We describe the design and performance of the near-infrared (1.51-1.70μm), fiber-fed, multi-object (300 fibers), high resolution (R = λ/∆λ ̃22,500) spectrograph built for the Apache Point Observatory GalacticEvolution Experiment (APOGEE). APOGEE is a survey of ̃10⁵ redgiant stars that systematically sampled all Milky Way populations(bulge, disk, and halo) to study the Galaxy’s chemical and kinematicalhistory. It was part of the Sloan Digital Sky Survey III (SDSS-III) from2011 to 2014 using the 2.5 m Sloan Foundation Telescope at Apache PointObservatory, New Mexico. The APOGEE-2 survey is now using thespectrograph as part of SDSS-IV, as well as a second spectrograph, aclose copy of the first, operating at the 2.5 m du Pont Telescope at LasCampanas Observatory in Chile. Although several fiber-fed, multi-object,high resolution spectrographs have been built for visual wavelengthspectroscopy, the APOGEE spectrograph is one of the first suchinstruments built for observations in the near-infrared. Theinstrument’s successful development was enabled by several keyinnovations, including a “gang connector” to allow simultaneousconnections of 300 fibers; hermetically sealed feedthroughs to allowfibers to pass through the cryostat wall continuously; the firstcryogenically deployed mosaic volume phase holographic grating; and alarge refractive camera that includes mono-crystalline silicon and fusedsilica elements with diameters as large as ̃400 mm. This paper containsa comprehensive description of all aspects of the instrument includingthe fiber system, optics and opto-mechanics, detector arrays, mechanicsand cryogenics, instrument control, calibration system, opticalperformance and stability, lessons learned, and design changes for thesecond instrument.

We apply an iterative reconstruction method to galaxy mocks in redshiftspace obtained from N-body simulations. Comparing the two-pointcorrelation functions for the reconstructed density field, we find thatalthough the performance is limited by shot noise and galaxy biascompared to the matter field, the iterative method can still reconstructthe initial linear density field from the galaxy field better than thestandard method both in real and in redshift space. Furthermore, theiterative method is able to reconstruct both the monopole and quadrupolemore precisely, unlike the standard method. We see that as the numberdensity of galaxies gets smaller, the performance of reconstruction getsworse due to the sparseness. However, the precision in the determinationof bias ({̃ }20{{ per cent}}) hardly impacts on the reconstructionprocesses.

We probe the higher-order galaxy clustering in the final data release ofthe Sloan Digital Sky Survey Baryon Oscillation Spectroscopic Survey(BOSS) using germ-grain Minkowski functionals (MFs). Our data selectioncontains 979 430 BOSS galaxies from both the Northern and SouthernGalactic Caps over the redshift range z = 0.2-0.6. We extract thehigher-order part of the MFs, detecting the deviation from the purelyGaussian case with χ ^2 ̃ O(10^3) on 24 degrees of freedom across theentire data selection. We measure significant redshift evolution in thehigher-order functionals for the first time. We find 15-35{{ per cent}}growth, depending on functional and scale, between our redshift binscentred at z = 0.325 and z = 0.525. We show that the structure inhigher-order correlations grows faster than that in the two-pointcorrelations, especially on small scales where the excess approaches afactor of 2. We demonstrate how this trend is generalizable by findinggood agreement of the data with a hierarchical model in which the higherorders grow faster than the lower-order correlations. We find that thenon-Gaussianity of the underlying dark matter field grows even fasterthan the one of the galaxies. Our method can be adapted to study theredshift evolution of the three-point and higher functions individually.

Covariance matrix estimation is a persistent challenge for cosmology,often requiring a large number of synthetic mock catalogues. The off-diagonal components of the covariance matrix also make it difficult toshow representative error bars on the 2-point correlation function(2PCF) since errors computed from the diagonal values of the covariancematrix greatly underestimate the uncertainties. We develop a routine fordecorrelating the projected and anisotropic 2PCF with simple and scale-compact transformations on the 2PCF. These transformation matrices aremodelled after the Cholesky decomposition and the symmetric square rootof the Fisher matrix. Using mock catalogues, we show that thetransformed projected and anisotropic 2PCF recover the same structure asthe original 2PCF while producing largely decorrelated error bars.Specifically, we propose simple Cholesky-based transformation matricesthat suppress the off-diagonal covariances on the projected 2PCF by {̃ }95{{ per cent}} and that on the anisotropic 2PCF by {̃ } 87{{ percent}}. These transformations also serve as highly regularized models ofthe Fisher matrix, compressing the degrees of freedom so that one canfit for the Fisher matrix with a much smaller number of mocks.

