distinct phases, which is a prediction of the two-phase model. As discussed in Section 6.2, we find thatat least 5 % of DLAs have a significant fraction of coldgas, consistent with a canonical CNM (Figure 6). Unfor-tunately, we are not able to confirm the existence of gasin a WNM as our method provides weak constraints athigh temperatures. However, from other studies such asCarswell et al. (2012) and Cooke et al. (2014), we knowthat such gas exists.In conclusion, the results in this paper are in gen-eral agreement with the two-phase model of Wolfe et al.(2003a,b). There are several DLAs, however, which havehigher than predicted temperatures and densities; theseDLAs are discussed further in Section 6.4.6.4. Implications for High- z Galaxies
In this section we will speculate about the implicationsthese results have on the physical conditions of DLA gasand the implication on the formation of high- z galaxies.6.4.1. Implications for DLA Gas
In this paper we have focussed on the physical condi-tions of the bulk of the neutral gas for a large sample ofDLAs. This is unlike previous absorption studies usingC I , which focus solely on the coldest and densest gasof DLAs as was noted by Jorgenson et al. (2010). Theresults from this paper corroborates this assessment asthe results from the C I analysis are consistent with thecoldest and densest gas measurements from the Si II ∗ and C II ∗ method.he Physical Conditions of High- z Atomic Gas 15 −3.5 −3.0 −2.5 −2.0 −1.5 −1.0r C −5−4−3−2−1 r S i −3.5 −3.0 −2.5 −2.0 −1.5 −1.0r C −5−4−3−2−1 r S i n H < 0.1 cm −3 n H > 10 cm −3 Figure 11. r C and r Si ratio for the complete sample. The filledcircle are the 46 DLAs that are part of the unbiased sample ofNeeleman et al. (2013) and have a well-defined r C and r Si . Theregion to the left of the dashed (red) line can only be occupied byDLAs with neutral hydrogen density less than 0.1 cm − , whereasthe region to the left of the solid (blue) line can be occupied solelyby DLAs with neutral hydrogen densities greater than 10 cm − . We find that the fraction of DLAs which have theirbulk of gas in such a cold and dense phase must be atleast 5 %. For the remaining DLAs the amount of gasin such a phase is unknown as the Si + and C + ratiosdo not provide stringent constraints on the temperatureand density of the gas. Interestingly, we find that theupper levels of both the ground state of Si + and C + have the same velocity structure as the lower levels forthe majority of DLAs, suggesting similar conditions forthe majority of the velocity components in these DLAs.The range of allowed temperatures and densities issummarized in Figure 10. The unbiased sample coversthe parameter space between the metal-poor sample andthe sample of C I detections. Several DLAs have tem-perature and density measurements which are inconsis-tent with the two-phase model of Wolfe et al. (2003a).The enhanced densities and temperatures in these DLAsincreases r C (see Figure 2), and therefore the observedcooling rate, ℓ c . As a result, setting the observed coolingrate equal to the calculated cooling rate from a two-phasemodel will over predict the star formation rate for theseDLAs.This could partially provide the answer to the un-successful attempts of observing the DLAs in emission(Fumagalli et al. 2014). If a large fraction of DLAs haveenhanced densities and temperatures compared to thetwo-phase model, the average star formation rate forDLAs will be systematically overestimated. One possi-ble explanation for the enhanced temperatures and pres-sures could be turbulence. At least in numerical simu-lations turbulence is able to drive some of the gas intothe classically forbidden region (Gazol et al. 2005). Suchgas will have enhanced r C ratios compared to those pre-dicted from the two-phase model and therefore the twophase model will overestimate the star formation rate.In conclusion, we suggest that the gas in DLAs followsin general the two-phase model of Wolfe et al. (2003a), as several studies have measured gas with propertiesvery similar to both the WNM (Lehner et al. 2008;Carswell et al. 2012; Kanekar et al. 2014; Cooke et al.2014) and CNM (Howk et al. 2005; Srianand et al. 2005;Carswell et al. 2010; Jorgenson et al. 2010). However,the detection of Si II ∗ in DLAs suggest that a fractionof DLAs have significantly higher densities and tempera-tures than expected from the two-phase model. For thesesystems the star formation rate obtained from the two-phase model is overestimated. Unfortunately, the exactfraction of these DLAs cannot be estimated from thisstudy. However, in the local ISM this fraction is quitesignificant ( ∼
30 %) (Roy et al. 2013b).We would like to note that these higher temperaturesare in agreement with the results from 21 cm absorption,thereby resolving two problems plaguing DLA studies atonce; the lack of detections of DLAs in emission andthe discrepancy between temperatures expected from thetwo-phase model and measurements from 21 cm studies.One possible adaptation to the two-phase model whichcould provide such a solution is to include turbulence, asnumerical simulations show that turbulence could drivegas into the classically forbidden region.6.4.2.
