Obesity in the Universe: How Did Early Type Galaxies Become so Fat? Richard Ellis & Drew Newman (Caltech) CIFAR April 6th 2012

Bimodal Galaxy Distribution Bell et al Star forming Blue Late type Young Passive Red Early type Old Hubble Sequence - morphology shows dynamically distinct populations Gas content/integrated colors - different ages and star formation histories

Color-magnitude Relation at z~2 Kriek et al 2009 Ap J 705, L71 Photometric and dynamical studies of early-type galaxies over 0<z<1 confirm that the most massive examples formed the bulk of their stars prior to z~2 And, sure enough, red and dead galaxies were eventually found at z~2 Logically these would seem to be the precursors of most of today’s massive early- types (although continued formation is also possible)

Modest Increase in Mass Density of Red Galaxies Log Fractional Contribution to Total Stellar Mass 0.4<z<1.4 Bundy et al Ap J 651, 120 (2006) Spectroscopically- complete surveys of galaxies with IR- based stellar masses show only a modest increase (30%) in the abundance of M>10 11 M  red galaxies over 0.4<z<1.4 (5 Gyr) at the expense of truncated SF in blue galaxies Rising Reds Declining Blues

The big surprise: z~2 early-types are SMALL! SDSS 2 < z < 3 HST sizes of a representative sample of z~2-3 red galaxies with M >10 11 M  : r e ~0.9 kpc 2-5 times smaller than comparably massive z~0 ellipticals! Growth in size but not mass? half- light radius van Dokkum et al (2008) Earlier claims by: Daddi et al (2005), Trujillo et al (2006)

The `Red Nuggets’ Problem Initial skepticism at observational claims: result depends on combination of photometrically-determined masses and HST sizes of distant galaxies Perhaps mass overestimated e.g. “bottom light” IMF, AGB stars Perhaps size underestimated due to surface brightness effects Dynamical data to confirm masses Improved HST data over 0<z<2 How much growth occurred in self-consistent samples? Separating growth of long-lived populations vs new arrivals What does it all mean?!

How Big Should a Massive Galaxy Be? Ask a Theorist Wuyts et al 2010 Ap J 722, 1666

Size Depends: I – On Gas Fraction of Initial Merger local spheroids z~2 red galaxies Wuyts et al 2010 Ap J 722, 1666 increasing dissipational gas fraction f gas Equal mass merger SPH simulations using GADGET-2 with gas cooling, multi- phase ISM and SN/AGN feedback (Springel, Hernquist et al) Remnant is smaller for suitably-scaled z~3 disks with high gas fractions f gas  r e ~ exp( -f gas /0.3)

Size Depends: II – On Epoch of Merger Since gas fraction f gas declines with time, later merger products are larger NB: In principle this could account for expansion in size from z~2 to 0 but such a simple explanation is ruled out by low rate of major merging and absence of significant decline in abundance of fixed mass spheroids for z<1 Hopkins et al 2010 MN 401, 1099

Keck LRIS-R : I AB <23.5; hr exposures, 1.1 < z < 1.60 Newman et al 2010 Ap J 707, L103

Size evolution at fixed dynamical mass Standard test conducted in literature: size ~ (1+z) -x Unlikely evolutionary path For M>10 11 M  x = 0.70 ± 0.11 (40% by z=1) Only massive early-types  are significantly growing in size z > 2 objects appear ultra-compact implying very fast growth??

Size evolution at fixed velocity dispersion More physically meaningful Mergers should increase size but not velocity dispersion Exploits unique dynamical data Tests “progenitor bias” (cut in M dyn restricts in σ, R so could give false evolutionary trend) For σ > 225 km s -1 x = 0.69 ± 0.21 Growth × since z~2 Treu et al DEIMOS Newman et al LRIS-R Cappellari stack van Dokkum z=2.2 Matched SDSS=LRIS Growth Keck  SDSS

Size growth for M  galaxy of × 3.5 ± 0.3 over 0.4 < z < 2.5 But scatter (1σ region) is significant (and valuable information) Growth rate consistent with that found in limited dynamical data and, again, particularly rapid in 2 Gyr period from 1.5<z<2.5 Size Growth Rate in CANDELS data Newman et al (2012) Ap J 746, 162

How Did Early Galaxies Enlarge? Improved observational data (dynamical masses, HST images) confirms size growth is real! What, physically, could lead to this growth in size? Major mergers Minor mergers Mass loss/adiabatic expansion (implausible as inactive systems)

Size Growth During Dissipationless Merging Naab et al 2009 Ap J 699, L108 ( see also Khochfar & Silk 2006 Ap J 648, L21; Khochfar & Silk 2009 MNRAS 397, 506) From virial theorem, total energy Consider merger such that and define Assuming conservation of energy (e.g. parabolic orbits, Binney & Tremaine 2008) Major merger : no change in v, M and R double, d log R / d log M = 1 Lots of minor mergers find d log R / d log M = 2 SPH simulations of minor mergers indicate d log R / d log M ~ 1.3 – 1.6

