Cassucsd.edu

The Astrophysical Journal, 533:L5–L8, 2000 April 10᭧ 2000. The American Astronomical Society. All rights reserved. Printed in U.S.A.
METALLICITY EVOLUTION IN THE EARLY UNIVERSE Observatories of the Carnegie Institute of Washington, 813 Santa Barbara Street, Pasadena, CA 91101; [email protected] Department of Physics and Center for Astrophysics and Space Sciences, University of California, San Diego, C-0424, La Jolla, CA 92093 Received 1999 October 12; accepted 2000 February 28; published 2000 March 16 Observations of the damped Lya systems provide direct measurements on the chemical enrichment history of neutral gas in the early universe. In this Letter, we present new measurements for four damped Lya systems athigh redshift. Combining these data with [Fe/H] values culled from the literature, we investigate the metallicityevolution of the universe from z ≈ 1.5 to 4.5. Contrary to our expectations and the predictions of essentiallyevery chemical evolution model, the N(H i)–weighted mean [Fe/H] metallicity exhibits minimal evolution overthis epoch. For the individual systems, we report tentative evidence for an evolution in the unweighted [Fe/H]mean and the scatter in [Fe/H], with the higher redshift systems showing lower scatter and lower typical [Fe/H]values. We also note that no damped Lya system has [Fe/H] ! Ϫ2.7 dex. Finally, we discuss the potential impactof small number statistics and dust on our conclusions and consider the implications of these results on chemicalevolution in the early universe.
Subject headings: galaxies: abundances — quasars: absorption lines 1999). To date, the chemical abundances of over 40 systemshave been measured, the majority with z = 1.5–3 where the The damped Lya systems—neutral hydrogen gas layers identification and follow-up observations of damped Lya sys- identified in the absorption-line spectra of background qua- tems are most efficient. These studies argue that at z ≈ 2, the sars—dominate the neutral hydrogen content of the universe mean metallicity of the damped systems is approximately at all epochs. At high redshift, these systems are widely 1/10–1/30 solar metallicity ([Zn/H] ≈ Ϫ1.1, [Fe/H] ≈ Ϫ1.5) accepted as the progenitors of present-day galaxies for the with a large scatter from nearly solar to less than 1/100 solar following reasons: (1) their very large H i column densi- metallicity. At very high redshift (z 1 3), the picture is far less certain. Focusing on a sample of seven z 1 3 damped Lya dr/r k 100, i.e., these are virialized systems at high redshift; systems, Lu, Sargent, & Barlow (1997) noted that the metal- (2) they contain the majority of neutral gas in the early universe licity of these systems is significantly lower than the z ! 3 and are therefore the reservoirs for galaxy formation; and observations. In turn, they argued that z ≈ 3 marked the epoch at z ≈ 2–3 is consistent with the mass at which significant star formation begins, a claim with im- density of stars today (Wolfe et al. 1995). While the physical portant implications for the processes of galaxy formation.
nature of the damped Lya systems is still controversial (Pro- In this Letter, we present new measurements on the metal- chaska & Wolfe 1997; Haehnelt, Steinmetz, & Rauch 1998; licity of four damped Lya systems (including three at z 1 Maller et al. 2000; Le Brun et al. 1997), by studying the chem- 3.5 and together with the data from Prochaska & Wolfe (1999) ical abundances of the damped Lya system one directly traces double the sample at z 1 3. The new full sample—including the chemical enrichment history of the universe at high redshift.
the systems from Lu et al. (1996, 1997)—reveals evidence for Observing damped Lya systems is equivalent to poking sight little change in the N(H i)–weighted mean metallicity of the lines through the interstellar medium of protogalaxies. Because neutral universe from z ≈ 1.5 to 4.5, contrary to the predictions these observations are biased to H i cross section and the H i of essentially every chemical evolution model. On the other gas mass of a system is proportional to ∫ jN, one can measure hand, we find tentative evidence for an evolution in the un- global properties of the universe simply by weighting the mea- weighted mean and scatter of [Fe/H] for individual damped surement from each damped system by N(H i). The observa- Lya systems. Finally, we comment on the robustness of these tions also afford an efficient means for examining the char- results, speculate on the implications for chemical enrichment, acteristics of individual protogalaxies in the early universe. In and discuss the prospects for future advances.