Implications for High- z Galaxy Formation
By the nature of their selection, DLA sightlines rep-resent a cross-section weighted sampling of high-surfacedensity, neutral hydrogen gas at high- z . In aggregate,these systems also represent a major reservoir to fuelgalaxy formation during the first few Gyr of the uni-verse (Wolfe et al. 1995; Prochaska et al. 2005). There-fore, one generally associates this gas with the ISM ofyoung galaxies. As such, the results presented here of-fer new insight into the nature of this ISM gas and itsrelationship to ongoing or future star-formation.Restricting the discussion first to our random sample,we find that the incidence of very strong fine-structureabsorption is rare: r C exceeds 10 − in only 4 out of the46 DLAs from the random sample with measured r C (seeFigure 11). Such high r C values require gas densities n H &
10 cm − for the majority of neutral, atomic gas athigh- z . In conjunction with the paucity of systems show-ing molecular gas and/or C I detection, which are bothindicators of dense gas, these results suggest that a largefraction of DLA gas is unconducive to star-formation.A result corroborated by the difficulty to directly mea-sure the in-situ star-formation of a typical DLA (see e.g.Fumagalli et al. 2014). Indeed, this material may evenform so-called ‘dark galaxies’ (Cantalupo et al. 2012).A significant fraction of DLAs (9 out 46) have mea-sured r C value in the range 10 − . to 10 − . For thesingle-phase analysis performed in this manuscript, wederive n H & − which exceeds the canonical valuefor the WNM. Together with the 18 DLAs which have up-per limits to r C that exceed 10 − . , about half of all sys-tems are consistent with moderate densities which exceedthose expected in a canonical WNM. One might argue,however, that the systems with intermediate r C valuesrepresent a mixture of dense and more diffuse gas withthe dense gas contributing nearly all of the observed C II ∗ absorption. Then, the bulk of the gas could (in principle)be very diffuse. However, as mentioned in Section 4.1,we disfavor extreme scenarios of this kind because the6 Neeleman et al.absorption profiles of C II ∗ λ H value derived from a single-phaseanalysis only overestimates the mass-weighted value bya factor of a few.Despite a large fraction of DLAs favoring modest den-sities, the majority of the DLAs have significant gaspressures ( P > K cm − ), which is a characteris-tic of an active ISM. Recent models of galaxy forma-tion within hierarchical cosmology predict highly tur-bulent conditions driven by the accretion of cool gasand violent disk instabilities within the galaxies (e.g.Kereˇs et al. 2005; Dekel & Birnboim 2006; Burkert et al.2010; Fumagalli et al. 2011). Perhaps such processes ex-plain the small, but non-negligible subset of DLA sight-lines with P > K cm − . As noted in Section 6.1,these pressures are predominantly recorded in gas withhigh metallicity and large velocity widths. SUMMARYIn this paper we have presented a new method of de-termining the physical conditions of gas in high redshiftgalaxies. Using the fine-structure lines of Si + and C + ,we are able to provide constraints on the temperatureand neutral hydrogen density of DLAs. We have appliedthis method to a sample of 80 DLAs, for which we areable to provide limits or detections of these fine-structureline transitions. This sample contains 5 new detectionsof the excited fine-structure line of Si + , which more thandoubles the previously know detections. The results ofthis analysis are:1. We find that 9 DLAs have temperatures consistentwith gas in a cold neutral medium. The remainingDLAs provide less stringent constraints