First stars in Damped Lyman Alpha systems
aa r X i v : . [ a s t r o - ph . C O ] N ov Mon. Not. R. Astron. Soc. , 1–5 () Printed 27 August 2018 (MN L A TEX style file v1.4)
First stars in Damped Lyman Alpha systems
Stefania Salvadori ⋆ & Andrea Ferrara Kapteyn Astronomical Institute, Landleven 12, 9747 AD Groningen, The Netherlands Scuola Normale Superiore, Piazza dei Cavalieri 7, 56126 Pisa, Italy
ABSTRACT
In order to characterize Damped Lyman Alpha systems (DLAs) potentially host-ing first stars, we present a novel approach to investigate DLAs in the context ofMilky Way (MW) formation, along with their connection with the most metal-poorstars and local dwarf galaxies. The merger tree method previously developed is ex-tended to include inhomogeneous reionization and metal mixing, and it is validatedby matching both the Metallicity Distribution Function of Galactic halo stars andthe Fe-Luminosity relation of dSph galaxies. The model explains the observed N HI -Ferelation of DLAs along with the chemical abundances of [Fe/H] < − z abs ≈ .
34 C-enhanced DLA (Cooke et al. 2011a),pertains to a new class of absorbers hosting first stars along with second-generationlong-living low-mass stars. These ”PopIII DLAs” are the descendants of H -coolingminihalos with M h ≈ M ⊙ , that virialize at z > T g ≈ (40 − M ∗ ≈ − M ⊙ . Key words: galaxies: abundances, evolution, stellar content - cosmology : theory -stars: Population II
Damped Ly α absorption systems (DLAs) are high-columndensity neutral gas reservoirs, N HI ≥ . cm − , observedat intermediate redshifts, z ≤
5, in the spectra of dis-tant quasars. Although their nature is still unclear (Pet-tini 2004), the key role of DLAs to understand galaxyformation (Wolfe, Gawiser & Prochaska 2005) is widelyrecognized. So far, more than 1000 DLAs have been ob-served and the iron abundance measured in ≈
150 sys-tems [
F e/H ] ≈ [ − . , − .
5] (Prochaska et al. 2007). Amongthese, very metal-poor (VMP) DLAs with [Fe/H] < −
2, canbe used to study the initial phases of heavy element en-richment of the interstellar medium (ISM) of early galaxies.Indeed, if VMP stars observed today in the Galactic haloand in nearby dwarf spheroidal galaxies (dSphs) are the liv-ing fossils of the first stellar generations, VMP DLAs maywell constitute the gas reservoir out of which such pristinestellar populations formed. Following the medium resolu-tion study of VMP DLAs by Penprase et al. (2010), Cookeet al. (2011b) have recently presented a high spectral res-olution sample, including 22 VMP systems. In these DLAs[C/O] ≈ − . ≈ − . ⋆ E-mail:[email protected]
VMP Galactic halo stars (Fabbian et al. 2009). So far, theonly exception to this general trend is represented by a DLAwith [Fe/H] ≈ − N HI = 10 . ± . cm − observed inthe spectrum of the QSO J0035-0918 at z abs ≈ .
34 (Cookeet al. 2011a). This system has [C/Fe]= 1 .
53, i.e. ≈
20 timeslarger than any other DLA. Moreover, its abundance pat-tern shows a clear ’odd-even’ effect and is consistent withthe predictions for the yields of Z = 0 faint supernovae (SN)with m ∗ ≈ M ⊙ (Kobayashi et al. 2011). Are we observingfor the first time a DLA whose gas retains the imprint leftby the first stars?To characterize DLAs potentially hosting the first starsor their ashes, we propose a novel approach that simultane-ously follows the evolution and chemical properties of DLAsand their connection with the most metal-poor stars andgalaxies observed in the Local Universe based on the re-sults obtained using the merger-tree code GAMETE (Sal-vadori, Schneider & Ferrara 2007, SSF07; Salvadori, Ferrara& Schneider 2008; Salvadori, & Ferrara 2009, SF09). The model key points can be summarized as follows (seeSSF07 and SF09 for details):(i) Hierarchical merger histories of a MW-sized dark mat- c (cid:13) RAS
S. Salvadori & A. Ferrara
Figure 1.