We present a machine-learning (ML) approach for estimating galaxycluster masses from Chandra mock images. We utilize a ConvolutionalNeural Network (CNN), a deep ML tool commonly used in image recognitiontasks. The CNN is trained and tested on our sample of 7896 Chandra X-raymock observations, which are based on 329 massive clusters from the{\text{}}{IllustrisTNG} simulation. Our CNN learns from a low resolutionspatial distribution of photon counts and does not use spectralinformation. Despite our simplifying assumption to neglect spectralinformation, the resulting mass values estimated by the CNN exhibitsmall bias in comparison to the true masses of the simulated clusters(-0.02 dex) and reproduce the cluster masses with low intrinsic scatter,

The combination of galaxy-galaxy lensing (GGL) with galaxy clustering isone of the most promising routes to determining the amplitude of matterclustering at low redshifts. We show that extending clustering+GGLanalyses from the linear regime down to {̃ } 0.5 h^{-1} Mpc scalesincreases their constraining power considerably, even aftermarginalizing over a flexible model of non-linear galaxy bias. Using agrid of cosmological N-body simulations, we construct a Taylor-expansionemulator that predicts the galaxy autocorrelation ξ_gg(r) andgalaxy-matter cross-correlation ξ_gm(r) as a function ofσ₈, Ω_m, and halo occupation distribution (HOD)parameters, which are allowed to vary with large-scale environment torepresent possible effects of galaxy assembly bias. We present forecastsfor a fiducial case that corresponds to BOSS LOWZ galaxy clustering andSDSS-depth weak lensing (effective source density ̃0.3arcmin^-2). Using tangential shear and projected correlationfunction measurements over 0.5 ≤ r_ p ≤ 30 h^{-1} Mpc yields a 2 percent constraint on the parameter combination σ _8Ω _ m^{0.6}, a factorof two better than a constraint that excludes non-linear scales (r_ p> 2 h^{-1} Mpc, 4 h^{-1} Mpc for γ_t, w_p). Muchof this improvement comes from the non-linear clustering information,which breaks degeneracies among HOD parameters. Increasing the effectivesource density to 3 arcmin^-2 sharpens the constraint on σ _8Ω_ m^{0.6} by a further factor of two. With robust modelling into thenon-linear regime, low-redshift measurements of matter clustering at the1-per cent level with clustering+GGL alone are well within reach ofcurrent data sets such as those provided by the Dark Energy Survey.

We present configuration-space estimators for the auto- and cross-covariance of two- and three-point correlation functions (2PCF and 3PCF)in general survey geometries. These are derived in the Gaussian limit(setting higher order correlation functions to zero), but for arbitrarynon-linear 2PCFs (which may be estimated from the survey itself), with ashot-noise rescaling parameter included to capture non-Gaussianity. Wegeneralize previous approaches to include Legendre moments via ageometry-correction function calibrated from measured pair and triplecounts. Making use of importance sampling and random particlecatalogues, we can estimate model covariances in fractions of the timerequired to do so with mocks, obtaining estimates with negligiblesampling noise in ̃10 (̃100) CPU-hours for the 2PCF (3PCF)autocovariance. We compare results to sample covariances from a suite ofBOSS DR12 mocks and find the matrices to be in good agreement, assuminga shot-noise rescaling parameter of 1.03 (1.20) for the 2PCF (3PCF). Toobtain strongest constraints on cosmological parameters, we must usemultiple statistics in concert; having robust methods to measure theircovariances at low computational cost is thus of great relevance toupcoming surveys.

Although photometric redshifts (photo-z's) are crucial ingredients forcurrent and upcoming large-scale surveys, the high-quality spectroscopicredshifts currently available to train, validate, and test them aresubstantially non-representative in both magnitude and colour. Weinvestigate the nature and structure of this bias by tracking howobjects from a heterogeneous training sample contribute to photo-zpredictions as a function of magnitude and colour, and illustrate thatthe underlying redshift distribution at fixed colour can evolve stronglyas a function of magnitude. We then test the robustness of the galaxy-galaxy lensing signal in 120 deg² of HSC-SSP DR1 data tospectroscopic completeness and photo-z biases, and find that theirimpacts are sub-dominant to current statistical uncertainties. Ourmethodology provides a framework to investigate how spectroscopicincompleteness can impact photo-z-based weak lensing predictions infuture surveys such as LSST and WFIRST.