on theirtemperature for two reasons. Firstly, the ratiosof fine-structure lines become insensitive to tem-perature changes above 500 K. Secondly, the lowdensity of Si II ∗ and C II ∗ in these systems makesdetection difficult; resulting in weak upper limitsto the column density measurements of both fine-structure lines in these systems.2. From the ‘unbiased’ subsample of DLAs part ofthe sample described in Neeleman et al. (2013), wefind that at least 5 % of all DLAs have significantfractions of gas with properties similar to a canon-ical CNM along their line-of-sight. This result isconsistent with the locally measured volume fillingfraction of 0.01 for the CNM.3. The results of the method show that the neutralhydrogen density of DLAs vary significantly fromDLA to DLA. On the other hand the electron den-sity varies little between DLAs with a median elec-tron density of 0 . ± . − . Furthermore,we can rule out the parameter space of high tem-perature and high neutral hydrogen density (seeFigure 6) as such gas would produce upper to lowerlevel fine-structure state ratios in excess of what weobserve. 4. We find that there exist a correlation between theneutral hydrogen density and the cooling rate of theDLA. This is consistent with the predictions fromthe two phase model, where stronger star forma-tion rates and therefore larger cooling rates resultin higher stable neutral hydrogen equilibrium den-sities. Furthermore, the comparison between theneutral hydrogen density and the total H I columndensity gives a rough estimate of the total absorp-tion length along the line-of-sight. These valuesrange from about 1 kpc to only a few pc, suggestingthat the bulk of the neutral gas at high redshift canbe located in reasonably small dense components.5. Finally, we find that the typical pressure of theDLAs in the sample is log( P/k B ) = 3.4 [K cm − ],which is comparable to the pressure of the localISM. However, the DLAs show a larger range inpressures, which can be easily explained by the factthat DLAs measure a range of different galaxies,with a wide range of different physical conditions.We speculate that these results indicate that DLAsgenerally follow the two-phase model of Wolfe et al.(2003b). However, a fraction of DLAs have tempera-tures and densities inconsistent with this model. As aresult, the two-phase model will over predict the starformation rate of these systems. By including a mech-anism in the two-phase model which will increase thetemperature and density of the gas for these DLAs, wecan account for both the higher spin temperatures seenin 21 cm absorption (Kanekar et al. 2014), and lower thestar formation rates of DLAs as is suggested by recent ob-servations (Fumagalli et al. 2014). One such mechanismis turbulence, which is able to drive gas into the unsta-ble temperature regime (Gazol et al. 2005; Walch et al.2011). ACKNOWLEDGEMENTSWe wish to thank the referee for the helpful commentsand N. Kanekar for reading a previous version of themanuscript. Support for this work was provided by NSFaward AST-1109452. The data presented herein were ob-tained at the W.M. Keck Observatory, which is operatedas a scientific partnership among the California Insti-tute of Technology, the University of California and theNational Aeronautics and Space Administration. TheObservatory was made possible by the generous finan-cial support of the W.M. Keck Foundation. The authorswish to recognize and acknowledge the very significantcultural role and reverence that the summit of MaunaKea has always had within the indigenous Hawaiian com-munity. We are most fortunate to have the opportunityto conduct observations from this mountain.