Left panel : observed (points with error-bars) and simulated (histogram) metallicity distribution function of Galactic halo stars.The data points are the sample by Beers & Christlieb (2006) with the inclusion of the three hyper-iron poor stars (Christlieb et al. 2002,2006; Frebel et al. 2005). The histogram is the average MDF value over 50 merger history of the MW, re-normalized to the total numberof observed stars with [Fe/H] ≤
2. The shaded area represents ± σ errors. Right panel : observed (points with error-bars) and simulated(contours) iron-luminosity relation for the MW dSph galaxies. The data points are by Kirby et al. (2008) with the inclusion of the twofaintest dSphs (Willman et al. 2005, Geha et al. 2009). The colored shaded areas correspond to regions that include the (99 , , ter (DM) halo are reconstructed from z = 20 via a MonteCarlo algorithm based on the Extended Press-Schechter the-ory (SSF07).(ii) Star formation (SF) is followed along the tree in halosexceeding a mass threshold, M sf , whose evolution (Fig. 2)is governed by: (a) the photo-dissociating Lyman-Werner(LW) background, quenching H -formation in T vir < Kminihalos; (b) the gas temperature in ionized regions of theGalactic Medium (GM), preventing gas-infall in halos with T vir lower than a threshold value, T th . We assume that SFis active in T vir > T th = 2 × K minihalos at z > z , T th (and hence M sf ) isassumed to linearly increase up to the value set by the endof reionization T th ≈ × K (Kitayama et al. 2000) for z < z rei = 6.(iii)
Inhomogeneous reionization is modeled by randomsampling the reionization history implied by M sf ( z ) toswitch off (on) gas accretion in minihalos that form in ion-ized (neutral) regions.(iv) The SF rate is taken to be proportional to the massof cold gas, ˙ M = ǫ ∗ M g /t ff , where ǫ ∗ is the SF efficiencyand t ff the halo free-fall time. In minihalos ǫ ∗ is reducedas ǫ = ǫ ∗ [1 + ( T vir / × K) − ] − due to the ineffectivecooling by H molecules (SF09). Low-mass, Population II(PopII) stars form according to a Larson IMF when the gasmetallicity exceeds Z cr = 10 − . Z ⊙ (Schneider et al. 2002).At lower metallicity, PopIII stars form with a reference massvalue m ∗ = 25 M ⊙ and explosion energy E SN = 10 ergconsistent with faint SNe.(v) The abundance evolution of different chemicalelements † (from C to Zn) is traced in both the ISM and † For m ∗ < M ⊙ stars we use yields by van den Hoek &Groenewegen (1997) ( Z ≥ − ) and by Meynet & Maeder(2002) ( Z ≤ − ); for more massive stars we use Woosley &Weaver (1995) with a systematic halving of the Fe yield (Timmes, in the GM by taking into account mass- and metallicity-dependent stellar evolutionary timescales (Raiteri, Villata& Navarro 1996) and SN feedback (Salvadori, Ferrara &Schneider 2008).(vi) To account for the incomplete mixing of SN ejectawithin the ISM of gas-poor galaxies, gas outflows have ametallicity Z w = Z ISM + η ( M g ) M Z /M ej , where M Z is themass of newly formed metals, M ej is the mass of gas ejectedout of the halo, and η is a function of the gas mass η =0 . .
65 tanh[(log10( M g ) − . / .
0] that varies in the range η = [0 , P ( z ) = Q Z /Q δ>δ c , where Q Z ( z ) = 1 − exp(Σ i πR b ( i ) /V MW ( z )) is the filling factorof metal bubbles within the MW physical volume ‡ , and Q δ>δ c ( z ) is the volume filling factor of fluctuations withoverdensity above the critical threshold, δ > δ c = 1 . first penetrate (Tornatore,Ferrara & Schneider 2007). Objects in enriched (primordial)regions are assigned an initial metallicity Z vir = Z GM /Q Z ( Z vir = 0).The model is calibrated by best-fitting the SF and feed-back efficiencies to reproduce the global properties of theMW (stellar/gas mass and metallicity). In Fig. 1 we showthat the average metallicity distribution of [Fe/H] < − M h < M σ ,candidates to remain MW satellites (Diemand, Madau & Woosley & Weaver 1995); yields for faint SNe are from Kobayashiet al. (2011). ‡ We assume V MW ( z ) = 5(1 + z ) − Mpc .c (cid:13) RAS, MNRAS , 1–5 irst stars in Damped Lyman Alpha systems Figure 2.
