Project ALCATRAZ: Areal Lyman-Continuum andTheoretical Reionization Analysis vs. z: Last Chance toObserve First Escape at z~4

Abstract Project ALCATRAZ will study escaping Lyman continuum (LyC; 780-831 A) from galaxies and AGN in theproto-cluster TN J1338-1942 at z=4.10+/-0.006, the highest redshift where it will still be visible (the IGMbecomes opaque at z~4.8). In two pointings of 25 orbits, WFC3 F410M can uniquely detect (>5sigma) such LyC emission when stacking>20 z=4.10 objects, reaching equivalent depths of >500 orbits with careful attention to systematics. HSTresolution with multi-filter (shallow) SED-fits are essential for rejection of foreground interlopers, and toestimate stellar mass, age, SFR, and extinction, needed to derive robust escape fractions. The new z=4.10 datawill constrain where, when, how and how much LyC escapes, tracing how galaxies may have started and AGNmaintained reionization. We aim to answer: 1. HOW MUCH LyC escapes? Stacking WFC3 images of 11-37 galaxies and weak AGN in four filters atz=2.3-5 yielded m(LyC)~29-30.5 mag (>3-4 sigma). For >20 objects at z=4.10, we can measure escapefractions of 30-80% for Lya galaxies, LBGs and AGN. 2. HOW does LyC escape? HST stacking will measure LyC light-profiles at r

Investigators:

Target Summary:

Observing Summary:

Investigator Institution CountryPI& R Windhorst Arizona State University USA/AZ

CoI B Smith Arizona State University USA/AZ

CoI R Jansen Arizona State University USA/AZ

CoI S Cohen Arizona State University USA/AZ

CoI* M Dijkstra University of Oslo NOR

CoI A Inoue Osaka Sangyo University, College of GeneralEducation,

JPN

CoI* R Bielby University of Durham GBR

CoI* C Conselice University of Nottingham GBR

CoI A Koekemoer Space Telescope Science Institute USA/MD

CoI J MacKenty Space Telescope Science Institute USA/MD

CoI R Overzier Observatorio Nacional BRA

CoI* H Rottgering Sterrewacht Leiden NLD

CoI* B Venemans Max-Planck-Institut fur Astronomie, Heidelberg DEU

Number of investigators: 13* ESA investigators: 5& Phase I contacts: 1

Target RA Dec MagnitudeTN-J1338-1942-UVFLD1 13 28 30.4600 -19 43 0.80 V = 23.05 +/- 0.05

TN-J1338-1942-UVFLD2 13 28 23.8600 -19 45 20.00 V = 24.5 +/- 0.10

TN-J1338-1942-IRFLD1 13 28 31.9400 -19 43 2.60 V = 24.5 +/- 0.10

TN-J1338-1942-IRFLD2 13 28 26.0600 -19 43 48.80 V = 24.5 +/- 0.10

TN-J1338-1942-IRFLD3 13 28 23.3400 -19 45 39.80 V = 24.5 +/- 0.10

Target Config Mode and Spectral Elements Flags OrbitsTN-J1338-1942-UVFLD1 WFC3/UVIS Imaging F410M 25

TN-J1338-1942-IRFLD1 WFC3/IR Imaging F098M 1

TN-J1338-1942-IRFLD1 WFC3/IR Imaging F125W 1

TN-J1338-1942-IRFLD1 WFC3/IR Imaging F160W 1

R Windhorst : Project ALCATRAZ: Areal Lyman-Continuum and Theoretical ReionizationAnalysis vs. z: Last Chance to Observe First Escape at z~4

Target Config Mode and Spectral Elements Flags OrbitsTN-J1338-1942-UVFLD2 WFC3/UVIS Imaging F410M 25

TN-J1338-1942-UVFLD2 ACS/WFC Imaging F475W 1

TN-J1338-1942-UVFLD2 ACS/WFC Imaging F625W 1

TN-J1338-1942-UVFLD2 ACS/WFC Imaging F814W 1

TN-J1338-1942-IRFLD2 WFC3/IR Imaging F098M 1

TN-J1338-1942-IRFLD2 WFC3/IR Imaging F125W 1

TN-J1338-1942-IRFLD2 WFC3/IR Imaging F160W 1

TN-J1338-1942-IRFLD3 WFC3/IR Imaging F098M 1

TN-J1338-1942-IRFLD3 WFC3/IR Imaging F125W 1

TN-J1338-1942-IRFLD3 WFC3/IR Imaging F160W 1

Total prime orbits: 62

R Windhorst : Project ALCATRAZ: Areal Lyman-Continuum and Theoretical ReionizationAnalysis vs. z: Last Chance to Observe First Escape at z~4