this Letter, we examine the metallicity of the damped Lyasystems from z ≈ 1.5 to 4.5, which places tight constraints on chemical evolution models (e.g., Pei, Fall, & Hauser 1999), aswell as a valuable consistency check on star formation rate To determine the metallicity of a damped Lya system, one must accurately measure the neutral hydrogen column density Over the past decade, several groups have surveyed the me- N(H i) and a metallicity indicator, typically either Zn or Fe. In tallicity of the damped Lya systems from z ≈ 1 to 4 (Pettini stellar population studies of the Galaxy, one traditionally uses et al. 1994, 1997, 1999; Lu et al. 1996; Prochaska & Wolfe Fe as the metallicity indicator, primarily as a matter of con-venience. Since we are studying gas-phase abundances, how-ever, we must account for the possible depletion of Fe onto 1 Visiting Astronomer, W. M. Keck Telescope. The Keck Observatory is a joint facility of the University of California and the California Institute of dust grains or instead choose an element like Zn which is minimally affected by depletion. Unfortunately, there are both a Average of Fe ii l1144 (log gf = 0.105) and Fe ii l1608.
b Fe ii l1608.
c Fe ii l1611.
theoretical and observational disadvantages to using Zn as themetallicity indicator. Theoretically, Zn has a very uncertainchemical origin. It is referred to as an Fe peak element becauseit traces Fe in Galactic stars (Sneden, Gratton, & Crocker 1991),yet the leading theory on the production of Zn proposes thatit forms in the neutrino-driven winds of Type II supernovae(Hoffman et al. 1996). Furthermore, recent measurements of[Zn/Fe] in metal-poor stars (Johnson 1999) and thick disk stars pairs for the damped Lya systems observed by Prochaska & Wolfe (1999) and in this study (squares) and Lu et al. (1996, (J. X. Prochaska et al. 2000, in preparation) suggest that Zn/ 1997; stars) with HIRES on the Keck I telescope. The open circles correspond Fe is enhanced relative to the Sun by ϩ0.1 to ϩ0.3 dex, perhaps to the N(H i)–weighted mean metallicity for the systems at z consistent with a Type II origin. Observationally there are com- 4.5] Note that the difference in these means is small, plications with measuring Zn in the damped Lya systems,where one must rely on two weak Zn ii transitions with and Lu et al. (1996, 1997), the total [Fe/H] sample is 39 sys- l rest ≈ 2000 A˚. The transitions are so weak that even at high tems, 15 with z 1 3. The systems were chosen independent of resolution and high signal-to-noise ratio, Zn can only be de- any prior metallicity measurements; the only possible biases tected in damped systems when log [N(H i)] ϩ [Zn/H] 1 are due to the magnitude-limited selection of the quasars (e.g., 19.0 (e.g., [Zn/H] 1 Ϫ1.3 for systems with N(H i) ≈ N Fall & Pei 1993), which will be discussed in the following Most important to this study, however, the large rest wavelength of the Zn ii transitions prevents one from readily measuringZn in z 1 3 damped Lya systems because it is difficult to make resolution spectrographs. In fact, at the time of publication, we Figure 1 plots the 39 [Fe/H] values versus z are not aware of a single accurate Zn measurement for any chaska & Wolfe sample (squares) and the sample of damped z 1 3 damped Lya system. Therefore, we will focus on Fe in Lya systems observed by Lu et al. (1996, 1997) (stars). To this Letter, which has two singly ionized transitions at l explore evolution in the metallicity of the damped Lya systems, ˚ with a complement of f-values ideal for measuring the we consider three moments of the metallicity data in two red- abundance of Fe in damped systems. We restrict the analysis to Fe measurements made with HIRES on the Keck I telescope (1) the N(H