Dark matter content of MW satellites ( M h < M σ ) asa function of their formation redshift. The two lines show the evo-lution of: the minimum mass of star-forming haloes, M sf (solid),the halo mass corresponding to 2 σ density peaks (dashed). Thedotted shaded area identifies the region populated by H -coolingminihalos with T vir < K. The number labels and colorsspecify different populations of satellites (see the text). The grayshaded area delimits the evolution of M sf predicted by assumingan early (upper limit) and a late (lower limit) reionization history(Gallerani et al. 2007). Moore 2001; SF09 for details). In Fig. 2 M sf ( z ) is comparedwith that obtained by using data-constrained reionizationmodels. To this end we compute the dissociating LW back-ground intensity associated with an early/late reionizationhistory (Gallerani, Choudhury & Ferrara 2006) by using theresults provided in Fig. 6 by Ahn et al. (2009); we then con-vert the flux into a critical mass following Machacek, Brian& Abel (2001) and extrapolate to higher masses. At eachredshift our M sf is within the range allowed by the tworeionization histories (gray shaded area). We now apply our model to DLAs starting from the ori-gin of the C-enhanced, [Fe/H] ≈ − z abs = 2 . h z i ≈ . ± . P = Q Z /Q δ> . >
1. Afterward, newly virializ-ing halos form from metal enriched GM gas, whose initialabundances [X/H] vir = h [X/H] i GM /Q Z . Since at z = 2 . Q Z = 0 . ± .
005 and h [Fe/H] i = − . ± .
02, proto-galaxies with an initial iron abundance [Fe/H] < − . < σ at its final assembling epoch in order to evolve in iso-lation (no further merger or accretion) and become a satel-lite. By inspecting Fig. 2 we can distinguish among differentpopulations of objects that satisfy this condition: (1) Ly α -cooling halos that assembled after the end of reionization; Fornax and LeoI, the most luminous among the observeddSphs, are predicted to belong to this domain; (2) star-forming Ly α -cooling halos that assembled before the end ofreionization; classical dSphs, L > L ⊙ , such as Sculptor(SF09) are members of this population. (3) H -cooling, inef-ficiently star-forming minihalos that appeared at z ≈ − ”sterile” halos that formed before the end of reionization and unableto trigger SF ( M h < M sf ). We selected the satellite candi-dates in 50 possible merger histories of the MW, and evolvethem in isolation down to redshift z = 2 .
34, when we com-pute their H I column density, N HI , assuming that the gasis fully neutral with a molecular weight µ = 1 . N HI ≈ n HI r g ≈ π M g /µm p α r vir = 5 . × M g α (cid:16) M h M ⊙ (cid:17) − / (cid:16)
101 + z form (cid:17) − cm − (1)the gas radius is written as r g = αr vir , with α = 0 . α = 0 . N HI > × cm − (Wolfe et al. 2005). Interestingly, wefind that systems which populate Zone (2), i.e. the progeni-tors of classical dSphs, have already exhausted most of theirgas by z = 2 .
34, i.e. log( N HI / cm − ) ≪ .
3. On the otherhand, many objects among those formed in Zones (1), (3)and (4), have H I column densities compatible with DLAs.We will then focus on these candidates. A comparison between the properties of candidate and ob-served DLAs at z ≈ .