Scientific Justification

A. Scientific Background: Cosmic Reionization was the second major phase transition ofhydro-gen in the universe, following recombination. It began at z∼2.5 (Madau

+ 1999; Fan+

2001; Hunt+ 2004; Siana+ 2007), they likely did not reionize the IGM at z∼>3. However, AGNcontributed much of the LyC background from their peak epochat z∼2 until today, maintainingthe current ionization of the IGM (Cowie+ 2009). No significant escaping LyC flux was detectedin rest-frame far-UV data of SF galaxies at 0.5∼4.35 (Fig.1a). S15 showed that, by carefully correcting for systematics (§D), stacking images with a native2-orbit depth for 133 objects can reach sensitivities equivalent to 100s of UV orbits. The higherIGM opacity at z∼>4.5 makes it more difficult to detect LyC in visible images, although IGMtransmission at z≃4.10 could still be∼20% (Inoue+ 2014; Fig. 1b here). Therefore, Project AL-CATRAZ will observe the best available cluster at z≃4.10, for which LyC can still be seen throughthe IGM transmission. Well studied with VLT and ACS, TN J1338–1942 has∼>38 known redshiftsat z≃4.099±0.006 (de Breuck+ 2004; Miley+ 2004; Venemans+ 2007; Overzier+ 2008).ALCA-TRAZ will stack 25-orbit images for ∼>20 objects at z≃4.10, reaching a∼>500-orbit depth atHST resolution, essential to measure LyC andfesc for Ly α galaxies, LBGs, and AGN.

1

Fig. 1a: Composite rest-frame spectra of SDSS QSOs atz≃1.3 (van den Berk+ 2001 [blue]) and of LBGs atz≃2–4 (Bielby+ 2013 [greenand

orange]; Shapley+ 2003 [red]). The F225W, F275W, F336W and F435W (not shown) transmission curves capture LyC (λ3–4σ) in images stacked over 11–37 galaxies, and for someweak AGN (Fig. 3). Cen & Kimm (2015) showed thatfesc will converge to more reliable valueswhen averaged over several dozen, compared to N∼20 objects (Fig. 6a). HST resolution multi-band images, photo-z’s and SED-fits (Overzier

+

2008; Fig. 4b &§D here) will mask-out all foreground interlopers (Fig. 3), which ground-basedimages cannot do.ALCATRAZ provides critical new LyC detections at ∼>5σ, using rigorousmapping and removal of low-level systematics, yielding robust fesc at z≃4.1.

Goal 2. HOW?: What is the radial dependence of escaping LyC radiation?Fig. 3 shows 51 (118)galaxies in the spectroscopic “Gold” (“Gold+Silver”) samples of S15 (§D.2a). Their weightedaverage LyC emission isnot centrally concentrated, perhaps vaguely resembling a “ring”. Fig. 4ashows that the stacked LyC profiles are extended with respectto the PSFs (S15),andmuch flatterthan non-ionizing UV-continuum (UVC; 1500Å) profiles for r∼

Fig. 2a: Absolute and apparent WFC3/IR F125W magnitude distributions of the Gold and Gold+Silver samples with spectroscopic redshifts of

S15, sampling rest-frame near-UV at 2.26∼< z∼

< 5. The blue dotted curve indicates the faint-end power-law slope of 0.16 dex/mag of the galaxy

counts of W11, suggesting sample incompleteness for AB∼> 24 mag. [2b] Same, for galaxies hosting aweakAGN, which have〈MAB〉≃–21.5±1

mag. [2c] Same, for galaxies without AGN.ALCATRAZ will add new VLT spectra for objects at z ≃4.10 selected by multi-band photo-z’s.