i)–weighted mean metallicity of neutral gas, AZ S; (Vogt 1992), specifically the systems observed by Prochaska (2) the unweighted mean metallicity, A[Fe/H]S, of the set of & Wolfe (1999) and Lu et al. (1996, 1997) and the additional systems introduced here. In addition to providing a homoge- deviation of [Fe/H] in these protogalaxies, j([Fe/ neous data set which has been analyzed with the same tech- The first moment represents the global metallicity of all neu- niques, these observations account for nearly every damped system with an accurate Fe abundance at z 1 1.5 and every ing each [Fe/H] measurement by the corresponding H i column AZ S { log [͸ N(Fe )/ ͸ N(H i)] Ϫ log (Fe/H),. Com- puting the mean for the damped Lya systems at the two epochs, from observations we acquired in 1998 February and 1999 March with HIRES on the Keck I 10 m telescope. The data 49 The errors on the AZ S values reflect only the statistical was reduced with the MAKEE software package developed by uncertainty in measuring N(Feϩ) and N(H i) and were derived T. Barlow, and the column densities were derived primarily with standard error propagation techniques. Below we estimate from the Fe ii ll1608, 1611 transitions with the apparent op- the uncertainty due to small number statistics. Comparing the tical depth method (Savage & Sembach 1991). We adopt the AZ S values, we note that they favor no significant evolution in oscillator strengths from Cardelli & Savage (1995), noting that the mean metallicity of neutral gas from z = 1.5 to 4.5. If we our conclusions on the evolution of [Fe/H] in the damped Lya include the tentative result from Pettini et al. (1999) that the systems are not sensitive to their values. The N(H i) values for Zn mean metallicity does not change from z ≈ 1 to 3, then one these systems are taken from the literature (Wolfe et al. 1995; concludes there is no evidence for significant metallicity evo- Storrie-Lombardi & Wolfe 2000) and are the dominant source lution from z = 1 to 4.5, an interval spanning more than of error in the [Fe/H] values. Finally, we evaluate [Fe/H] as- 3 Gyr. The other two moments, the unweighted mean suming the meteoritic Fe abundance [e(Fe) = 7.50; Grevesse A[Fe/H]S = (1/n) ͸ [Fe/H] and the scatter & Sauval 1999] without adopting any ionization corrections, sensitive to the chemical enrichment history within individual which is an excellent assumption for all but possibly the lowest protogalaxies since each damped system is given equal weight.
N(H i) damped systems (Prochaska & Wolfe 1996). Together For the two intervals, we find that the mean logarithmic abun- with the published measurements of Prochaska & Wolfe (1999) Performing the Student’s t-test and the F-test on the two mo-ments, we find that the A[Fe/H]S and j([Fe/H]) statistics for thetwo epochs are inconsistent at the 90% and 80% confidencelevel. Therefore, there is tentative evidence for chemical evo-lution in the individual damped Lya systems with the z ! 3sample showing a higher typical metallicity and a larger scatterin [Fe/H] from system to system.
To address the robustness of these results, one must consider several issues. First, because AZ S is dominated by the systemswith the largest N(H i) values, this mean is robust only in sofar as the total N(H i), H iT of a single damped Lya system. Figure 2 plots the [Fe/H],N(H i) pairs for all 39 systems; the circles are members of thez sample and the triangles those of the z cmϪ2, which is a factor of 4 larger than the largest Fig. 2.—Thirty-nine [Fe/H], N(H i) pairs for the damped Ly the full sample. The circles correspond to z N(H i) ! 10 cmϪ2 show a large scatter ] and 10 times greater than most of the known in [Fe/H], the large N(H i) systems all have [Fe/H] ≈ Ϫ1.8.