We present a method for generating suites of dark matter halo catalogueswith only a few N-body simulations, focusing on making small changes tothe underlying cosmology of a simulation with high precision. In thecontext of blind challenges, this allows us to re-use a simulation bygiving it a new cosmology after the original cosmology is revealed.Starting with full N-body realizations of an original cosmology and atarget cosmology, we fit a transfer function that displaces haloes inthe original so that the galaxy/HOD power spectrum matches that of thetarget cosmology. This measured transfer function can then be applied toa new realization of the original cosmology to create a new realizationof the target cosmology. For a 1 per cent change in σ₈, weachieve 0.1 per cent accuracy to k = 1 h Mpc^{-1} in the real-spacepower spectrum; this degrades to 0.3 per cent when the transfer functionis applied to a new realization. We achieve similar accuracy in theredshift-space monopole and quadrupole. In all cases, the result isbetter than the sample variance of our 1.1 h^{-1} Gpc simulation boxes.

We present a high-fidelity realization of the cosmological N-bodysimulation from the Schneider et al. code comparison project. Thesimulation was performed with our ABACUSN-body code, which offers high-force accuracy, high performance, and minimal particle integrationerrors. The simulation consists of 2048³ particles in a 500h^{-1} Mpc box for a particle mass of 1.2× 10^9 h^{-1} M_\odot with 10h^{-1} kpc spline softening. ABACUS executed 1052 global time-steps to z= 0 in 107 h on one dual-Xeon, dual-GPU node, for a mean rate of 23million particles per second per step. We find ABACUS is in goodagreement with RAMSES and PKDGRAV3 and less so with GADGET3. We validateour choice of time-step by halving the step size and find sub-percentdifferences in the power spectrum and 2PCF at nearly all measuredscales, with {\lt }0.3{{ per cent}} errors at k\lt 10 Mpc^{-1} h. Onlarge scales, ABACUS reproduces linear theory better than 0.01 per cent.Simulation snapshots are available athttp://nbody.rc.fas.harvard.edu/public/S2016.

Covariance matrix estimation is a persistent challenge for cosmology. Wefocus on a class of model covariance matrices that can be generated withhigh accuracy and precision, using a tiny fraction of the computationalresources that would be required to achieve comparably precisecovariance matrices using mock catalogues. In previous work, the freeparameters in these models were determined using sample covariancematrices computed using a large number of mocks, but we demonstrate thatthose parameters can be estimated consistently and with good precisionby applying jackknife methods to a single survey volume. This enablesmodel covariance matrices that are calibrated from data alone, with noreference to mocks.

The DESI Legacy Imaging Surveys (http://legacysurvey.org/) are acombination of three public projects (the Dark Energy Camera LegacySurvey, the Beijing-Arizona Sky Survey, and the Mayall z-band LegacySurvey) that will jointly image ≈14,000 deg² of theextragalactic sky visible from the northern hemisphere in three opticalbands (g, r, and z) using telescopes at the Kitt Peak NationalObservatory and the Cerro Tololo Inter-American Observatory. Thecombined survey footprint is split into two contiguous areas by theGalactic plane. The optical imaging is conducted using a unique strategyof dynamically adjusting the exposure times and pointing selectionduring observing that results in a survey of nearly uniform depth. Inaddition to calibrated images, the project is delivering a catalog,constructed by using a probabilistic inference-based approach toestimate source shapes and brightnesses. The catalog includes photometryfrom the grz optical bands and from four mid-infrared bands (at 3.4,4.6, 12, and 22 μm) observed by the Wide-field Infrared Survey Explorersatellite during its full operational lifetime. The project plans twopublic data releases each year. All the software used to generate thecatalogs is also released with the data. This paper provides an overviewof the Legacy Surveys project.