REFERENCESAbazajian, K. N., Adelman-McCarthy, J. K., Ag¨ueros, M. A.,et al. 2009, ApJS, 182, 543Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009,ARA&A, 47, 481Berg, T. A. M., Ellison, S. L., Venn, K. A., & Prochaska, J. X.2013, MNRAS, 000, 00Berg, T. A. M., Neeleman, M., Ellison, S. L., & Prochaska, J. X.2014, PSAP, submittedBird, S., Vogelsberger, M., Haehnelt, M., et al. 2014, MNRAS he Physical Conditions of High- z Atomic Gas 17
Burkert, A., Genzel, R., Bouch´e, N., et al. 2010, ApJ, 725, 2324Cantalupo, S., Lilly, S. J., & Haehnelt, M. G. 2012, MNRAS, 425,1992Carswell, R., Becker, G., Jorgenson, R. A., Murphy, M., & Wolfe,A. M. 2012, MNRAS, 422, 1700Carswell, R., Jorgenson, R. A., Wolfe, A. M., & Murphy, M. 2010,MNRAS, 411, 2319Catinella, B., Haynes, M. P., Giovanelli, R., Gardner, J. P., &Connolly, A. J. 2008, ApJ, 685, L13Cen, R. 2012, ApJ, 748, 121Cooke, R., Pettini, M., & Jorgenson, R. A. 2014, AJ, In PrepDekel, A., & Birnboim, Y. 2006, MNRAS, 368, 2Dessauges-Zavadsky, M., Calura, F., Prochaska, J. X., D’Odorico,S., & Matteucci, F. 2004, A&A, 416, 79—. 2007, A&A, 470, 431Draine, B. T. 2011, Physics of the Interstellar and IntergalacticMedium (Princeton: Princeton University Press)Dutta, R., Srianand, R., Rahmani, H., et al. 2014, MNRAS, 440,307Field, G., Goldsmith, D., & Habing, H. 1969, ApJ, 155, L149Fumagalli, M., O’Meara, J. M., Prochaska, J. X., & Rafelski, M.2014, MNRAS, submittedFumagalli, M., Prochaska, J. X., Kasen, D., et al. 2011, MNRAS,418, 1796Gazol, A., V´azquez-Semadeni, E., & Kim, J. 2005, ApJ, 630, 911Heiles, C., & Haverkorn, M. 2012, Space Sci Rev, 166, 293Heiles, C., & Troland, T. 2003a, ApJS, 145, 329—. 2003b, ApJ, 586, 1067Herbert-Fort, S., Prochaska, J. X., Dessauges-Zavadsky, M., et al.2006, PASP, 118, 1077Howk, J. C., Wolfe, A. M., & Prochaska, J. X. 2005, ApJ, 622,L81Jenkins, E., & Tripp, T. M. 2011, ApJ, 734, 33Jorgenson, R. A., Wolfe, A. M., & Prochaska, J. X. 2010, ApJ,722, 460Kanekar, N., Prochaska, J. X., Smette, A., et al. 2014, MNRAS,438, 2131Kanekar, N. a. 2003, A&A, 399, 857Kaplan, K. F., Prochaska, J. X., Herbert-Fort, S., Ellison, S. L.,& Dessauges-Zavadsky, M. 2010, MNRAS, 122, 619Kereˇs, D., Katz, N., Weinberg, D. H., & Dav´e, R. 2005, MNRAS,363, 2Khare, P., Kulkarni, V. P., Lauroesch, J. T., et al. 2004, ApJ,616, 86Krogager, J.-K., Fynbo, J., Møller, P., et al. 2012, MNRAS, 424, 1Kulkarni, V. P., Meiring, J., Som, D., et al. 2012, ApJ, 749, 176Ledoux, C., Petitjean, P., Fynbo, J., Møller, P., & Srianand, R.2006, A&A, 457, 71Ledoux, C., Petitjean, P., & Srianand, R. 2003, MNRAS, 346, 209Lehner, N., Howk, J. C., Prochaska, J. X., & Wolfe, A. M. 2008,MNRAS, 390, 2Levshakov, S. A., Dessauges-Zavadsky, M., D’Odorico, S., &Molaro, P. 2002, ApJ, 565, 696Lu, L., Sargent, W. L., & Barlow, T. A. 1996, ApJ Supp., 107, 475—. 