damped Lya systems. As such, we consider the mean derivedfrom the z sample to be reasonably robust. The primary potential pitfall is if the optical surveys have systematically the same epoch? (2) Dust obscuration could remove damped Lya systems from the magnitude-limited samples, which would if dust obscuration is significant (discussed further below). The significantly alter the conclusions [e.g., metal-rich, high situation is far more uncertain for the z 1 3 sample where N(H i) systems]. With respect to the first concern, we can cmϪ2, comparable to the N(H i) of the Q0458 estimate the maximum dust correction to [Fe/H] via the mea- sured Zn/Fe ratio. Again, Zn is essentially undepleted in the the largest N(H i) system in the z gas phase so that [Zn/H] = [Fe/H] ϩ [Zn/Fe] may be more recent surveys suggest there are very few z 1 3 damped systems representative of the true metallicity in the damped Lya sys- cmϪ2 (Storrie-Lombardi & Wolfe 2000), tems. This practice is limited, however, by the fact that Zn may we caution that the mean we have derived for the z be produced in Type II supernovae (Hoffman et al. 1996) such is a tentative result. For example, the system toward that supersolar Zn/Fe ratios would be representative of nucle- Q0000Ϫ2619 has significant bearing on AZ S osynthesis, not dust depletion. Furthermore, recent results on dance has been difficult to determine (Prochaska & Wolfe 1999; the [Zn/Fe] ratio measured in Galactic stars shows that Lu et al. 1996). Ironically, removing it from the z [Zn/Fe] 1 ϩ0.2 dex in very metal poor stars ([Fe/H] ! Ϫ ([Zn/Fe] ≈ ϩ0.1 in 10 stars; J. X. Prochaska et al. 2000, in Fe ii l1611 profile from this system. Meanwhile, lowering preparation) in thick disk stars with [Fe/H] 1 Ϫ1. Therefore, [Fe/H] by 0.6 dex to establish consistency with the Ni and Cr while the typical [Zn/Fe] value in the damped Lya systems is ϩ0.4 dex with relatively small scatter (Pettini et al. 1997; Pro- chaska & Wolfe 1999), it is unclear what fraction is due to that small number statistics are still important in evaluating the dust depletion. Nonetheless, if we take [Zn/H] as the true me- AZ S . One can estimate the uncertainty associated with the tallicity indicator, AZ S and the unweighted mean are enhanced small number statistics of the two samples by performing a by ≈0.4 dex, but there is very little change in the observed bootstrap statistical analysis. For each sample, we indepen- scatter. The potential effects of dust depletion on the statistical dently calculated AZ S 500 times by randomly drawing n objects (n is the number of damped systems in a given redshift interval) is no accurate Zn determination for any z 1 3 damped Lya from each interval. In turn, we can estimate the effects of system. To estimate the depletion level, we can compare the cosmic variance on our results by calculating the standard de- relative abundance patterns (in particular the Si/Fe ratio) of Si/Fe has been measured in the z 1 3 systems, one finds which is based on only 15 systems, is considerably less certain [Si/Fe] ≈ ϩ0.3 dex, nearly identical to the typical value of the measurement. The difference in the AZS z ! 3 sample. While the similarity of a metal ratio like Si/Fe stresses the outstanding need for future observational programs does not require similar dust depletion levels, the z 1 3 [Si/Fe] to focus on z 1 3 damped systems.
