THE LOCAL HIGH REDSHIFT UNIVERSE THE EXTREMELY METAL

THE LOCAL HIGH REDSHIFT UNIVERSE – THE EXTREMELY METAL POOR STARS IN THE GALACTIC HALO Why study very metal poor stars ? We can study the early epochs of the Galaxy, the local equivalent of the z ~ 5 Universe, with objects that are bright enough to be found in large numbers and to be analyzed in considerable detail. We can study the kinematics of the early Galaxy and its halo. We can study the onset of chemical evolution in the Galaxy, the possible stellar sources which produced many elements at very early epochs (very massive stars and SNII). We can probe the age of the galaxy (Th and U dating) and the relationship between the halo field stars and the Galactic globular clusters. 1

Binding energy/nucleon – up to Fe/Ni, fusion releases energy. Heavier than Fe/Ni, fusion is energetically prohibited; fission releases energy. Most of the available energy is released when fusing 4 H atoms to He. 2

The onion-layer characteristic of highly evolved massive stars, a consequence of the ever increasing T required to burn various nuclear fuels, H to He, He to C, C to Si …to Fe. . Only SNII contribute to EMP stars, no SNI. 3

r and s-process neutron captures produce specific isotopes. From Sneden & Cowan, 2003, Science, 299, 70 4

Solar Abundances Beyond the Fe-peak, produced by n-addition, Separated by s and r processes of n-addition. 5

How Many SN Contribute to an EMP Star’s Chemical Inventory ? 6

Timescale to reach 1/3000 Solar Metallicity • Assume SN rate at least as large as now, 1 SN/100 yr in our galaxy • Assume solar abundance ratios. • All material for < -3. 5 dex is then made by t << 10**6 yr. • The EMP stars thus represent the relics of the early stages of the formation of our Galaxy. They are very old. • Their effective redshifts are very high (z>4). • The EMP stars represent the local high redshift Universe (near field cosmology) 7

Energy Considerations Assume 1 Msun of material (mostly Fe) is ejected per SNII at ejection velocity of 10, 000 km/sec. The amount of mass (assumed to be pristine H) swept up when the ejecta has decelerated to the level of typical clouds in the proto. Galaxy (assumed to be vf = 50 km/sec) is then (by conservation of energy) M = 4 x 10^4 Msun x (Mej/1 Msun) x (50/vf)^2, and the resulting [Fe/H] is So 1 SN cannot pollute the entire halo mass. Within this swept up mass, [Fe/H] is Solar, so there must be considerable mixing or overlap between the ejecta of individual SNII. 8

The Holy Grail of Cosmochronology – Does It Exist ? • Use radioactive decays to age date EMP halo stars, get age of halo, add a small amount for the initial formation of the proto-halo, get age of Universe • Must measure N(decay)(mostly lead)(now) / N(parent in the past), but lead is made in other ways as well • Easiest: Th/Eu ratio, but these are far apart in the periodic table, do we understand their production well enough ? Probably no • Results from Th/Eu from different EMP stars are inconsistent • Th/U is better, much closer in periodic table • Th lines are hard to detect, U lines are almost impossible to detect • Half life – Th 13. 5 Gyr, U 4. 5 Gyr 9

Galactic chemical evolution models for EMP stars • EMP stars are very old, so fewer production sources contribute (at high metallicity, many production sources with varying timescales can contribute, much more complex to model). • Only SNII and any pre-galactic stars (first stars, VMS, Pop III) contribute, lifetimes and mass distributions for these sources • Metal production and ejection into the ISM by each relevant source • Issues: volume of pollution, mixing of ejecta, mass outflows from the galaxy, continued accretion into the proto-galaxy halo • Stellar IMF for first generation, star formation rate f(t), mass loss • Role of binaries 10

Solar abundances, atomic numbers 1 -32. Isotopic ratios mostly measured just for Sun Sum for all stable isotopes of an element. Heavier than Fe-peak, very low abunds. 11

The 0 Z Project The core group: Judy Cohen (Caltech), Andy Mc. William, Steve Shectman, Ian Thompson (Carnegie), Norbert Christlieb (Hamburg) Current and former postdocs: Inese Ivans, Jorge Melendez, Solange Ramirez (Caltech), Innocenza Busa, Franz-Josef Zickgraf (Hamburg) Undergraduate students: Amber Swenson (Caltech), Berit Behnke (Hamburg) 12