Goal 3. WHEN?: How exactly does the LyC escape fraction evolve with redshiftfor galaxiesand weak AGN? Doesfesc follow the cosmic SF and SMBH-formation history?Estimating theescape fraction of LyC photons from galaxies,fesc, is non-trivial as it requires modeling intrinsicLyC flux, f intLyC, and the wavelength-dependent fraction of LyC photons transmitted through theIGM, T LyCIGM(λ, z), for a galaxy at redshiftz. In Fig. 5a, we summarize publishedfesc-values asblue triangles. These came from different data sets (including both spectra and imaging; see§A),with different object selection, reduction techniques, assessment of systematics, and treatment ofthe estimated corrections for IGM absorption. With SED-fitting of the UVC longwards of Lya,the observed LyC fluxes of S15 correspond toaveragerelative LyC escape fractions that seemto rapidly increase fromfesc≃7% at〈z〉≃2.37–2.67 to∼50±30% at〈z〉≃3.5, and to∼100% at〈z〉≃5.1, where we do see some LyC flux at the∼>3σ level, although its interpretation is muchharder due to the considerable IGM opacity-correction at z∼>5. Fig. 5ab plot thesefesc-valuesfrom Table 1, and from MC simulations of our ERS data through IGM transmission models (Fig.1b). Several authors (Inoue+ 2006; Kuhlen & Faucher-Gigùere 2012; Finlator+ 2012; Becker &Bolton 2013; Dijkstra+ 2014) have suggested thatfesc may increase significantly with redshift,possibly as steeply as∝(1+z)3. The combined data in Fig. 5a suggest a trend in thefesc-valuesof galaxies with redshift that may not be a simple power-law in (1+z). The violet-shaded regionis bounded by:fesc ≃ (0.02 ± 0.01) × (1 + z)1.0±0.5, which assumesno dependence offesc onMAB. About 2/3 of the 21 independent data points in Fig. 5a deviates bymore than1σ from eitherone of the three power-laws, so thatno single (1+z)-regression fits all thefesc-data for galaxies.Therefore, we suggest thata more sudden increase offesc with redshiftmay instead have occurred:log(fesc) ≃ log(fesc,0) + F0 · tanh [(log((1 + z) − log(1 + z0))/Z0)].

Each of the four free parameters in this tanh-fit has a straightforward meaning. For galaxies,the steepest drop infesc occurs atz0 ≃2.9 — i.e., right around the peak in the cosmic star-formationhistory — over an interval less than±1 Gyr in cosmic time [i.e.,Z0≡∆log(1 + z)≃0.07]. Thepivot point for galaxies atz0 ≃2.9 occurs atfesc,0≃11%, andfesc may have dropped by a factor of∼30 [i.e.,F0≡∆logf≃(log 30)/2] from∼60% at z≃5 to∼2% at z≃0–1. For “Galaxies with weakAGN” these parameters are very poorly constrained, butif we assume that their UVC SEDs are stilldominated by their stellar population [〈MAB〉≃–21.5±1 mag in Fig. 2b], and that thefesc-valuesof weak AGN dropped similarly fast over±1 Gyr in cosmic time — but aroundz0 ≃2.4 duringthe peak in the QSO epoch — then theirfesc-values may have dropped fromfesc∼90% at z≃5 tofesc≃15% at z≃0–1, i.e., only by factor a of∼6 over most of cosmic time.Multiplying by theirnumbers in Table 1, galaxies then dominate the reionizing flux at z∼>3, while AGN take overat z∼

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Isophotal Semi-Major Axis (arcsec)

22

24

26

28

30

32

Surfa

ce Brig

htne

ss (m

agAB arcse

c−2)

F435W PSFF275W PSF

Model (z=2.65)

F225W (=2.38)

F275W (=2.68)

F336W (=3.47)

F435W (=5.02)

1 2 3 4 5 6 7 8 92 10

0

1

2

3

Fig. 4a: Radial surface brightness (SB) profiles of the non-ionizing UVC signal (solid curves) and of the LyC signal (dashed curves) detected in

the stacks for the combined spectroscopic Gold sample of S15, color-coded according to their mean redshift. The horizontaldashed line indicates

the 1σ SB-limit for stacked LyC profiles. All four UVC SB-profiles areextended with respect to the PSFs (dotted purple and cyan curves). The LyC

SB-profiles (dashed) are also clearly extended,and flatterthan the UVC profiles (solid), as predicted by our scattering model (light-blue dot-dashed

curves for z≃2.68). HST resolution is essential to measure LyC light-profiles atz≃4.10. Fig. 4b: Distribution of (discretized) dust extinction

AV -values from best-fit SEDs for all 6900 galaxies in the 10-band ERS data (W11; small black dots), compared to the LyC samples used for the

stacks in the four indicated redshift bins (colored open circles and asterisks for galaxies and AGN, resp.).ALCATRAZ will address this with

∼>20 new LyC objects at z≃4.10, providing the high-z anchor to see if dust accumulating over cosmic time has shut downfesc(z).