values do imply depletion levels of at least 0.3 dex. Therefore, Any study on the chemical abundances of the damped Lya unless one takes the unlikely stance that the z 1 3 systems are systems must assess the potential effects of dust. With respect significantly more depleted than the z to this analysis, where we have taken Fe as the metallicity minimal evolution in the depletion levels of the damped sys- indicator, there are two important aspects to consider. (1) If we tems and no significant impact on any of our conclusions. The need to correct the observed [Fe/H] values by some factor to effects of biasing due to dust obscuration are more difficult to obtain the true metallicity of each system, does the mean cor- address. Note in Figure 2 the absence of any rection evolve in time and/or differ from system to system at cmϪ2 systems with [Fe/H] ∼ Ϫ1. While this may be due to small number statistics or the fact that very few regions exist cmϪ2 systems exhibit low metallicity. In in the early universe with large N(H i) and [Fe/H] տ Ϫ1, the particular, in the z 1 3 sample only two of 15 damped systems trend could also be explained by dust obscuration. Fall & Pei show [Fe/H] 1 Ϫ1.5 dex. One possible explanation for the dif- (1993) have presented an excellent framework for addressing ference is systems that have just formed have preferentially the effects of dust depletion on damped Lya statistics. Their low N(H i) and [Fe/H]. The trend is also suggestive of the calculations indicate that if the logarithmic scatter in the dust- correlation Cen & Ostriker (1999) find between overdensity to-gas ratio k is small (less than 1 dex), then only a small and metallicity in their numerical simulations. The problem correction to the mean optical depth and in turn to AZ S is re- remains, however, in explaining why the highest metallicity quired (S. M. Fall 1999, private communication). For a constant sample also have low N(H i). Lastly, recall dust-to-metals ratio—implied by the nearly constant [Zn/Fe] that there is tentative support for an evolution in both the scatter values—the logarithmic scatter in k ≈ j([Fe/H]), and we have shown j([Fe/H]) ≤ 0.5 for the two samples. Therefore, we ex- more recently formed systems tend to have lower metallicity, pect dust obscuration to have a minimal effect (!0.2 dex) on then the evolution may easily be explained by a larger mean and scatter in the age of the damped Lya systems at z ≈ 2.
The results in this Letter on the evolution of the metallicity Furthermore, the systems at z ≈ 2 may have larger masses and of neutral gas in the universe and individual protogalaxies pre- sent an unexpected picture. A number of groups have estimated Finally, we stress that only two systems from the full sample the chemical evolution of neutral gas at high redshift (Malaney have [Fe/H] ! Ϫ2.5 and the large majority show [Fe/H] 1 Ϫ2.
& Chaboyer 1996; Edmunds & Phillips 1997; Pei et al. 1999), As first noted by Lu et al. (1997), there appears to be a threshold and essentially every treatment predicts a substantial (10.5 dex) to the minimum metallicity of the damped Lya systems at increase in the mean metallicity from z = 4 to 2. While a 0.5 ≈1/100 solar metallicity. Therefore, even out to z ≈ 4.5 there dex evolution is consistent with our results at the 3 j level, is no evidence for damped Lya systems with primordial abun- the current observations favor a very mild evolution in AZ S. If dance. This places a further constraint on chemical evolution future observations lend further support for this conclusion, the models. As we probe higher and higher redshift without de- theoretical models will require significant revision. Of course, tecting primordial gas, one may be forced toward one of the these theoretical treatments depend sensitively on a number of following conclusions: (1) star formation proceeds rapidly factors that are uncertain: (1) the SFR, (2) the initial mass (!107 yr) to bring the metallicity to 1/100 solar after the for- function, (3) the mass distribution of protogalaxies, (4) the loss mation of a damped system; (2) either the damped system or of metals to the intergalactic medium, (5) the yield of various its progenitors have been undergoing star formation for a elements, etc. Therefore, there is considerable theoretical free- lengthy time; and/or (3) a generation of Population III stars dom to bring the models into agreement with the observed lack of evolution. Nonetheless, the Lyman break galaxies offer in- The future prospects for improving the z 1 3 observational controvertible evidence that significant star formation is taking sample are excellent. While fast progress with HIRES and sim- place from z = 3 to 4 (Steidel et al. 1998) such that the total ilar instruments is limited by the faintness of z 1 4 quasars, the metal content of the universe must be increasing. Unless these new Echellette Spectrograph and Imager instrument on Keck metals are enriching only ionized regions (an unlikely sce- II will be ideal for surveying the N(H i) and metal content of nario), then to explain the minimal evolution in AZ S the total a large (N 1 20) sample of very high redshift damped Lya H i content of the universe must be increasing at nearly the systems. We intend to pursue such a program over the next same rate as the metal content. It is intriguing to note that this few years, taking full advantage of the ever increasing sample is qualitatively consistent with the evolution of Q of known z 1 4 quasars (Fan et al. 1999).
by Storrie-Lombardi & Wolfe (2000) for the damped Lyasystems.