Measuring the Minor Merger Rate in CANDELS Data WFC3/IR data is sufficiently deep (H<26.5) that we can secure photometric redshifts for secondaries 1/10 th as massive for 404 quiescent primaries with log M P /M  < 10.7 over 0.4 < z < 2 Search area 10 < R < 30 h -1 kpc δz < 0.1 (for z < 1) and δz < 0.2 (for 1 < z < 2) Mass ratio μ = M S /M P > 0.1 Caution: such photo-z associations could still lead to an over-estimate of pairs that will ultimately merge given environs in which red galaxies lie Satellite photo z precision

μ * =0.3 μ * =0.5 μ * =0.2 μ * =0.1 Measuring the Minor Merger Rate in CANDELS Data Find f pair = 0.16 ± 0.03 over full z range Majority of secondaries are also red μ (mass ratio) distribution quite flat as expected from SAMs Newman et al (2012) Ap J 746, 162

Can Minor Merging Explain Size Growth: I? Assuming: 1.Merger timescale τ e ~ 1-2 Gyr (Patton+, Lotz+, Kitzbichler+) 2. Bound fraction of projected pairs C mg (=f 3D )~ Size growth per mass increase dlogR / dlogM ~1.6 (Nipoti+) Size growth over 0.4<z<1 is broadly consistent with that expected from observer minor merger fraction IF merger timescale is fast Size growth over 1<z<2 is inconsistent with observed minor merger fraction for any reasonable choice of parameters

Accounting for New Arrivals Simple model is naïve as it assumes all sources enlarge in lockstep from z~2 progenitors. In reality population comprises old galaxies which formed at z~2 and perhaps expand via mergers AND Newly arrived quiescent systems whose size reflects their epoch of formation Comoving no. density of log M>10.7 quiescents Rapid size growth at high z may be associated with increase in no. density over 1.5<z<2.5

Evolution in Size Distribution Function Key to distinguishing growth of pre-existing sources and the arrival of new sources is the cumulative distribution of mass-normalized radius γ In addition to matching the evolution in mean size growth and number density of quiescent galaxies, a satisfactory model must also account for the rate of depletion of the most compact systems from high redshift to low redshift. Cumulative Distribution Function (CDF) is fit by a skew- normal distribution at various redshifts

A Two Phase Growth Model Consider CDF at z~2 and z~1: Mergers add mass and lead to enlargement. For “intra sample mergers”, the number also declines. Plausible model will shift some fraction P of the most compact z~2 sources to lie within the z~1 CDF with the remainder (1-P) as `new arrivals’ Defines Δlog γ min The test is thus whether the observed rate of minor mergers can deplete this fraction of the most compact sources in the observed CDFs

Can Minor Merging Explain Growth: II ? Conclusion unchanged: 0.4<z<1 size evolution is readily explained by observed rate of minor mergers, but rapid growth over 1<z<2.5 is harder to understand Newman et al (2012) Ap J 746, 162

Summary Present-day massive early type galaxies formed most of their stars by z~2 Evolving stellar mass functions place some limits on the continued appearance of massive early types: most are not genuine `new arrivals’ but represent some combination of dry mergers and truncated star formation in massive blue galaxies The compact nature of early types at z~2.5 is confirmed by CANDELS data; we observe a × 3.5 growth in mean size over M . Dynamical data has been key in verifying the relevant masses, at least to z~1.6; the `red nugget puzzle’ is unlikely to be due to observational errors/mis-interpretations Minor mergers are so far the only plausible mechanism for the size growth. Modeling suggests the observed merger rate can explain the growth observed since z~1 but explaining the rapid growth observed over 1.5<z<2.5 remains a challenge

Growth Rate in HST CANDELS WFC3/IR data Mass-selected sample of 935 massive (> M  ) sources over 311 arcmin 2 in UDS/GOODS-S with 0.4<z<2.5 to gauge growth rate. 90% complete to log M/M  = 9.7 to z=2 so can also search for minor mergers with 10:1 mass ratio around hosts with log M/M  > 10.7 λ (μm) Newman et al (2012) Ap J 746, 162

Adiabatic Expansion Through Mass Loss? Consider a galaxy that expels a significant fraction of its gas e.g. via AGN or SN driven galactic winds Stars and DM will expand in response to shallower central potential Simple homology criterion in spherical symmetric case: Differences from classical work on star clusters (e.g. Tutukov 1978) include role of DM halo and timescales Ragone-Figueroa & Granato astro-ph/ Loss of significant baryonic mass can induce size increase but simulations show this `puffing up’ occurs only when the stellar population are much younger (<0.5 Gyr) than for any of the early type galaxies under consideration.

Cosmology with the Subaru Prime Focus Spectrograph

HyperSuprimeCam – being commissioned now!