1998, AJ, 115, 55McKee, C., & Ostriker, J. P. 1977, ApJ, 218, 148 Morton, D. C. 2003, ApJS, 149, 205Neeleman, M., Wolfe, A. M., Prochaska, J. X., & Rafelski, M.2013, ApJ, 769, 54Noterdaeme, P., Laursen, P., Petitjean, P., et al. 2012, A&A, 540,63O’Meara, J. M., Burles, S., Prochaska, J. X., et al. 2006, ApJ,649, 61Petitjean, P., Srianand, R., & Ledoux, C. 2000, A&A, 364, L26Prochaska, J. X., Chen, H.-W., & Bloom, J. S. 2006, ApJ, 648, 95Prochaska, J. X., Gawiser, E., Wolfe, A. M., Castro, S., &Djorgovski, S. 2003a, ApJ, 595, L9Prochaska, J. X., Herbert-Fort, S., & Wolfe, A. M. 2005, ApJ,635, 123Prochaska, J. X., & Wolfe, A. M. 1997, ApJ, 487, 73—. 1999, ApJ Supp., 121, 369—. 2000, ApJ, 533, L5—. 2002a, ApJ, 566, 68Prochaska, J. X., Wolfe, A. M., Howk, J. C., et al. 2007, ApJS,171, 29Prochaska, J. X., Wolfe, A. M., Tytler, D., et al. 2001a, ApJSupp., 137, 21Rafelski, M., Wolfe, A. M., Prochaska, J. X., Neeleman, M., &Mendez, A. 2012, ApJ, 755, 89Richings, A., Schaye, J., & Oppenheimer, B. D. 2014b, MNRAS,442, 2780Roy, N., Kanekar, N., Braun, R., & Chengalur, J. 2013a,MNRAS, 436, 2352Roy, N., Kanekar, N., & Chengalur, J. 2013b, MNRAS, 436, 2366Sarazin, C., Rybicki, G., & Flannery, B. 1979, ApJ, 230, 456Savage, B. D., & Sembach, K. R. 1991, ApJ, 379, 245Schaye, J. 2001, ApJ Letters, 559, 1Silva, A., & Viegas, S. 2002, MNRAS, 329, 135Snowden, S., Egger, R., Freyberg, M., et al. 1997, ApJ, 485, 125Srianand, R., Petitjean, P., Ledoux, C., Ferland, G., & Shaw, G.2005, MNRAS, 362, 549Vladilo, G. 2004, A&A, 421, 479Vogt, S., Allen, S., Bigelow, B., et al. 1994, Proc. SPIE, 2198, 362Walch, S., Wunsch, R., Burkert, A., Glover, S., & Whitworth, A.2011, ApJ, 733, 47Wolfe, A. M., Gawiser, E., & Prochaska, J. X. 2003b, ApJ, 593,235—. 2005a, ARA&A, 43, 861Wolfe, A. M., Howk, J. C., Gawiser, E., Prochaska, J. X., &Lopez, S. 2004, ApJ, 615, 625Wolfe, A. M., Lanzetta, K. A., Foltz, C. B., & Chaffee, F. H.1995, ApJ, 454, 698Wolfe, A. M., Prochaska, J. X., & Gawiser, E. 2003a, ApJ, 593,215Wolfe, A. M., Prochaska, J. X., Jorgenson, R. A., & Rafelski, M.2008, ApJ, 681, 881Wolfire, M., Hollenbach, D., McKee, C., Tielens, A., & Bakes, E.1995, ApJ, 443, 152Wolfire, M., McKee, C., Hollenbach, D., & Tielens, A. 2003, ApJ,587, 278
Table 1
Fine Structure DLA sampleIndex QSO z abs log N HI Metallicity M a log N CII ∗ log N SiII ∗ log N SiII
N13 b ReferencesNumber (cm − ) [M/H] (cm − ) (cm − ) (cm − )1 Q1157+014 1.9437 21.70 ± − . ± > . c . ± . ± ± − . ± < . < .
56 15 . ± −
02 2.0395 21.65 ± − . ± > . c < .
58 16 . ± c Y 4, 74 J2340 − ± − . ± . ± < .
31 15 . ± −
19 2.0762 20.43 ± − . ± < . < .
83 13 . ± −
02 2.0950 20.70 ± − . ± . ± < .
05 15 . ± ± − . ± < . < .
22 14 . ± −
14 2.2794 20.56 ± − . ± < . < .
04 14 . ± − ± − . ± . ± < .
86 14 . ± ± − . ± . ± < .
30 15 . ± ± − . ± . ± < .
69 15 . ± ± − . ± . ± < .
23 14 . ± ± − . ± < . < .
23 14 . ± ± − . ± < . < .
96 14 . ± ± − . ± < . < .
41 13 . ± ± − . ± . ± < .
60 15 . ± ± − . ± > . < . > .