Other work (see text)

Other work (see text)

Fig. 5a: Relativeescape fractions for various galaxy samples vs. redshift. Orange filled circles indicate WFC3+ACS ERS data for the spectroscopic

Gold galaxy, and grey filled circles for the Gold+Silver galaxy samples of S15. Orange triangles connected by thin lines indicate the modal–median–

averagefesc-values plus their±1σ-range from MC simulations using Inoue+ (2014)’s code. Blue triangles are published data (§A), some of

which have different sample selection. The violet-shaded region is bounded by:fesc≃(0.02±0.01)×(1+z)1.0±0.5. The combined galaxy samples

suggests a trend offesc with redshift, although it may not be a simple power-law in (1+z). Hence, we also plot a simple tanh[log(1+z)] relation

(green curve), which captures the possibly rather sudden change infesc around z∼>3 fromfesc≃2% at z≃0–1 tofesc∼>60% at z∼>5, implying that

galaxies may have had high enoughfesc-values, and so produced sufficient LyC flux, to complete reionization byz≃6. Fig. 5b: Same as Fig. 5a,

but for weak AGN in the S15 sample, plus one from Bridge+ (2010).fesc-values were calculated under theassumptionthat the UVC-SED of weak

AGN is dominated by star-light and not QSO light. For (weak) AGN, fesc-values may have increased fromfesc≃15% at z≃0–1 tofesc≃90% at

z∼>5 (purple curve). The object-weighted ratio of these tanh-curves suggests that galaxies dominated reionization for z∼> 3, while AGN took over

at z∼< 3. ALCATRAZ will provide critical new LyC data at for

∼>20 new Lyα galaxies, LBGs & AGN to measurefesc(z) at z≃4.10.

5

RAJ

Fig. 6a: VLT spectra for 37 Lyα emitters in the cluster surrounding radio galaxy TN J1338–1942 (Venemans+ 2007), whose broad-line spectrum

is in the upper right. Some are weak broad-line AGN.Fig. 6b: Distribution of galaxies around the radio galaxy (ellipse). The color scale marks

velocities away from the cluster〈z〉≃4.099±0.006 (rms). Asterisks mark LBGs at z≃4.1 (Overzier+ 2008). This is the best available cluster at

the highest known redshift where LyC can still be measured through the IGM using careful image stacking. ALCATRAZ will observe it,

since it is our last chance (i.e., highest redshift) to see “first escaping” LyC radiation (i.e., LyC still visible throug h the high IGM opacity).

Goal 4b. WHERE?: We will study LyC escaping from galaxies near AGN with strong outflows.VLT/MUSE provided spatially resolved Lya spectra for radiogalaxy TN J1338–1942 (Swinbank+

2015), which has L1500∼6L∗ at z≃4.10 (Overzier+ 2008). It hasstrongly blue-shifted Lyα (at

∼>–2000 km/s; Fig. 6a), indicating a strong outflow in the center of its extended Lya-halo. This is agood tracer of escaping LyC (Heckman+ 2011, Borthakur+ 2014), since outflows may clear holesand reduce the HI-covering factor (§C.3), producing escape paths for LyC (and Ly-α) photons.Outflowing gas may be driven by the radio jets, or starburst-driven (Zirm+ 2005). Hence, weexpect a highfesc-value for the radio galaxy, and possibly spatially resolveits escaping LyC (e.g.,Fig. 3, 4a). There is strong evolution in the Lyα LF between z≃6.6 and z≃5.7, although thereis no evolution at the LF bright-end, which may deviate from aSchechter function. This may bea sign of reionization not being completed yet: LAEs are onlyobserved if they are in an ionizedsphere large enough for Lyα photons to redshift out of resonance. We will trace LyC separatelyfor LAEs closest in redshift (red dots) to the radio galaxy atz≃4.10, and compare to the ensemble-average LyC flux.ALCATRAZ will study if faint Ly- α emitters near bright AGN or nearmore luminous LAEs have higher LyC flux. It will do so where we can, while we can.