We would like to thank A. McWilliam, E. Gawiser, M. Pet- The other statistical moments are sensitive to the chemical tini, and M. Fall for insightful discussion and comments. We enrichment history within individual galaxies. Comparing the thank T. Barlow for providing the HIRES reduction package.
unweighted mean with the weighted mean, we find that We acknowledge the very helpful Keck support staff for their A[Fe/H]S is less than AZ S at both epochs. While the difference efforts in performing these observations. J. X. P. acknowledges is not large (≈0.1–0.2 dex), it does highlight the fact that many support from a Carnegie postdoctoral fellowship.
Cardelli, J. A., & Savage, B. D. 1995, ApJ, 452, 275 Pettini, M. 2000, in Lecture Notes in Physics, Proc. ESO Workshop, ed. J.
Cen, R., & Ostriker, J. P. 1999, ApJ, 519, L109 Walsh & M. Rosa (Berlin: Springer), in press (astro-ph/9902173) Edmunds, M. G., & Phillips, S. 1997, MNRAS, 292, 733 Pettini, M., Ellison, S. L., Steidel, C. C., & Bowen, D. V. 1999, ApJ, 510, Fall, S. M., & Pei, Y. C. 1993, ApJ, 402, 479 Pettini, M., Smith, L. J., Hunstead, R. W., & King, D. L. 1994, ApJ, 426, 79 Grevesse, N., & Sauval, A. J. 1999, A&A, 347, 348 Pettini, M., Smith, L. J., King, D. L., & Hunstead, R. W. 1997, ApJ, 486, 665 Haehnelt, M. G., Steinmetz, M., & Rauch, M. 1998, ApJ, 495, 647 Prochaska, J. X., & Wolfe, A. M. 1996, ApJ, 470, 403 Hoffman, R. D., et al. 1996, ApJ, 460, 478 Johnson, J. 1999, Ph.D. thesis, Univ. California, Santa Cruz Le Brun, V., Bergeron, J., Boisse´, P., & Deharveng, J. M. 1997, A&A, 321, Savage, B. D., & Sembach, K. R. 1991, ApJ, 379, 245 Sneden, C., Gratton, R. G., & Crocker, D. A. 1991, A&A, 246, 354 Lu, L., Sargent, W. L. W., & Barlow, T. A. 1997, preprint (astro-ph/9711298) Steidel, C. C., Adelberger, K., Dickinson, M., Giavalisco, M., Pettini, M., & Lu, L., Sargent, W. L. W., Barlow, T. A., Churchill, C. W., & Vogt, S. 1996, Storrie-Lombardi, L. J., & Wolfe, A. M. 2000, ApJ, submitted Malaney, R. A., & Chaboyer, B. 1996, ApJ, 462, 57 Vogt, S. S. 1992, in ESO Workshop on High Resolution Spectroscopy with Maller, A. H., Prochaska, J. X., Somerville, R. S., & Primack, J. R. 2000, the VLT, ed. M.-H. Ulrich (Garching: ESO), 223 Wolfe, A. M., Lanzetta, K. M., Foltz, C. B., & Chaffee, F. H. 1995, ApJ, 454, Pei, Y. C., Fall, S. M., & Hauser, M. G. 1999, ApJ, 522, 604

Source: http://casswww.ucsd.edu/archive/personal/awolfe/pub/metal_evol_2000.pdf