10 Y 1818 PKS1354 −
17 2.7800 20.30 ± − . ± . ± < .
56 14 . ± ± − . ± < . < . > .
48 Y 1020 Q1337+11 2.7958 20.95 ± − . ± < . < .
63 14 . ± ± − . ± . ± < .
46 14 . ± ± − . ± < . < .
64 13 . ± ± − . ± < . < .
58 14 . ± ± − . ± < . < . > .
85 Y 2125 Q1021+30 2.9489 20.70 ± − . ± < . < .
60 14 . ± ± − . ± . ± < .
82 15 . ± ± − . ± < . < .
78 14 . ± ± − . ± < . < .
08 14 . ± ± − . ± < . < . > .
37 Y 1830 Q0347 −
38 3.0247 20.60 ± − . ± . ± < .
17 15 . ± −
01 3.0620 21.20 ± − . ± . ± < . > .
87 Y 732 J1200+4015 3.2200 20.85 ± − . ± . ± < . > .
21 Y 2133 Q0930+28 3.2352 20.30 ± − . ± < . < .
16 13 . ± ± − . ± . ± < .
23 15 . ± ± − . ± . ± < .
26 15 . ± ± − . ± < . < .
64 14 . ± −
15 3.4388 20.92 ± − . ± . ± < .
41 15 . ± ± − . ± . ± < .
12 14 . ± −
03 3.7358 20.72 ± − . ± . ± < .
30 13 . ± ± − . ± < . < .
91 13 . ± ± − . ± < . < .
52 14 . ± ± − . ± < . < . > .
32 Y 6, 743 J0817+1351 4.2584 21.30 ± − . ± > . < . > .
93 Y 2144 J1100+1122 4.3947 21.74 ± − . ± . ± < . > .
85 Y 2145 J1607+1604 4.4741 20.30 ± − . ± . ± < .
80 14 . ± ± − . ± < . < . > .
55 Y 2147 J1202+3235 4.7955 21.10 ± − . ± . ± < . > .
80 Y 2148 J1051+3545 4.8206 20.35 ± − . ± < . < .
34 13 . ± ± − . ± > . c . ± . ± c N 20, 24, 2550 J0044+0018 1.7250 20.35 ± − . ± . ± < .
36 15 . ± ± − . ± > . c < .
11 15 . ± − ± − . ± . ± < .
88 16 . ± − ± − . ± . ± < . > .
11 N 15, 2554 J0233+0103 1.7850 20.60 ± − . ± < . < .
44 14 . ± ± − . ± . ± < .
24 15 . ± ± − . ± > . c . ± . ± c N 24, 2557 J1310+5424 1.8005 21.45 ± − . ± > . c . ± . ± c N 20, 24, 2558 J1106+1044 1.8185 20.50 ± − . ± < . < . > .
21 N 2559 J1142+0701 1.8407 21.50 ± − . ± > . c . ± . ± c N 24, 2560 J0815+1037 1.8462 20.30 ± − . ± < . < .
57 15 . ± ± − . ± . ± < .
12 15 . ± ± − . ± . ± < .
05 15 . ± ± − . ± > . c < . > . c N 24, 2564 J1042+0628 1.9429 20.70 ± − . ± < . < .
06 15 . ± ± − . ± > . c . ± > . c N 23, 24, 2566 J1552+4910 1.9599 21.15 ± − . ± . ± < .
49 15 . ± ± − . ± > . c . ± . ± c N 24, 2568 J1305+0924 2.0184 20.40 ± − . ± . ± < .
07 15 . ± ± − . ± . ± < .
94 16 . ± − ± − . ± > . c . ± . ± c N 22, 2471 J1211+0422 2.3765 20.70 ± − . ± < . < .
90 14 . ± ± − . ± . ± < . > .
67 N 2573 J0211+1241 2.5951 20.60 ± − . ± . ± < .
14 15 . ± he Physical Conditions of High- z Atomic Gas 19
Table 1 — Continued
Index QSO z abs log N HI Metallicity M a log N CII ∗ log N SiII ∗ log N SiII
N13 b ReferencesNumber (cm − ) [M/H] (cm − ) (cm − ) (cm − )74 J0812+3208 2.6263 21.35 ± − . ± . ± < .