We use the IllustrisTNG (TNG) simulations to explore the galaxy-haloconnection as inferred from state-of-the-art cosmological,magnetohydrodynamical simulations. With the high-mass resolution andlarge volume achieved by combining the 100 Mpc (TNG100) and 300 Mpc(TNG300) volumes, we establish the mean occupancy of central andsatellite galaxies and their dependence on the properties of the darkmatter haloes hosting them. We derive best-fitting HOD parameters fromTNG100 and TNG300 for target galaxy number densities of \bar{n}_g =0.032 and \bar{n}_g = 0.016 h^3 Mpc^-3, respectively,corresponding to a minimum galaxy stellar mass of M_\star ̃ 1.9× 10^9and M_\star ̃ 3.5× 10^9 M_☉, respectively, in hosts more massive than10^{11} M_☉. Consistent with previous work, we find that haloes locatedin dense environments, with low concentrations, later formation times,and high angular momenta are richest in their satellite population. Atlow mass, highly concentrated haloes and those located in overdenseregions are more likely to contain a central galaxy. The degree ofenvironmental dependence is sensitive to the definition adopted for thephysical boundary of the host halo. We examine the extent to whichcorrelations between galaxy occupancy and halo properties areindependent and demonstrate that HODs predicted by halo mass andpresent-day concentration capture the qualitative dependence on theremaining halo properties. At fixed halo mass, concentration is a strongpredictor of the stellar mass of the central galaxy, which may play adefining role in the fate of the satellite population. The radialdistribution of satellite galaxies, which exhibits a universal formacross a wide range of host halo mass, is described accurately by thebest-fitting NFW density profile of their host haloes.

In preparation for deep extragalactic imaging with the James Webb SpaceTelescope, we explore the clustering of massive halos at z = 8 and 10using a large N-body simulation. We find that halos with masses of10⁹-10¹¹ h ^-1 M _☉, which arethose expected to host galaxies detectable with JWST, are highlyclustered with bias factors ranging from 5 to 30 depending strongly onmass, as well as on redshift and scale. This results in correlationlengths of 5-10 h ^-1 Mpc, similar to those of today’sgalaxies. Our results are based on a simulation of 130 billion particlesin a box of size 250 h ^-1 Mpc using our new high-accuracyABACUS simulation code, the corrections to cosmological initialconditions of Garrison et al., and the Planck 2015 cosmology. We usevariations between sub-volumes to estimate the detectability of theclustering. Because of the very strong interhalo clustering, we findthat a medium-sized survey with a transverse size of the order of 25 h^-1 comoving Mpc (about 13′) may be able to detect theclustering of z = 8-10 galaxies with only 500-1000 survey objects if thegalaxies indeed occupy the most massive dark matter halos.

We utilize mock catalogues from high-accuracy cosmological N-bodysimulations to quantify shifts in the recovery of the acoustic scalethat could potentially result from galaxy clustering bias. Therelationship between galaxies and dark matter haloes presents acomplicated source of systematic errors in modern redshift surveys,particularly when aiming to make cosmological measurements to sub-percent precision. Apart from a scalar, linear bias parameter accountingfor the density contrast ratio between matter tracers and the truematter distribution, other types of galaxy bias, such as assembly andvelocity biases, may also significantly alter clustering signals fromsmall to large scales. We create mocks based on generalized halooccupation populations of 36 periodic boxes from the ABACUSCOSMOSrelease, and test various biased models along with an unbiasedbase case in a total volume of 48 h^{-3} Gpc³. Tworeconstruction methods are applied to galaxy samples and the apparentacoustic scale is derived by fitting the two-point correlation functionmultipoles. With respect to the baseline, we find a 0.3 per cent shiftin the line-of-sight acoustic scale for one variation in the satellitegalaxy population, and we find a 0.7 per cent shift for an extreme levelof velocity bias of the central galaxies. All other bias models areconsistent with zero shift at the 0.2 per cent level afterreconstruction. We note that the bias models explored are relativelylarge variations, producing sizeable and likely distinguishable changesin small-scale clustering, the modelling of which would furthercalibrate the baryon acoustic oscillations standard ruler.

We present a public data release of halo catalogs from a suite of 125cosmological N-body simulations from the ABACUS project. The simulationsspan 40 wCDM cosmologies centered on the Planck 2015 cosmology at twomass resolutions, 4 × 10¹⁰ h ^-1 M_⊙ and 1 × 10¹⁰ h ^-1 M_⊙, in 1.1 h ^-1 Gpc and 720 h^-1 Mpc boxes, respectively. The boxes are phase-matchedto suppress sample variance and isolate cosmology dependence. Additionalvolume is available via 16 boxes of fixed cosmology and varied phase; afew boxes of single-parameter excursions from Planck 2015 are alsoprovided. Catalogs spanning z = 1.5 to 0.1 are available forfriends-of-friends and ROCKSTAR halo finders and include particlesubsamples. All data products are available at https://lgarrison.github.io/AbacusCosmos.

