J. Lange

We present reduced data and data products from the 3D-HST survey, a 248-orbit $HST$ Treasury program. The survey obtained WFC3 G141 grism spectroscopy in four of the five CANDELS fields: AEGIS, COSMOS, GOODS-S, and UDS, along with WFC3 $H_{140}$ imaging, parallel ACS G800L spectroscopy, and parallel $I_{814}$ imaging. In a previous paper, we presented photometric catalogs in these four fields and in GOODS-N, the fifth CANDELS field. Here we describe and present the WFC3 G141 spectroscopic data, again augmented with data from GO-1600 in GOODS-N (PI: B. Weiner). We developed software to automatically and optimally extract interlaced two-dimensional (2D) and one-dimensional (1D) spectra for all objects in the Skelton et al. (2014) photometric catalogs. The 2D spectra and the multi-band photometry were fit simultaneously to determine redshifts and emission line strengths, taking the morphology of the galaxies explicitly into account. The resulting catalog has redshifts and line strengths (where available) for 22,548 unique objects down to ${{JH}}_{\\mathrm{IR}}\\leq 24$ (79,609 unique objects down to ${{JH}}_{\\mathrm{IR}}\\leq 26$). Of these, 5459 galaxies are at $z > 1.5$ and 9621 are at $0.7< z< 1.5$, where Hα falls in the G141 wavelength coverage. The typical redshift error for ${{JH}}_{\\mathrm{IR}}\\leq 24$ galaxies is ${\\sigma }_{z}\\approx 0.003\\times (1+z)$, i.e., one native WFC3 pixel. The $3\\sigma $ limit for emission line fluxes of point sources is $2.1\\times {10}^{-17}$ erg $s^{-1} cm^{-2}$. All 2D and 1D spectra, as well as redshifts, line fluxes, and other derived parameters, are publicly available.

ABSTRACT We present H α maps at 1 kpc spatial resolution for star-forming galaxies at z ∼ 1, made possible by the Wide Field Camera 3 grism on Hubble Space Telescope ( HST ). Employing this capability over all five 3D- HST /CANDELS fields provides a sample of 3200 galaxies enabling a division into subsamples based on stellar mass and star formation rate (SFR). By creating deep stacked H α images, we reach surface brightness limits of 1 × 10 −18 erg s −1 cm −2 arcsec −2 , allowing us to map the distribution of ionized gas to ∼10 kpc for typical L * galaxies at this epoch. We find that the spatial extent of the H α distribution increases with stellar mass as kpc. The H α emission is more extended than the stellar continuum emission, consistent with inside-out assembly of galactic disks. This effect grows stronger with mass as . We map the H α distribution as a function of SFR(IR+UV) and find evidence for “coherent star formation” across the SFR– M * plane: above the main sequence (MS), H α is enhanced at all radii; below the MS, H α is depressed at all radii. This suggests that at all masses the physical processes driving the enhancement or suppression of star formation act throughout the disks of galaxies. At high masses ( ), above the MS, H α is particularly enhanced in the center, potentially building bulges and/or supermassive black holes. Below the MS, a strong central dip in the EW(H α ), as well as the inferred specific SFR, appears. Importantly, though, across the entirety of the SFR– M * plane, the absolute SFR as traced by H α is always centrally peaked, even in galaxies below the MS.

ABSTRACT We introduce a novel approach to boost the efficiency of the importance nested sampling (INS) technique for Bayesian posterior and evidence estimation using deep learning. Unlike rejection-based sampling methods such as vanilla nested sampling (NS) or Markov chain Monte Carlo (MCMC) algorithms, importance sampling techniques can use all likelihood evaluations for posterior and evidence estimation. However, for efficient importance sampling, one needs proposal distributions that closely mimic the posterior distributions. We show how to combine INS with deep learning via neural network regression to accomplish this task. We also introduce nautilus, a reference open-source python implementation of this technique for Bayesian posterior and evidence estimation. We compare nautilus against popular NS and MCMC packages, including emcee, dynesty, ultranest, and pocomc, on a variety of challenging synthetic problems and real-world applications in exoplanet detection, galaxy SED fitting and cosmology. In all applications, the sampling efficiency of nautilus is substantially higher than that of all other samplers, often by more than an order of magnitude. Simultaneously, nautilus delivers highly accurate results and needs fewer likelihood evaluations than all other samplers tested. We also show that nautilus has good scaling with the dimensionality of the likelihood and is easily parallelizable to many CPUs.

ABSTRACT We constrain the newly introduced decorated halo occupation distribution (HOD) model using SDSS DR7 measurements of projected galaxy clustering, $w$p(rp) of galaxies in r-band luminosity-threshold samples. The decorated HOD is a model for the galaxy–halo connection that augments the traditional HOD by allowing for the possibility of galaxy assembly bias: galaxy luminosity may be correlated with dark matter halo properties besides mass, Mvir. We demonstrate that it is not possible to rule out galaxy assembly bias using DR7 measurements of galaxy clustering alone. Moreover, galaxy samples with Mr < −20 and Mr < −20.5 favour central galaxy assembly bias. These samples prefer scenarios in which high-concentration haloes are more likely to host a central galaxy relative to low-concentration haloes of the same Mvir. We formally rule out zero assembly bias with high significance for these samples. In the particular case of the Mr < −20 sample, zero assembly bias is excluded at greater than the 3σ level. Satellite galaxy assembly bias is preferred for the faintest sample we study, Mr < −19. We find no evidence for assembly bias in the Mr < −21 sample. Assembly bias should be accounted for in galaxy clustering analyses or attempts to exploit galaxy clustering to constrain cosmology. In addition to presenting the first constraints on HOD models that accommodate assembly bias, our analysis includes numerous improvements over previous analyses of this data set and supersedes previously published results, even in the case of a standard HOD analysis.