The 0 Z Project Plan • Goal: to find and study a large sample of extremely metal poor stars • We use the Hamburg/ESO survey to generate candidate lists. We have exclusive access to half of the HES fields. • We have taken about 1700 mod. resolution follow-up spectra, highest quality class candidates first, and by brightness. About 600 from 200 -inch at Palomar, about 1100 from 6. 5 m Magellan Telescope at Las Campanas. • We are 99. 9% complete (to B=17. 5 mag, i. e. the HES limit) in the fall -north fields (about 900 deg sq). • Those with [Fe/H](HES) < -2. 9 dex from P 200 spectra are to be observed at HIRES/Keck – about 75 such stars observed to date, at present only 8 such stars have not yet been observed 13

Finding EMP stars (EMP: Fe abund. < 1/1000 Solar) • Rare: roughly 1/square degree to B(lim) ~ 17 mag • Use objective prism spectra to search for candidate EMP stars. These are low quality, low disp spectra. • Since contamination by higher [Fe/H] stars is severe, candidates must be checked with mod. res. spectra. So 3 stage process, HES, then follow up, then HIRES for most interesting • HK Survey, late 1980 s, Preston & Shectman • Hamburg/ESO Survey, 2000+, digitzed and fainter. Original purpose – find bright QSOs. Norbert Christlieb, in charge of stellar survey. 14

Examples of HES objective prism spectra of candidate metal poor stars from the HES (digitized photographic plates) (from Christlieb 2003) 15

Sample Mod. Res. Follow Up Spectra of HES EMP candidates from 200 -inch Hale Telescope. Measure Ca. II and Balmer line indices. 16

3933/Ca. II index versus Hdelta index for ~800 stars from the HES with mod. res. specta from Magellan (from Ian Thompson) with Carnegie Fe/H calibration 17

The assignment of [Fe/H](HES) • We use the algorithm of Beers et al (1999) to assign [Fe/H](HES) from analysis of the moderate resolution spectra. • Two indices are used, KP, measuring the absorption in the 3933 A line of Ca. II, and HP 2, measuring the strength of Hd. • An index measuring the strength of the G band of CH (GP) is also measured. • We use the definitions of these indices from Beers et al (1999). 18

Size of analyzed sample of candidate EMP stars from the HES as f(time) (1700 Follow Up Spectra In Hand) • Papers through 2004, abundance analyses of individual stars of interest, no attempt at a statistical sample • Ap. JL now in press, early 2005, 497 stars, all DBSP/P 200 through spring 2004) • Current sample shown here : 732 spectra of EMP candidates (~600 DBSP/P 200 + 120 Magellan), remove duplicates and rejects, yields 663 different EMP/VMP stars with ~21 rejects (galaxies or M dwarfs, d. Me) • Final sample (expected ready summer 2006), all 1700 follow-up spectra in hand analyzed, expect about 1600 stars when duplicates and rejects removed. Now assembled, working on this. 19

• Accuracy checks of our results from the moderate resolution spectra: • 57 stars with duplicate spectra (runs in 2 different months on P 200 or P 200+Mag obs. ) • “Standard stars” – comparison of values inferred from our spectra with published indices and [Fe/H] values. • The HP 2 index (Hd) is the most uncertain. It’s the weakest in giants, and located adjacent to other strong features. 20

HIRES/Keck Spectra obtained for the most interesting EMP stars only: Spectra for EMP dwarfs (and 1 EMP giant) in the region of the 4215 A Sr. II line 21

Spectral synthesis of the region of the CH and C 2 bandheads in the dwarf carbon star HE 0007 -1832 22

Teff, log(g) for 62 EMP candidates from the HES analyzed HIRES spectra so far. Note: log(g) is from isochrone and Teff. Includes 16 C*, 3 C-enh. , 24 C-normal dwarfs, 19 C -normal giants 23