We develop a formalism for measuring the cosmological distance scalefrom baryon acoustic oscillations (BAO) using the cross-correlation of asparse redshift survey with a denser photometric sample. This reducesthe shot noise that would otherwise affect the autocorrelation of thesparse spectroscopic map. As a proof of principle, we make the firston-sky application of this method to a sparse sample defined as the z> 0.6 tail of the Sloan Digital Sky Survey's (SDSS) BOSS/CMASS sampleof galaxies and a dense photometric sample from SDSS DR9. We find a2.8σ preference for the BAO peak in the cross-correlation at aneffective z = 0.64, from which we measure the angular diameter distanceD_M(z = 0.64) = (2418 ± 73 Mpc)(r_s/r_s,fid). Accordingly, we expect that using this method to combinesparse spectroscopy with the deep, high-quality imaging that is just nowbecoming available will enable higher precision BAO measurements thanpossible with the spectroscopy alone.

We investigate the potential sources of theoretical systematics in theanisotropic Baryon Acoustic Oscillation (BAO) distance scalemeasurements from the clustering of galaxies in configuration spaceusing the final Data Release (DR12) of the Baryon OscillationSpectroscopic Survey (BOSS). We perform a detailed study of the impacton BAO measurements from choices in the methodology such as fiducialcosmology, clustering estimators, random catalogues, fitting templates,and covariance matrices. The theoretical systematic uncertainties in BAOparameters are found to be 0.002 in the isotropic dilation α and0.003 in the quadrupolar dilation ɛ. The leading source ofsystematic uncertainty is related to the reconstruction techniques.Theoretical uncertainties are sub-dominant compared with the statisticaluncertainties for BOSS survey, accounting 0.2σ_stat forα and 0.25σ_stat for ɛ(σ_{α, stat} ˜ 0.010 andσ_{ɛ, stat} ˜ 0.012, respectively). We alsopresent BAO-only distance scale constraints from the anisotropicanalysis of the correlation function. Our constraints on the angulardiameter distance D_A(z) and the Hubble parameter H(z),including both statistical and theoretical systematic uncertainties, are1.5 per cent and 2.8 per cent at z_eff = 0.38, 1.4 per centand 2.4 per cent at z_eff = 0.51, and 1.7 per cent and 2.6 percent at z_eff = 0.61. This paper is part of a set thatanalyses the final galaxy clustering data set from BOSS. Themeasurements and likelihoods presented here are cross-checked with otherBAO analysis in Alam et al. The systematic error budget concerning themethodology on post-reconstruction BAO analysis presented here is usedin Alam et al. to produce the final cosmological constraints from BOSS.

We present measurements of the Baryon Acoustic Oscillation (BAO) scalein redshift-space using the clustering of quasars. We consider a sampleof 147 000 quasars from the extended Baryon Oscillation SpectroscopicSurvey (eBOSS) distributed over 2044 square degrees with redshifts 0.8< z < 2.2 and measure their spherically averaged clustering inboth configuration and Fourier space. Our observational data set and the1400 simulated realizations of the data set allow us to detect apreference for BAO that is greater than 2.8σ. We determine thespherically averaged BAO distance to z = 1.52 to 3.8 per cent precision:D_V(z = 1.52) = 3843 ± 147(r_d/r_d,fid)Mpc. This is the first time the location of the BAO featurehas been measured between redshifts 1 and 2. Our result is fullyconsistent with the prediction obtained by extrapolating the Planck flatΛCDM best-fitting cosmology. All of our results are consistentwith basic large-scale structure (LSS) theory, confirming quasars to bea reliable tracer of LSS, and provide a starting point for numerouscosmological tests to be performed with eBOSS quasar samples. We combineour result with previous, independent, BAO distance measurements toconstruct an updated BAO distance-ladder. Using these BAO data alone andmarginalizing over the length of the standard ruler, we findΩ_Λ > 0 at 6.6σ significance whentesting a ΛCDM model with free curvature.

We search for a galaxy clustering bias due to a modulation of galaxynumber with the baryon-dark matter relative velocity resulting fromrecombination-era physics. We find no detected signal and place theconstraint b_v < 0.01 on the relative velocity bias for theCMASS galaxies. This bias is an important potential systematic of baryonacoustic oscillation (BAO) method measurements of the cosmic distancescale using the two-point clustering. Our limit on the relative velocitybias indicates a systematic shift of no more than 0.3 per cent rms inthe distance scale inferred from the BAO feature in the BOSS two-pointclustering, well below the 1 per cent statistical error of thismeasurement. This constraint is the most stringent currently availableand has important implications for the ability of upcoming large-scalestructure surveys such as the Dark Energy Spectroscopic Instrument(DESI) to self-protect against the relative velocity as a possiblesystematic.