34 is given in Fig. 3 and Fig. 4. Ob-jects formed in different Zones result in distinct DLA popu-lations, differing in their chemical abundances and N HI . Sys-tems in Zone (3) are segregated in the C-rich, metal-poor,low- N HI regions of the plots, matching the properties of theC-enhanced DLA. What is the origin of these systems?According to our model, C-enhanced DLAs are hostedby M h ≈ . − . M ⊙ H -cooling minihalos, which virial-ized before the end of reionization out of metal-free , neu-tral regions of the MW environment. Because of their lowDM content the star formation history of such a primordialproto-galaxies is extremely short. Some tens of Myrs afterthe onset of star-formation the intensity of LW backgroundbecomes high enough ( M sf > M h ) to suppress further SFactivity, thus turning them into ”sterile” minihalos. Metal-free stars are hosted by this population of DLAs, to whomwe will hereafter refer as PopIII DLAs.The metal-enrichment by first stars gently proceeds inthese objects, mainly because of ineffective cooling by H molecules and their low gas-mass content, M g ≈ . − . M ⊙ , favoring metals and gas loss ( M ejg ≈ − M ⊙ ).As soon as Z > Z cr = 10 − . Z ⊙ , however, ”normal” PopIIstars can form and contribute to enrichment. Given theyields by faint SNe, the PopIII-to-PopII transition occurswhen [Fe/H] cr ≈ − .
8, implying that most of the iron ob-served in these systems originates from these second gener-ations of stars. The mass of relic stars in PopIII DLAs is c (cid:13) RAS, MNRAS , 1–5
S. Salvadori & A. Ferrara
Figure 3.
Gas abundances of DLAs with respect to the solar values (Asplund et al. 2009). Points are the observed values in VMP DLAsat z abs = (2 . − .
21) (triangles, Cooke+2011b) and in the C-enhanced DLA at z abs = 2 .
34 (square, Cooke+2011a). The color shadedareas show the results of the model for the three different populations of DLAs at z ≈ .
34: PopIII (blue), very metal-poor (violet), andmetal-poor DLAs (orange). For each population the intensity of the colors correspond to regions containing the (99 , , expected to vary between M ∗ ≈ − M ⊙ , the most star-rich systems being the most enriched ones. The rise of theC abundance at increasing metallicity (Fig. 3) reflects thegradual contribution by low-metallicity AGB stars, mainlyproducing C, some O, and limited amounts of N (for sys-tems with [Fe/H]= − . ± . − . ± . − . ± . − . ± . N HI decrease (Fig. 4) is caused by gas consumption in themost star-rich systems, due to both astration and gas loss.Note that our model does not predict any [C/O] increase to-wards low [O/H] values, as reported by medium resolutionobservations (Penprase et al. 2010).The properties of VMP DLAs are well reproduced (Fig.3) by [Fe/H] < − starless systems formed in Zone (4), pas-sively evolving since their assembly epoch, 6 < z <
10. Theyform through merging of primordial and metal-enriched pro-genitors, virialized out of neutral regions of the MW envi-ronment. The metallicity spread of these systems dependson the gas enrichment in progenitor minihalos. Abundanceratios, instead, closely reflect those of the GM at the timeof formation. Although the overall GM metallicity increaseswith time, abundances ratios get locked to the dominantstellar population which contribute to the enrichment, i.e. type II SNe . As a consequence, [C/O] and [C/Fe] ratiosshow little dispersion and are independent of Z , but alsoof redshifts, because of the passive evolution of these star-less DLAs. These findings are in perfect agreement with thenew observational results by Becker et al. (2011).Finally, [Fe/H] > − T vir > K halos form-ing through merging of star-rich progenitors either before(Zone 4) or after (Zone 1) the end of reionization. Their sys-tematically higher Z is a consequence of self-enrichment by asubstantial ( M ∗ ≈ − . M ⊙ ) population of long-livingstars. Similarly, the high [C/Fe] ratios reflect the strong con-tribution by AGB stars. The gas mass in these DLAs is M g = 10 − . M ⊙ , thus resulting in higher N HI values,despite of the lower formation z and larger M h ≈ − M ⊙ (eq. 1). Note however that the N HI derived for these DLAsmust be interpreted as an upper limit. In fact, M h > M sf objects (Fig. 3) are actively star-forming at z = 2 .
