arXiv subjects

Philip Mansfield

Publications and source records attributed to Philip Mansfield.

27 records · Page 2Linked to original sources

arXiv2021

Approximating Density Probability Distribution Functions Across Cosmologies

Using a suite of self-similar cosmological simulations, we measure the probability distribution functions (PDFs) of real-space density, redshift-space density, and their geometric mean. We find that the real-space density PDF is well-described by a function of two parameters: $n_s$, the spectral slope, and $σ_L$, the linear rms density fluctuation. For redshift-space density and the geometric mean of real- and redshift-space densities, we introduce a third parameter, $s_L={\sqrt{\langle(dv^L_{\rm pec}/dr)^2\rangle}}/{H}$. We find that density PDFs for the LCDM cosmology is also well-parameterized by these three parameters. As a result, we are able to use a suite of self-similar cosmological simulations to approximate density PDFs for a range of cosmologies. We make the density PDFs publicly available and provide an analytical fitting formula for them.

astro-ph.CO↗

arXiv2021

Contrastive Learning for Label-Efficient Semantic Segmentation

Collecting labeled data for the task of semantic segmentation is expensive and time-consuming, as it requires dense pixel-level annotations. While recent Convolutional Neural Network (CNN) based semantic segmentation approaches have achieved impressive results by using large amounts of labeled training data, their performance drops significantly as the amount of labeled data decreases. This happens because deep CNNs trained with the de facto cross-entropy loss can easily overfit to small amounts of labeled data. To address this issue, we propose a simple and effective contrastive learning-based training strategy in which we first pretrain the network using a pixel-wise, label-based contrastive loss, and then fine-tune it using the cross-entropy loss. This approach increases intra-class compactness and inter-class separability, thereby resulting in a better pixel classifier. We demonstrate the effectiveness of the proposed training strategy using the Cityscapes and PASCAL VOC 2012 segmentation datasets. Our results show that pretraining with the proposed contrastive loss results in large performance gains (more than 20% absolute improvement in some settings) when the amount of labeled data is limited. In many settings, the proposed contrastive pretraining strategy, which does not use any additional data, is able to match or outperform the widely-used ImageNet pretraining strategy that uses more than a million additional labeled images.

cs.CV↗

arXiv2021

Debiased Galaxy Cluster Pressure Profiles from X-ray Observations and Simulations

We present an updated model for the average cluster pressure profile, adjusted for hydrostatic mass bias by combining results from X-ray observations with cosmological simulations. Our model estimates this bias by fitting a power-law to the relation between the "true" halo mass and X-ray cluster mass in hydrodynamic simulations (IllustrisTNG, BAHAMAS, and MACSIS). As an example application, we consider the REXCESS X-ray cluster sample and the Universal Pressure Profile (UPP) derived from scaled and stacked pressure profiles. We find adjusted masses, $M_\mathrm{500c},$ that are $\lesssim$15% higher and scaled pressures $P/P_\mathrm{500c}$ that have $\lesssim$35% lower normalization than previously inferred. Our Debiased Pressure Profile (DPP) is well-fit by a Generalized Navarro-Frenk-White (GNFW) function, with parameters $[P_0,c_{500},α,β,γ]=[5.048,1.217,1.192,5.490,0.433]$ and does not require a mass-dependent correction term. When the DPP is used to model the Sunyaev-Zel'dovich (SZ) effect, we find that the integrated Compton $Y-M$ relation has only minor deviations from self-similar scaling. The thermal SZ angular power spectrum is lower in amplitude by approximately 30%, assuming nominal cosmological parameters (e.g. $Ω_\text{m}=0.3$, $σ_8 = 0.8$), and is broadly consistent with recent Planck results without requiring additional bias corrections.

astro-ph.CO↗

arXiv2020

How Biased Are Halo Properties in Cosmological Simulations?