78 15 . ± − ± − . ± < . < . > .
15 N 1476 FJ2334 −
09 3.0569 20.45 ± − . ± < . < .
97 14 . ± − ± − . ± . ± < .
31 15 . ± ± − . ± . ± < .
29 15 . ± ± − . ± . ± < .
00 14 . ± ± − . ± . ± < .
50 14 . ± References . — (1) Lu et al. (1996);(2) Prochaska & Wolfe (1997);(3) Lu et al. (1998);(4) Prochaska & Wolfe (1999);(5) Petitjean et al.(2000);(6) Prochaska & Wolfe (2000);(7) Prochaska et al. (2001a);(8) Levshakov et al. (2002);(9) Prochaska & Wolfe (2002a);(10)Prochaska et al. (2003a);(11) Dessauges-Zavadsky et al. (2004);(12) Khare et al. (2004);(13) Ledoux et al. (2006);(14) O’Meara et al. (2006);(15)Herbert-Fort et al. (2006);(16) Dessauges-Zavadsky et al. (2007);(17) Prochaska et al. (2007);(18) Wolfe et al. (2008);(19) Jorgenson et al.(2010);(20) Kaplan et al. (2010);(21) Rafelski et al. (2012);(22) Kulkarni et al. (2012);(23) Berg et al. (2013);(24) This Work;(25) Berg et al.(2014)a Ion used for metallicity determinationb Part of the Neeleman et al. (2013) samplec VPFIT used to determine the column density
Table 2
Results of Si II ∗ and C II ∗ TechniqueIndex QSO z abs log n H (2- σ ) log n e (2- σ ) log T(2- σ ) log P/k B (2- σ )Number [cm − ] [cm − ] [K] [K cm − ]1 Q1157+014 1.9437 1 . . . . − . − . − . − .
0) 2 . . . .
1) 3 . . . . − . . − . . − . − . − . − .
6) 1 . . . .
3) 0 . . − . . −
02 2.0395 1 . . . . − . − . − . − .
9) 1 . . . .
2) 3 . . . . − . . . . − . − . − . − .
1) 1 . . . .
2) 2 . . . . −
19 2.0762 − . . − . . − . − . − . − .
3) 1 . . . .
3) 0 . . . . −
02 2.0950 0 . . . . − . − . − . − .
3) 1 . . . .
3) 2 . . . . − . . − . . − . − . − . − .
6) 1 . . . .
3) 0 . . − . . −
14 2.2794 − . . − . . − . − . − . − .
4) 1 . . . .
3) 0 . . . . − . . . . − . − . − . − .
1) 1 . . . .
4) 2 . . . . − . . − . . − . − . − . − .
3) 1 . . . .
3) 1 . . . . − . . − . . − . − . − . − .
6) 1 . . . .
3) 0 . . . . . . − . . − . − . − . − .
3) 1 . . . .
4) 2 . . . . − . . − . . − . − . − . − .
6) 1 . . . .
3) 0 . . − . . − . . − . . − . − . − . − .
4) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
1) 1 . . . .
3) 1 . . . . − . . − . . − . − . − . − .
5) 1 . . . .
3) 1 . . . . − . . − . . − . − . − . − .
1) 1 . . . .
3) 1 . . . . −
17 2.7800 0 . . − . . − . − . − . − .
3) 1 . . . .
4) 2 . . . . − . . − . . − . − . − . − .
6) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
6) 1 . . . .
3) 0 . . − . . . . . . − . − . − . − .
2) 1 . . . .
3) 2 . . . . − . . − . . − . − . − . − .
4) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
5) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
3) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
5) 1 . . . .
3) 0 . . − . . − . . − . . − . − . − . − .
6) 1 . . . .
3) 0 . . − . . − . . − . . − . − . − . − .
5) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
6) 1 . . . .
3) 0 . . − . . − . . − . . − . − . − . − .
3) 1 . . . .
3) 1 . . . . −
38 3.0247 0 . . . . − . − . − . − .
2) 1 . . . .
4) 2 . . . . −
01 3.0620 − . . − . . − . − . − . − .
1) 1 . . . .