Errors in our detailed abundance analysis using 1 d model stellar atmosphere, (mostly) LTE, line by line analysis • Random errors: Uncertainty in Teff (from broad band colors) is the biggest contributor, surface gravity and vt also Unc. in the measurements of the strength of the absorption features in the spectra Random errors in transition probs. and other atomic data • Systematic errors: Does the model atmosphere adequately represent T(depth) ? Is the assumption of non-LTE or the specific non-LTE corrections adopted for several elements adequate ? Is the absolute scale of the gf values OK (lab measure) ? 24

A check of the analysis for Fe/H in a star using the many Fe. I lines in its spectrum. Is the deduced Fe abundance independent of the EP of the line ? of the strength of the line ? of the wavelength of the line ? If not, something is wrong in the analysis. 25

Fe ionization equilibrium – a stringent test of the validity of our abundance analyses 26

Deviations from the mean for Ca/Fe versus Ti/Fe. Circular distribution suggests random errors. From Cohen et al (2004) 27

Deviations from the mean for Cr/Fe versus those for Ti/Fe. Circular distribution suggests random errors. (Cohen et al 2004) One star is a probable real outlier. 28

Current measurements of s[X/Fe] by all groups are very small ! (examples from Arnone et al 2005) 29

Comparison of our results with the SNII yields of Umeda & Nomoto for a range of progenitor mass and explosion energy, from Cohen et al (2004). Note oddeven problems. 30

Ti/Fe for a sample of halo field stars and galactic globular clusters. Red – GCs, green – stars from the HES 31

Ba/Fe as a function of Fe/H for halo field stars, stars from the HES, and globular clusters. Ba production is disconnected from Fe production at low metallicity. 32

Ways to explain the low s values for [X/Fe] in EMP stars and, very recently, also found for DLAs at z 2 to 4 • All early SN (all SNII) were essentially identical, same mass, same production and ejection yields • Many SNII contributed to the gas which became a single low mass EMP star • But the observed abundance patterns require SNII of relatively high mass, where yields depend a lot on M, on the mass cut • The proto-galactic halo was very well mixed 33

Peculiar Stars among the EMP stars • Highly C-enhanced stars, including Carbon stars • UMP stars ([Fe/H] ~ -5. 3 dex) • Extreme s-process stars (all of these are C-rich stars) • Extreme r-process stars • Stars with symptoms of substantial r and s-process • Occasional genuine outlier in some Fe-peak element, strong Cr and Mn enhancements seen in 1 star. 34

Roads to Peculiar Abundances These stars are very metal poor. A very small amount of additional material of a relatively rare element can produce a big change in its abundance, assuming the material does not diffuse into the interior. Mass transfer in a binary system where the primary is more massive, in the AGB phase, and transferring material onto the low mass secondary. Today the primary is a WD, and the secondary, which we see, has a surface contaminated with enhanced C and s-process elements. 35

Problem with Cstar [Fe/H](HES). For coolest Cstars its too low by a factor of 10 compared to results of HIRES analyses. Offset for dwarfs due to lower Teff scale we adopt 36

Spectra of 2 C-stars and a C-normal star, KP and HP 2 index feature and continuum bandpasses 37

e(C) and C/N Ratio Among C-rich Stars, HIRES Suggestion for a Constant C/H of ~ 1/5 Solar 38

Strength of G Band of CH versus V-K color (Teff) (663 stars) 39

GP indices (G band of CH) measured from synthetic spectra of M. Briley. Grid calculated for 2 values of [Fe/H] 40

Histogram of C/Fe for candidate EMP giants from the HES for 2 ranges of [Fe/H]. HIRES [Fe/H] used when available. 41

Fraction of Cenhanced giants as a function of [Fe/H], with and without HIRES Fe. Much smaller when correct [Fe/H] is used for C-stars ! 42

C/H vs Teff for EMP giants. note trend of Cdepletion for cooler, more luminous, more evolved giants Stars at bottom, GP index too low for calibration 43

Large isotopic separation for 4740 A band of C 2 (but not for stronger (0, 0) band at 5160 A. Easy to derive C 12/C 13 ratio ! 44

C 12/C 13 For EMP C-stars from CH and C 2 45

Ba/Fe for EMP C-stars, 85% high, 15% low, C-normal stars from HES also shown 46

Ba/C abundances for the C-rich stars, with 10 EMP C-stars from the literature added, 85% high Ba (s-process) 47