Cosmological N-body simulations have been a major tool of theorists for decades, yet many of the numerical issues that these simulations face are still unexplored. This paper measures numerical biases in these large, dark matter-only simulations that affect the properties of their dark matter haloes. We compare many simulation suites in order to provide several tools for simulators and analysts which help mitigate these biases. We summarise our comparisons with practical `convergence limits' that can be applied to a wide range of halo properties, including halo properties which are traditionally overlooked by the testing literature. We also find that the halo properties predicted by different simulations can diverge from one another at unexpectedly high resolutions. We demonstrate that many halo properties depend strongly on force softening scale and that this dependence leads to much of the measured divergence between simulations. We offer an empirical model to estimate the impact of such effects on the rotation curves of a halo population. This model can serve as a template for future empirical models of the biases in other halo properties.

astro-ph.CO↗

arXiv2020

The Sheet of Giants: Unusual Properties of the Milky Way's Immediate Neighbourhood

We quantify the shape and overdensity of the galaxy distribution in the `Local Sheet' within a sphere of $R=8$ Mpc, and compare these properties with the expectations of the $Λ$CDM model. We measure ellipsoidal axis ratios of $c/a\approx0.16$ and $b/a\approx0.79$, indicating that the distribution of galaxies in the Local Volume can be approximated by a flattened oblate ellipsoid, consistent with the `sheet'-like configuration noted in previous studies. In contrast with previous estimates that the Local Sheet has a density close to average, we find that the number density of faint and bright galaxies in the Local Volume is $\approx1.7$ and $\approx5.2$ times denser, respectively, than the mean number density of galaxies of the same luminosity. Comparison with simulations shows that the number density contrasts of bright and faint galaxies within $8$ Mpc alone make the Local Volume a $\approx 2.5σ$ outlier in the $Λ$CDM cosmology. Our results indicate that the cosmic neighbourhood of the Milky Way may be unusual for galaxies of similar luminosity. The impact of the peculiar properties of our neighbourhood on the properties of the Milky Way and other nearby galaxies is not yet understood and warrants further study.

astro-ph.GA↗

arXiv2019

The Three Causes of Low-Mass Assembly Bias

We present a detailed analysis of the physical processes that cause halo assembly bias -- the dependence of halo clustering on proxies of halo formation time. We focus on the origin of assembly bias in the mass range corresponding to the hosts of typical galaxies and use halo concentration as our chief proxy of halo formation time. We also repeat our key analyses across a broad range of halo masses and for alternative formation time definitions. We show that splashback subhaloes are responsible for two thirds of the assembly bias signal, but do not account for the entire effect. After splashback subhaloes have been removed, we find that the remaining assembly bias signal is due to a relatively small fraction ($\lesssim 10\%$) of haloes in dense regions. We test a number of additional physical processes thought to contribute to assembly bias and demonstrate that the two key processes are the slowing of mass growth by large-scale tidal fields and by the high velocities of ambient matter in sheets and filaments. We also rule out several other proposed physical causes of halo assembly bias. Based on our results, we argue that there are three processes that contribute to assembly bias of low-mass halos: large-scale tidal fields, gravitational heating due to the collapse of large-scale structures, and splashback subhaloes located outside the virial radius.

astro-ph.CO↗

arXiv2017

The splashback radius of halos from particle dynamics. II. Dependence on mass, accretion rate, redshift, and cosmology

The splashback radius $R_{\rm sp}$, the apocentric radius of particles on their first orbit after falling into a dark matter halo, has recently been suggested as a physically motivated halo boundary that separates accreting from orbiting material. Using the SPARTA code presented in Paper I, we analyze the orbits of billions of particles in cosmological simulations of structure formation and measure $R_{\rm sp}$ for a large sample of halos that span a mass range from dwarf galaxy to massive cluster halos, reach redshift 8, and include WMAP, Planck, and self-similar cosmologies. We analyze the dependence of $R_{\rm sp}/R_{\rm 200m}$ and $M_{\rm sp}/M_{\rm 200m}$ on the mass accretion rate $Γ$, halo mass, redshift, and cosmology. The scatter in these relations varies between 0.02 and 0.1 dex. While we confirm the known trend that $R_{\rm sp}/R_{\rm 200m}$ decreases with $Γ$, the relationships turn out to be more complex than previously thought, demonstrating that $R_{\rm sp}$ is an independent definition of the halo boundary that cannot trivially be reconstructed from spherical overdensity definitions. We present fitting functions for $R_{\rm sp}/R_{\rm 200m}$ and $M_{\rm sp}/M_{\rm 200m}$ as a function of accretion rate, peak height, and redshift, achieving an accuracy of 5% or better everywhere in the parameter space explored. We discuss the physical meaning of the distribution of particle apocenters and show that the previously proposed definition of $R_{\rm sp}$ as the radius of the steepest logarithmic density slope encloses roughly three-quarters of the apocenters. Finally, we conclude that no analytical model presented thus far can fully explain our results.