3) 1 . . . . − . . − . . − . − . − . − .
4) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
4) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
4) 1 . . . .
3) 1 . . . . . . − . . − . − . − . − .
4) 1 . . . .
3) 2 . . . . − . . − . . − . − . − . − .
4) 1 . . . .
3) 0 . . . . −
15 3.4388 0 . . . . − . − . − . − .
2) 1 . . . .
4) 2 . . . . . . . . − . − . − . − .
1) 1 . . . .
4) 2 . . . . −
03 3.7358 0 . . − . . − . − . − . − .
1) 1 . . . .
4) 2 . . . . − . . − . . − . − . − . − .
4) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
4) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
3) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
0) 1 . . . .
3) 1 . . . . − . . − . . − . − . − . − .
0) 1 . . . .
3) 1 . . . . . . . . − . − . − . − .
9) 1 . . . .
4) 3 . . . . − . . − . . − . − . − . − .
2) 1 . . . .
3) 1 . . . . − . . − . . − . − . − . − .
0) 1 . . . .
4) 1 . . . . − . . − . . − . − . − . − .
2) 1 . . . .
3) 0 . . . . . . . . − . − . − . − .
9) 2 . . . .
2) 4 . . . . . . . . − . − . − . − .
1) 1 . . . .
4) 2 . . . . . . . . − . − . − . − .
9) 1 . . . .
3) 3 . . . . − . . . . − . − . − . − .
1) 1 . . . .
9) 2 . . . . − − . . − . . − . − . − . − .
2) 1 . . . .
3) 1 . . . . − . . − . . − . − . − . − .
5) 1 . . . .
3) 0 . . − . . . . − . . − . − . − . − .
3) 1 . . . .
4) 2 . . . . . . . . − . − . − . − .
9) 2 . . . .
1) 4 . . . . . . . . − . − . − . − .
9) 2 . . . .
0) 3 . . . . − . . − . . − . − . − . − .
6) 1 . . . .
3) 0 . . − . . . . . . − . − . − . − .
9) 2 . . . .
4) 3 . . . . − . . − . . − . − . − . − .
5) 1 . . . .
3) 0 . . − . . . . . . − . − . − . − .
3) 1 . . . .
3) 2 . . . . . . . . − . − . − . − .
1) 1 . . . .
8) 2 . . . . − . . − . . − . − . − . − .
2) 1 . . . .
2) 0 . . − . . − . . − . . − . − . − . − .
5) 1 . . . .
3) 0 . . . . − . . − . . − . − . − . − .
2) 1 . . . .
3) 1 . . . . − . . − . . − . − . − . − .
4) 1 . . . .
2) 1 . . . . . . . . − . − . − . − .
0) 2 . . . .
4) 3 . . . . . . . . − . − . − . − .
2) 1 . . . .
2) 2 . . . . . . . . − . − . − . − .
2) 1 . . . .
3) 2 . . . . − . . . . − . − . − . − .
6) 2 . . . .
4) 4 . . . . − . . − . . − . − . − . − .
6) 1 . . . .
3) 0 . . − . . − . . − . . − . − . − . − .
2) 1 . . . .
3) 1 . . . . he Physical Conditions of High- z Atomic Gas 21
Table 2 — Continued
Index QSO z abs log n H (2- σ ) log n e (2- σ ) log T(2- σ ) log P/k B (2- σ )Number [cm − ] [cm − ] [K] [K cm − ]73 J0211+1241 2.5951 − . . − . . − . − . − . − .
5) 1 . . . .
4) 1 . . . . . . . . − . − . − . − .
2) 1 . . . .
7) 2 . . . . − − . . − . . − . − . − . − .
4) 1 . . . .
3) 0 . . . . −
09 3.0569 − . . − . . − . − . − . − .
6) 1 . . . . − . . − . . − . . . . − . − . − . − .
1) 1 . . . .
5) 2 . . . . − . . − . . − . − . − . − .
3) 1 . . . .
8) 1 . . . . − . . − . . − . − . − . − .
4) 1 . . . .
3) 1 . . . . . . − . . − . − . − . − .
1) 1 . . . .
4) 2 . . . ..