Abbreviated References:Avila R+ ACS Handbook (STScI)Baggett S+ 2006 SPIE 6265 626532Becker G+ 2013 MNRAS 436 1023Becker R+ 2001 AJ 122 2850Bertin Arnouts 1996 A&ApS 117 393Bielby R+ 2013 MNRAS 430 425Borthakur S+ 2014 Science 346 216Boutsia K+ 2011 ApJ 736 41Bouwens R+ 2012 ApJ 752 L5Bridge C+ 2010 ApJ 720 465Bruzual+ 2003 MN 344 1000Calzetti D+ 2000 ApJ 533 682Cen & Kimm 2015 ApJL 801 L25Coe D+ 2006 AJ 132 926Cohen S+ 2006 ApJ 639 731Cooke J+ 2014 MNRAS 837 51Cowie L. L+ 2009 ApJ 692 1476Dahlen T+ 2013 ApJ 775 93de Breuck C+ 2004 AA 424 1Dijkstra Kramer 2012 MN 424 1672Dijkstra M+ 2014 MN 440 3309

Dressel L. WFC3 Handbook (STScI)Fan X+ 2001 AJ 122 2833Fan X+ 2002 AJ 123 1247Finlator+ 2012 MNRAS 427 2464Giallongo E+ A. 2002 ApJ 568 L9Giavalisco M+ 2004 ApJ 600 L93Gnedin N. 2008 ApJ 673 L1Grazian A+ 2011 A&A 532 A33Haardt & Madau 1996 ApJ 461 20Haardt & Madau 2012 ApJ 746 215Hathi N+ 2008 AJ 135 156Hopkins P+ 2006 ApJS 163 1Hunt M. P+ 2004 ApJ 605 625Inoue A. K+ 2006 MNRAS 371 L1Inoue A. K+ 2014 MNRAS 442 1805Ishigaki M+ 2015 ApJ 799 12Iwata I+ 2009 ApJ 692 1287Koekemoer A+ 2013 ApJS 209 3Kozhurina V+ WFC3 ISR 2013-14Kuhlen Faucher 2012 MN 423 862Le Fevre O+ 2004 A&A 428 1043Leitherer C+ 1999 ApJS 123 3

Mack J+ WFC3 ISR 2013-10Madau P. 1995 ApJ 441 18Madau P+ 1999 ApJ 514 648Malkan M+ 2003 ApJ 598 878Mesinger Haiman 2004 ApJ 611 L69Miley G+ 2004 Nature 427 47Mostardi R. E+ 2013 ApJ 779 65Nestor D+ 2011 ApJ 736 18Nestor D+ 2013 ApJ 765 47Oesch P+ 2013 ApJ 773 75Overzier R+ 2008 ApJ 673 143Planck Collab. arXiv:1502.01589Prochaska J+ 2009 ApJ 705 L113Rauch M+ 2011 MN 418 1115Ricotti M. 2002 MNRAS 336 L33Robertson B+ 2013 ApJ 768 71Sabbi E+ WFC3 ISR 2009-19Saito T+ 2015MNRAS434 3069Shapley A+ 2003 ApJ 588 65Shapley A+ 2006 ApJ 651 688Siana B+ 2007 ApJ 668 62Siana B+ 2009 ApJ 698 1273

Siana B+ 2010 ApJ 723 241Siana B+ 2015 arXiv:1502.06978Smail I+ 2013 MNRAS434 3246Smith B+ astro-ph/1504. (S15)Springel V+ 2005 ApJ 620 79Stark D+ 2010 Mnras 408 1628Steidel C+ 2001 ApJ 546 665Swinbank A+ 2015 MN 449 1298Vanden Berk D. E+ 2001 AJ 122 549Vanzella E+ 2008 A&A 478 83Vanzella E+ 2010 MNRAS 404 1672Vanzella E+ 2012 ApJ 751 70Vanzella E+ 2015 arXiv:1502.04708Venemans B+ 2005 AA 431 793Venemans B+ 2007 AA 461 823Windhorst R+ 1994 PASP 106 798Windhorst R+ 1998 ApJ 494 L27Windhorst Cohen 2010 AIP 1291 225Windhorst+ 2011 ApJS 193 27Wolf C+ 2004 A&A 421 913Worseck G+ 2014MN 445 1745Zirm A+ 2005 ApJ 630 68

6

Table 1. Summary of LyC Stacking thus far, and Proposed New LyC-Data at z≃4.10Filter z-range 〈z〉 Nobj mLyC SNLyC mUV C SNUV C f1500/fLyC AV med < TIGM > frelesc,700 f

relesc(MC)