astro-ph.CO↗

arXiv2017

Splashback Shells of Cold Dark Matter Halos

The density field in the outskirts of dark matter halos is discontinuous due to a caustic formed by matter at its first apocenter after infall. In this paper, we present an algorithm to identify the "splashback shell" formed by these apocenters in individual simulated halos using only a single snapshot of the density field. We implement this algorithm in the code SHELLFISH (SHELL Finding In Spheroidal Halos) and demonstrate that the code identifies splashback shells correctly and measures their properties with an accuracy of $<5\%$ for halos with more than 50,000 particles and mass accretion rates of $Γ_\textrm{DK14}>0.5$. Using SHELLFISH, we present the first estimates for several basic properties of individual splashback shells, such as radius, $R_\textrm{sp}$, mass, and overdensity, and provide fits to the distribution of these quantities as functions of $Γ_\textrm{DK14}$, $ν_\textrm{200m}$, and $z.$ We confirm previous findings that $R_\textrm{sp}$ decreases with increasing $Γ_\textrm{DK14}$, but we show that independent of accretion rate, it also decreases with increasing $ν_\textrm{200m}$. We also study the 3D structures of these shells and find that they generally have non-ellipsoidal oval shapes. We find that splashback radii estimated by SHELLFISH are $20\%-30\%$ larger than those estimated in previous studies from stacked density profiles at high accretion rates. We demonstrate that the latter are biased low due to the contribution of high-mass subhalos to these profiles and show that using the median instead of mean density in each radial bin mitigates the effect of substructure on density profiles and removes the bias.

astro-ph.CO↗

arXiv2015

SCORCH I: The Galaxy-Halo Connection in the First Billion Years

SCORCH (Simulations and Constructions of the Reionization of Cosmic Hydrogen) is a new project to study the Epoch of Reionization (EoR). In this first paper, we probe the connection between observed high-redshift galaxies and simulated dark matter halos to better understand the abundance and evolution of the primary source of ionizing radiation. High-resolution N-body simulations are run to quantify the abundance of dark matter halos as a function of mass $M$, accretion rate $\dot{M}$, and redshift $z$. A new fit for the halo mass function $dn/dM$ is $\approx 20\%$ more accurate at the high-mass end where bright galaxies are expected to reside. A novel approach is used to fit the halo accretion rate function $dn/d\dot{M}$ in terms of the halo mass function. Abundance matching against the observed galaxy luminosity function is used to estimate the luminosity-mass relation and the luminosity-accretion-rate relation. The inferred star formation efficiency is not monotonic with $M$ nor $\dot{M}$, but reaches a maximum value at a characteristic mass $\sim 2 \times 10^{11}\ M_\odot$ and a characteristic accretion rate $\sim 6 \times 10^2\ M_\odot/{\rm yr}$ at $z \approx 6$. We find a universal EoR luminosity-accretion-rate relation and construct a fiducial model for the galaxy luminosity function. The Schechter parameters evolve such that $ϕ_\star$ decreases, $M_\star$ is more positive (fainter), and $α$ is more negative (steeper) at higher redshifts. We forecast for the upcoming James Webb Space Telescope and show that with apparent magnitude limit $m_{\rm AB} \approx 31\ (32)$, it can observe $\gtrsim 11\ (24)$ unlensed galaxies per square degree per unit redshift at least down to $M_\star$ at $z \lesssim 13\ (14)$.

astro-ph.CO↗