34, with˙ M ≈ (0 . − M ⊙ yr − , and hence part of their gas is pre- Figure 4. N HI distribution vs. iron abundances of DLAs. Pointswith error-bars are observed values for metal-poor DLAs at z abs = (1 . − .
9) (circles, Prochaska et al. 2007), VMP DLAsat z abs = (2 . − .
21) (triangles, Cooke+2011b) and for the C-enhanced DLA at z abs = 2 .
34 (square, Cooke+2011a). The colorshaded areas show the results of the model for the three differentpopulations of DLAs predicted to exist at z = 2 .
34 (see Fig. 3). sumably ionized. Moreover, the [Fe/H] value ([C/Fe]) hasto be interpreted as a lower (upper) limit, since the contri-bution by SN type Ia may be significant in star-rich DLAs(Calura, Matteucci & Vladilo 2002). The same interpreta-tion has to be applied to [Fe/H] measurements, as Fe isdepleted onto dust grains (Noterdaeme et al. 2008). Barredthese limitations, included the unknown number of surviv-ing satellites, we note that the apparent concentration of[Fe/H] > − N HI − F e plane(Fig. 4) implies that these absorbers are only partially rep-resentative of the progenitors of MW-like systems (Pontzenet al. 2008).
The recently discovered z abs ≈ .
34 C-enhanced DLA per-tains to a new class of systems, dubbed PopIII DLAs, host- c (cid:13) RAS, MNRAS , 1–5 irst stars in Damped Lyman Alpha systems ing the first stars along with second generations of long-living stars. These systems are associated to H -coolingminihalos that virialize at z > . Once ”sterilized” these minihalos pas-sively evolve as an inert gas reservoir, M g ≈ . − . M ⊙ ,retaining the imprint of the stellar generations they hosted.Their gas temperature, T g , is regulated by the balance be-tween molecular/metal radiative cooling and photo-heatingby the external ionizing radiation (Black 1981): n HI Λ( T g , Z ) = h ǫ i K ph (2)where n HI is the gas density, Λ( T g , Z ) the gas coolingrate (Maio et al. 2007), h ǫ i ≈
20 eV the mean UVbackground photon energy, and K ph the optically-thickH I photo-ionization rate (Abel & Mo 1998). K ph ≈ . × − J ( N HI / cm − ) − β s − with β = 1 . J = 1.We find that [Fe/H] > − T g ≈ (40 − < T vir , implying that these systems can likely sur-vive photo-evaporation thanks to self-shielding and metalcooling. Their gas temperature increases with decreasingmetallicity, in agreement with the results by Kanekar et al.(2009). The C-enhanced DLA has T g ≈
70 K while in[Fe/H] ≈ − Z ≈ − Z ⊙ ) systems T g ≈ z abs ≈ .
34 DLA does not result from Z = 0 faint SNe, butrather from the enrichment by low-metallicity SNII andAGB stars, which may start to form as soon as Z > Z cr =10 − ± Z ⊙ . While SNII nucleosynthetic products are mostlylost in winds, AGB metals are retained in the ISM, caus-ing a dramatic increase of [C/Fe]. The chemical evolution of[Fe/H] > − independent on the assumed yields andIMF of PopIII stars, confirming the role of ordinary PopIIstars in driving the enrichment of VMP systems (SSF07),recently emphasized by the detection of a Z ≤ × − Z ⊙ star with a ”normal” chemical abundance pattern (Caffauet al. 2011). If Z cr < − Z ⊙ , as suggested by the exis-tence of such star, the PopIII-PopII transition would be evenquicker. The mass of relic stars in PopIII DLAs is foundto be M ∗ ≈ − M ⊙ , implying that they are the gas-richcounterpart of the faintest dSphs.As stated, the C-enhanced DLA is a minihalo. However,we cannot exclude a different interpretation, in which suchabsorber might be a newly formed halo virializing at z ≈ . M h > . M ⊙ ≈ M sf ( z = 2 . N HI . By determining the darkmatter content and SF rate of the C-enhanced DLA, it wouldbe possible to disentangle these two pictures. We finally notethat our simple semi-analytical model, which holds similar-ities with that proposed by Abel & Mo (1998) for LymanLimit Systems, prevent us from making specific predictionson the number of DLAs at z = 2 .