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)

ERS GALAXIES WITH AGN (SMITH+ 2015):F225W 2.291–2.291 2.291 1 ∼

> 30.12 ∼< 2.34 27.90 7.85 3.44+0.13

−0.10 0.90+0.14−0.14 0.297

+0.081−0.083 ∼>100% —

F275W 2.470–3.008 2.697 7 28.92 8.77 25.00 156.9 2.98+0.08−0.07 1.23

+1.14−1.13 0.247

+0.085−0.085 33

+24−22% —

F336W 3.217–3.474 3.349 3 29.69 3.58 24.45 118.2 11.4+0.20−0.14 0.10

+0.14−0.10 0.112

+0.049−0.049 82

+50−50% —

F435W 4.760–4.823 4.792 2 28.58 4.48 24.66 79.0 3.55+0.37−0.26 1.90

+0.50−0.50 0.0011

+0.0012−0.0010 ∼100% —

ERS GOLD GALAXIES WITHOUT AGN (SMITH+ 2015):F225W 2.302–2.450 2.380 14 29.98 5.64 24.43 237.5 3.44+0.13

−0.10 0.55+0.70−0.44 0.297

+0.081−0.083 7.0

+5.3−5.3% 0.76

+15−0.4%

F275W 2.559–3.076 2.682 11 30.09 5.71 24.51 192.2 2.98+0.08−0.07 0.58

+0.89−0.40 0.247

+0.085−0.085 7.1

+6.0−6.0% 3.22

+35−1.1%

F336W 3.132–3.917 3.472 11 30.66 4.48 24.88 101.9 11.4+0.20−0.14 0.18

+0.64−0.12 0.112

+0.049−0.049 50

+31−31% 34

+63−16%

F435W 4.414–5.786 5.015 15 30.37 3.28 26.12 70.3 3.55+0.37−0.26 0.17

+0.67−0.12 0.0011

+0.0012−0.0010 ∼100% ∼100%

ERS GOLD + SILVER GALAXIES WITHOUT AGN (SMITH+ 2015):F225W 2.262–2.450 2.362 31 29.79 9.46 24.56 303.6 3.74+0.12

−0.10 0.55+0.70−0.44 0.306

+0.055−0.055 9.9

+7.0−7.0% 1.76

+15−0.7%

F275W 2.481–3.076 2.692 26 29.46 11.9 24.76 229.6 3.25+0.06−0.06 0.58

+0.89−0.40 0.249

+0.052−0.054 17

+9.7−11 % 6.2

+27−2.1%

F336W 3.110–4.149 3.524 24 29.96 6.85 24.73 164.9 4.33+0.34−0.30 0.18

+0.64−0.12 0.089

+0.027−0.027 39

+18−21% 6.5

+25−3.1%

F435W 4.414–6.277 5.312 37 30.35 5.79 26.72 92.7 2.97+0.13−0.15 0.17

+0.67−0.12 0.0002

+0.0015−0.0015 ∼100% 87

+113−55 %

PROPOSED TN J1338–1942 Radio Galaxy + Weak AGN at z≃4.10 (estimated):F410M 4.08–4.11 4.10 ∼

> 3 ∼29.1 ∼> 5 ∼24.7 ∼

> 100 4.33+0.34−0.34 0.2

+0.6−0.1 0.089

+0.027−0.027 60

+20−20%? 60

+25−3 %?

PROPOSED TN J1338–1942 Proto-Cluster Lyα Galaxies + LBGs at z≃4.10 (estimated):F410M 4.08–4.11 4.10 ∼

> 20 ∼30.3 ∼> 5 ∼24.7 ∼

> 250 4.33+0.34−0.34 0.2

+0.6−0.1 0.089

+0.027−0.027 50

+20−20%? 50

+13−1.5%?