Are elements besides CNO and s-process affected in Cstars ? HIRES results say NO for Na through Ni, median [X/Fe] Species [Na/Fe] [Mg/Fe] [Al/Fe] [Ca/Fe] [Sc/Fe] [Ti/Fe] [Cr/Fe] [Mn/Fe] N(C-stars) [X/Fe] s EMP C-normal dwarfs 3 0. 27 0. 22 0. 41 12 0. 55 0. 27 0. 56 10 0. 27 0. 39 -0. 09 *** 14 0. 54 0. 36 0. 31 5 0. 39 0. 26 0. 24 *** 15 0. 43 0. 26 0. 36 14 0. 43 0. 21 0. 36 12 -0. 30 0. 21 -0. 23 *** only 1 line used, possible blending by CH or CN 48

CMD diagram, C-stars, Ba-poor stars, known binaries marked 49

Mass Transfer in Binary Systems Abundances Afterward 50

Our Hypothesis for the Origin of C-rich ([C/Fe] > 1. 0 dex) Stars Important clue – constant C/H, ~1/5 Sun is consistent with observations in all C -stars analyzed from our sample 85% must be binary mass transfer, high C and high s-process is the signature of ~4 Msun AGB stars Remaining 15% tends to be more metal poor, suggesting this is result of mass transfer in more metal poor stars, where the s-process runs all the way to lead, the last stable element, no big enhancement of Ba produced, or insufficient n for s-proc. Predictions: expect Ba-poor C-stars to have low [Fe/H] as is seen. Maybe high Pb in most metal poor C-stars, very hard to detect, At [Fe/H] ~ -1. 8 dex, C-stars become CH stars (C up, but C<O), at still higher [Fe/H], CH stars become Ba stars (as observed) 51

A new short period double-lined spec. binary from HES 52

53

Fe/H histogram of HES sample, C-stars black, 663 stars, note absence of C-stars with [Fe/H] -1. 8 dex 54

The Yield of the HES for Metal. Poor Stars Cleaned: 3. 5 % <-3. 0 1. 8% < -3. 2 0. 7% < -3. 4 55

UMP stars with [Fe/H] < -5 dex • There are no such stars among the 1700 spectra we have taken at Palomar or at Las Campanas. There are no stars in our sample below -4. 0 dex. • There are only 2 known for about 7, 000 searched • UMP stars are very hard to find and may masquerade as [Fe/H] -4 stars. • Both are very C-rich with ~1/10 C/H of Sun • There are no stars known between ~-4 and -5. 3 dex. • Are these stars a continuation of the [Fe/H] -3 to -4 stars ? • Are these also binary mass transfer remnants ? 56

The first UMP found, having Fe/H < 10**-5 that of the Sun (See Christlieb 2002, Nature) 57

The UMP stars as a continuation of the EMP stars 58

The Extreme r-process Stars • Normal EMP stars show no sign of the s-process, this does not begin to contribute until [Fe/H] ~ -2 dex, • They show evidence for a small r process component • Extreme r-process stars show r-dominated elements (Eu, Th, U) enhanced by a factor of more than 100 • Extreme r-process events require close proximity to a source of a lot of neutrons • HST role: UV spectra (mostly STIS) yields lines for some key elements, Ga Ge Os Pt Au, and a better Pb line than in the optical 59

Classic extreme r-process star CS 22892 -052 (Sneden. . 2005) 60

CS 22892 -052 vs Solar r, Sneden et al (2003) 61

Suggested phenomenological model: N(elem E)=A x p(E, light) + B x p(E, Fe-peak) + C x p(E, r, Fe-seed) + D x p(E, s, Fe-seed) A, B, C, and D are independent ! Key elements in each process: Light – CNO through Mg - SNII SNIa, Fe-peak - Fe, Ni (Ca to Zn) s process – Ba, La r process – Eu (hard to observe all r-process elements) 62

Summary of our observational results for C-star Fe/H from analysis of HIRES spectra a) The [Fe/H] values for cool C-stars given by the standard tools used by the HES (and formerly by the HK project) give values too low by ~1. 0 dex. b) The sample appears divided into C-normal stars and a much smaller number of C-rich stars, many of which are C -stars. c) 85% of the C-rich stars are also s-process rich with low C 12/C 13 d) When one corrects for (a), the fraction of C-rich stars with [C/Fe] > 1. 0 dex is ~14%, and is probably independent of 63 [Fe/H]