34. However, the relativecontribution of C-enhanced DLAs to the total population isexpected to be extremely low, ≈ . ACKNOWLEDGEMENTS
We thank R. Cooke, P. Molaro, P. Petitjean, X. Prochaska &R. Schneider for enlightening comments on the draft versionof the paper. S.S. acknowledges a NOVA fellowship grantedby the Netherlands Research School for Astronomy.
REFERENCES
Abel T. & H. J. Mo, 1998, ApJL, 494, 151Ahn K., Shapiro P., Iliev I. T., Mellema G. & Pen U. L., 2009,ApJ, 695, 1430Asplund M., Grevesse N., Sauval A. J., Scott P., 2009, ARA&A,47, 481Becker G. D., Sargent W. L. W., Rauch M. & Carswell R. F.,ApJ acceptedBeers T. C., Christlieb N., 2006, private communicationBlack J. H., 1981, MNRAS, 197, 553Calura F., Matteucci F., Vladilo G., 2003, MNRAS, 340, 59Caffau E. et al., 2011, Nature, 477, 67Cooke R. et al. , 2011a, MNRAS, 412, 1047Cooke R. et al. , 2011b, MNRAS, 417, 1534Christlieb N. et al. , 2002, Nat, 904, 419Diemand J., Madau P., Moore B., 2005, MNRAS, 364, 367Dijkstra M., Haiman Z., Rees M. J., Weinberg D. H., 2004, ApJ,601, 666Fabbian D., Nissen P. E., Asplund M., Pettini M. & Akerman C.,2009, A&A, 500, 1143Ferrara A., Pettini M. & Shchekinov Y., 2000, MNRAS, 319, 539Frebel A., Christlieb N., Norris J. E., Aoki W., Asplund M., 2005,Nat, 434, 871Kanekar N. et al., 2009, ApJ, 705, L40Kobayashi C., Tominaga N., Nomoto K., 2011, ApJ, 730, 14Kirby E. N., Simon J. D., Geha M., Guhathakurta P., Frebel A.,2008, ApJ, 685, 43Kitayama T., Tajiri Y., Umemura M., Susa H., Ikeuci S., 2000,MNRAS, 315, 1Gallerani S., Choudhury T. & Ferrara A., 2006, MNRAS, 360,1401Geha M. et al., 2009, ApJ, 629, 1464Machacek M. E., Brian G. L. & Abel T., 2001, ApJ, 548, 509Meynet G. & Maeder A., 2002, A&A, 390, 561Maio U., Dolag K., Ciardi B. & Tornatore L., 2007, MNRAS, 379,963Miralda-Escud J., Haehnelt M. & Rees M. J., 2000, ApJ, 530, 1Noterdaeme et al. 2008, A&A, 481, 327Penprase et al. 2010, ApJ, 721, 1Pettini M., 2004, Cambridge University Press 257Pontzen A. et al., 2008, MNRAS, 390, 1349Prochaska J. X. et al., 2007, ApJs, 171, 29Raiteri C. M., Villata M. & Navarro J. F., 1996, A&A, 105, 315Salvadori S., Schneider R. & Ferrara A., 2007, MNRAS, 381, 647Salvadori S., Ferrara A. & Schneider R., 2008, MNRAS, 386, 348Salvadori S. & Ferrara A., 2009, MNRAS, 395, 6Schneider R., Ferrara A., Natarajan P., Omukai K., 2002, ApJ,571, 30Timmes F. X., Woosley S. E. & Weaver T. A., 1995, ApJs, 98,617Tornatore, A. Ferrara & Schneider R., 2007, MNRAS, 382, 945van den Hoek L. B., Groenewegen M. A. T., 1997, A&AS, 123,305Willman B. et al. , 2005, ApJ, 129, 2692Wolfe A. M., Gawiser E., Prochaska J. X., 2005, ARA&A, 43, 861Woosley S. E., Weaver T. A., 1995, ApJ, 101, 181c (cid:13)
RAS, MNRAS000