TOTAL New objects: 4.10 ∼> 23

Table columns:(1) WFC3 or ACS filter; (2) Redshift range used in LyC/UVC stacks; Each lower redshift bound samplesno light atλ>912Å below the filter’s red edge(defined at 0.5% of the filter’s peak transmission; see S15). (3) Average redshift of stack; (4) Number of galaxies with reliable spectroscopic redshifts used in stack; (5)Observed total AB magnitude in LyC stack (using SEXTRACTORMAG AUTO); (6) S/N ratio of LyC stack; (7) Observed total AB magnitude in UVC stack; (8) S/Nratio of UVC stack; (9) Averageintrinsic model flux ratiof1500/fLyC and its±1σ error; (10) Median dust extinction AV and its±1σ error from the 10-band SEDBC03-model fits; (11) Average IGM transmission〈TIGM〉 and its±1σ range, derived from the Inoue+ (2014) models for the sample of actual redshifts; (12) Inferredrelative escape fraction of LyC photons in percent and its±1σ range, assuming a constantintrinsic LyC flux density overλ=700–900̊A. (13) Modal relativefesc-valuesderived from our Monte Carlo test of the IGM transmission using the code of Inoue+ (2014) plus their±1σ ranges, usingfabsesc ≡ fLyC,obs/fLyC,⋆. These ERSUV-stacks are equivalent in depth to 22–236 HST orbits.ALCATRAZ adds ∼

> 20 new galaxies and∼> 3 AGN at z≃4.10, reaching∼

> 500 orbits stacking depth. All theproposed numbers of new objects are indicated in violet.For the new z≃4.10 cluster data, the numbers in black are the expected values from ALCATRAZ (§D).

D. Description of the Observations(0a) Existing Data: The new WFC3 F410M data will be combined with existing 2-orbit ACS/WFCimages in F475W, F625W, F775, F850LP, and the G800L grism (9291, PI Ford), so they do notneed to be taken for the main eastern field (Fig. 6b), which contains the radio galaxy.

(0b) New WFC3 and ACS Data Required:The WFC3 ETC yields the following, consistent withour analysis of actual WFC3 data (W11, S15). First, 25 drizzled full-orbit F410M exposures willyield very reliable CR-removal (Windhorst+ 1994). In Fall 2015–Spring 2016 (6.5–7 years afterWFC3 launch), the WFC3 UVIS dark current will be∼>3.5e

− /hour. The full-orbit sky in F410Mdoes not quit beat the UVIS dark current nor its read-noise, but helps elevate the total sky-signallevel. Hence, the CCD post-flash level can be kept at∼12e− required for reliable

filling of all charge traps, so thatno faint LyC signal gets lost upon readout(see S15 for details ofCTE mitigation, and 1b below). We use the actual ETC Zodi-background for this field, and takea single∼5. The stack of 41 objects at z≃3.45–5 (Table 1 &Fig. 3) shows that their LyC flux ison averageAB≃30.3±0.3 mag. For Lya galaxies and LBGs atz≃4.10 it is likelyat least this bright, sincefesc likely rises fast with redshift (Fig. 5a).The ETCshows that 25 orbits in F410M will yield SNR≃1.2–1.3on average per object above skyover thesmall area where LyC is seenin individual galaxies.The LyC stacking method of S15 (Table 1,

7

Fig. 3–4a, 5 here) has clearly shown that WFC3 and ACS images can be stacked to well over 200orbits depth (see also Hathi+ 2008, who stacked ACS BViz images to∼>5000-orbit depth to getUV-continuum light-profiles for HUDF galaxies at z≃4–6).Hence, stacking a minimum of∼>20galaxies at z≃4.10 in 25-orbits F410M will yield for galaxies a stacked LyCflux above skywith SNR≃ 1.25 ·

√20 ∼> 5 (see further Fig. 3 and Table 1 for feasibility, and S15 for details).

Using the same UVIS parameters as above, in 500 full-orbits of 2700 sec the WFC3 ETC doesindeed yield SNR∼>5.1 for AB≃30.3 mag, using small detection apertures as in Table 1.

The second WFC3 F410M field also needs 1-orbit ACS exposures in F475W, F625W, F814W,since it is not covered by existing ACS images (Fig. 6b). The ACSETC shows these will reachAB∼>27.0 (SNR∼>9σ) for compact objects in 3×900 sec exposures. ForbothWFC3 F410M fields,we also needthree WFC3/IR fieldswith 1-orbit exposures in F098M, F125W, and F160W (Fig.6b), also reaching AB∼>27 mag (SNR∼>5σ), as shown by the WFC3 ETC and W11. This is neededto measure thestacked UV-continuumat SNR∼>250 (Table 1), so it yields reliable SED-fittingand Monte Carlo modeling offesc. (Using F814W and F098W prevents the need for the moreinefficient F850LP). Together, the combination of the shallow ACS+WFC3 grI+YJH filters willcover both deep WFC3 F410M fields in restframe UVC continuum filters. For reliable LyCstacking, it is absolutely essential to obtain all images atHST resolution to: (1) remove allforeground interlopers to AB∼5 pixels (W11), which have been addressed with new geometrical distortioncorrections (GDCs; Kozhurina-Platais 2013) to within〈∆(X,Y )〉∼

of ∼32.3 mag arcsec−2 across our 6.′′39×6.′′39 LyC boxes (Fig. 3). We will map any low-levelresidual gradients across all images as a 2D-surface, and subtract them from the individual imagesafter super-darkframe subtraction, and if needed again after flat-fielding and before drizzling.