Our hypothesis for the formation of C-rich EMP stars • We believe that all C-rich EMP stars are the remnants of mass transfer in a binary system when the (former) primary was an AGB star • C-rich EMP stars are not produced by exotic SN ! • This includes both s-process highly enhanced and s-normal stars, with the latter having lower Fe/H and perhaps very high lead • We suggest that such mass transfer involves material with a fixed upper bound to C/H of ~1/5 Solar • We suggest that at higher [Fe/H], addition of the same AGB material produces CH stars, then at near Solar Fe-metallicity, produces Ba stars 64

Remaining Puzzles - What to do next ? • We have reduced the zoo of EMP stars to order. There are Cnormal stars and descendents of mass transfer binary systems. • For the C-normal stars, how is their very low s of X/Fe produced? No current theoretical models for the first SN and the early formation of the Milky Way reproduce this. • Can we confirm that more of the C-rich stars are/were binaries ? Is there any other way to do this than extensive and expensive vr surveys ? • We need to find more Ba-por C-rich stars and check their binarity – are they rich in Pb ? • Can we measure a definitive isotope ratio in these stars (Ba !) 65

Next steps, part 2 • How can we find more UMP stars in an efficient manner • How are UMP stars formed ? • What is the relationship between the UMP and the EMP stars? 66

• End of STSc. I Colloq. Feb 15, 2006 67

A Periodic Table with the Production Mechanism Indicated. 68

Solar abundances, atomic numbers 32 to 92 69

Predictions of s as a Function of [Fe/H] from Argast (2001) (SN progenitor mass in circles) 70

Dwarfs versus Giants/Subgiants • Dwarfs are hotter and have weaker spectral lines (BAD) • Many more elements can be analyzed in a giant for a given SNR spectrum (GOOD) • Dwarfs are unevolved, no nuclear processed material can get to the surface from the stellar interior (GOOD) • The surface of a luminous giant may be contaminated by internal nuclear processing + convective mixing (BAD) • Overall EMP giants are much easier to isolate and study than EMP dwarfs • Dwarf/giant boundary used here is Teff 6000 K 71

Analysis procedures for the HIRES spectra • Only freedom is above or below turnoff for stars near MSTO. • Use automatic eq. width program of Ramirez & Cohen, then a well defined procedure of manual checks for weak lines. • Fixed template list of lines, gf values, HFS corrections, damping constants. Fixed list of molecular lines for C and N abundances. • Use MOOG code of Sneden modified for batch operation. • v_t determined for some stars, v_t –Teff relation used for most • Except for the Keck Pilot Project (Sep 2000 run), I have done all the observing, data reduction and analysis thus far myself. • For the Keck Pilot Proj. , E. Carretta and R. Gratton did the analysis. I did the stellar parameters and data reduction. 72

Ca/Fe for 40 Cnormal stars, HIRES, red = giants, blue =dwarfs. s(dwarfs) = 0. 10 dex, <[Ca/Fe]> = +0. 25 dex Giant anal. not full checked yet. 73

Sc, Ti, and Mn/Fe versus Fe/H for normal -C candidate EMP stars (dwarfs blue, giants red) Each s(dwarfs) < 0. 15 dex Sc, Ti, Mn/Fe 35 to 39 stars, s 0. 21, 0. 18, 0. 28(trend ? ) dex <X/Fe]> 0. 20, 0. 28, 0. 63 dex Giant analyses not checked yet. 74

Mean [X/Fe] and s for 24 EMP dwarfs (Cohen et al 2004) 75

Histogram of log[e(C)] (i. e. C/H) for EMP candidate giants from the HES in 2 ranges of [Fe/H]. HIRES [Fe/H] used when available. 76

C 2 Bandhead Strengths for HES C-stars (HIRES) obs. and pred. 77

La/Eu (proxy for s/r) as a function of Fe/H (Simmerer/2005) Normal stars, EMP heavy is r, s rises gradually with Fe/H 78