Subaru Prime Focus Spectrograph (PFS) Team

Prime Focus Unit includes Wide Field Corrector (WFC) and Fiber Positioner. Spectrograph room located above Naysmith platform Fiber connector mounted on top end structure Fiber Cable routed around elevation axis and brings light to the Spectrographs PFS Concept is Caltech/JPL WFMOS design Design study based on Subaru-provided details, several visits to Subaru, interaction with HSC team Key requirements: survey speed and positioner reliability

Rotator Interface Ring Positioner Equipment Bench Cobra Optic Bench Alignment System Cobra Modules with Drive Electronics Cobra Fiber Positioners Rotating Portion of PFI in PFS Configuration Fibers are routed between fixed and rotating parts of PFI through strain relief box so that 500 degree rotator range is accommodated by twisting along rotator axis rather than bending around a spool

PFS Positioner Optical Bench with Positioner Units Positioner Unit - Cobra Cobra system tested at JPL in partnership with New Scale Technologies Designed to achieve 5μm accuracy in < 8 iterations (40 sec) 2400 positioners 8mm apart in hexagonal pattern to enable field tiling A&G Fiber Guides

Positioner Element – “Cobra” 7.7mm diameter, theta-phi system positioning within 9.5mm patrol area to 5μm precision. Optical fibers mounted in “fiber arm” which attaches to upper postioner axis: Fiber runs through the center of the positioner First axis of rotation Second axis of rotation Patrol Region Top View Fiber Tip Theta stage Phi stage Fiber arm

Triple-arm Spectrograph (one of 4) Unique feature of Jim Gunn design is continuous coverage from 370nm to 1.3μm with matched spectral resolutions (R~ ) for effective simultaneous exposures Detectors: Hamamatsu red- sensivity CCDs and expected Teledyne HgCdTe 4RG arrays

PFS Cosmology: Key Science Goals (Takada, Hirata, Kneib) Measure the cosmological distances, D A (z) and H(z), to 3% accuracy in each of 6 redshift bins over 0.8<z<2.4 via the BAO experiment Use the BAO distance constraints to reconstruct the dark energy densities in each redshift bin Use the BAO distance constraints to constrain the curvature parameter Ω K to 0.3% accuracy Measure the redshift-space distortion (RSD) effect to reconstruct the growth rate in each redshift bin to 6% accuracy

PFS Cosmology Survey The total volume: ~9 (Gpc/h)3 ~ 2 × BOSS survey Assumed galaxy bias (poorly known): b= z PFS survey will have n g P(k)~a in each of 6 redshift bins RedshiftV survey (h -3 Gpc 3 ) # of galaxies (per FoV) n g (10 -4 h 3 Mpc - 3 ) biasn g <z< <z< <z< <z< <z< <z< Assume 100 clear nights to meet the scientific goals → the area of PFS survey

Expected BAO constraints The PFS cosmology survey enables a 3% accuracy of measuring D A (z) and H(z) in each of 6 redshift bins, over 0.8<z<2.4 This accuracy is comparable with BOSS, but extending to higher redshift range Also very efficient given competitive situation –BOSS (2.5m): 5 yrs –PFS (8.2m): 100 nights BOSS PFS-RedPFS-NIR

DE constraints The BAO geometrical constraints achieved significantly constrain the DE parameters For a (w 0, w a )-DE model, H(z) and D A (z) are given as Survey Ω de w const wawa FoM ΩKΩK Planck+BOSS Planck+BOSS+P FS PFS can significantly improve the DE constraints over BOSS PFS can achieve a 0.3% accuracy of the curvature → a fundamental discovery A wider redshift coverage as well as more independent z-bins are powerful aspects in improving the overall constraints (which the CMB alone cannot)

DE reconstruction The wide-z coverage of PFS+SDSS enables a reconstruction of DE densities as a function of redshift → can constrain a broader range of DE models

DE reconstruction The wide-z coverage of PFS+BOSS enables a reconstruction of DE densities as a function of redshift → can constrain a broader range of DE models PFS can significantly improve the accuracy of the reconstruction due to the increased z- bins 7% accuracy of Ω de (z) in each of z- bins PFS+SDSS+Planck allows a detection of dark energy up to z~2, for a Λ-type model

But is PFS competitive with BigBOSS? Survey parameters taken from BigBOSS proposal 500 vs. 100 nights 14,000 vs. 1,420 deg 2 Naively, a factor (10) 1/2 difference BigBOSS better than PFS at z<1.2 (but note target selection a big issue for BigBOSS) PFS comparable with BigBOSS for 1.2<z<1.6 but PFS can uniquely target1.6<z<2.4 PFS more likely to complete its survey before Euclid BigBOSS (4m) vs. PFS (8.2m) BigBOSS

Redshift-Space Distortion (RSD) The redshift-space distortion can probe the growth rate It gives a complementary probe of DE and can test gravity on cosmological scales The PFS can reconstruct the growth rate in each of redshift bins, to a 7% accuracy Peacock et al (2dF team, 01)