(1d) Red-Leak and Filter-Pinhole Corrections:Fig. 1a and S15 predict the LyC red-leak fractionfor LBGs at z≃2–6 relative to their UVC flux as 0.0030%–0.0001%, in line with the red-leak wingsof these four filters. Redleak fractions are thus very small compared to our relativefesc-values.

(2) Panchromatic SED fitting: We will use panchromatic SED-fitting (Coe+ 2006; Dahlen+

2013) to estimate the following galaxy properties: zphot, total flux, & luminosity, stellar mass,stellar population age, star-formation rate (SFR), and extinction (AV , e.g., Windhorst+ 2010).

(2a) Spectroscopic Redshifts, Sample Selection, Reliability & Completeness: Spectroscopicredshifts (zspec) will be used wherever available, resulting in smaller but more reliable samples.We will inspect the available spectra, and assign a fidelity grade to each redshift, ranging from 1(Gold or “highly reliable”), 1 (Silver or “likely reliable”) to 2 (“indeterminable”).

(2b) Photometric Redshifts and Removal of Interlopers:To get larger and deeper samples,we will use the multi-filter P(zphot) distribution to maximize the probability that each photo-zobject belongs to the LyC candidate to be stacked. All contaminating neighbors visible in thedeepest multi-filterχ2-stacks will be masked out in the LyC apertures and surrounding sky usingSEXTRACTOR (Bertin & Arnouts 1996) “segmentation” maps (see Fig. 3).

(3) LyC Stacks and Quality Checks:We will perform various quality checks (see S15), verifyingthat the LyC flux remains present (within the errors in Table 1) when the images are rotated byrandom multiples of 90◦, or when the first and second independent data-halves are stacked. Wewill also verify that no spurious fluxis seen when stacking an equal number ofrandom emptysky-boxeswithout known objects, verifying the point-source and SB-sensitivity limits in Table 1.

(4a) Theoretical Modeling — Scattering Models:Dijkstra will predict LyC SB-profiles for a gridof wind model parameters, and investigate what constraintscan be placed from either detectionsand/or upper limits. He will perform full MC simulations of asub-set of models to account formultiple scattering events, and to test the accuracy of analytic models.

(4b) Theoretical Modeling — Improved Constraints on IGM Tran smission Models:Inouewill investigate if significant z≃4.10 LyC detections require updates of IGM transmission models,which predicts very lowTIGM values (Table 1). The Inoue+ (2014) model is based on the redshift(dn/dz) and column density distributions from LAF, LLS, andDLAs, and best reproduces Lymanlimit mean-free-path (MFP) measurements at z∼

Past HST Usage

Note that the description of past HST usage DOES NOT count against the 9-page limits.

This Medium-sized program has∼3.5 pages of text (out of 4 allowed) and∼2.5 pages of figures.Together with§D, this is within the 9-page limit.

GO-12500, Cycle 19: PI: S. Kaviraj (Oxford): “ High-res WFC3 UV studies of SAURON galax-ies ” — WFC3 observations started in Fall 2011, data fully 6 papers published. Project took 10%of Windhorst’s time, and finished in fall 2014.

GO-12974, Cycle 20: PI: M. Mechtley (ASU):“WFC3 IR Imaging of z=6 QSO Host Galaxies”— Cycle 20 observations continuing through summer 2013. Datareduced and analyzed. Led toDr. M. Mechtley’s PhD Dissertation; ASU; Jan. 2014, see ASU link below). One paper publishedin ApJL, 2 more papers submitted in Spring 2015. Program took20% of Windhorst’s time, andwill finish in Spring 2015.

AR-13241, Cycle 21: PI: Cohen (ASU):“Pixel-by-pixel Resolved Stellar Populations” — Archivalpixel-by-pixel analysis of of resolved stellar populations in nearby galaxies compared with inter-mediate redshift galaxies (z∼