Prof. Matt Jarvis

Abstract:

© 2020. The Astronomical Society of the Pacific. The Very Large Array Sky Survey (VLASS) is a synoptic, all-sky radio sky survey with a unique combination of high angular resolution (≈2.″5), sensitivity (a 1σ goal of 70 μJy/beam in the coadded data), full linear Stokes polarimetry, time domain coverage, and wide bandwidth (2–4 GHz). The first observations began in 2017 September, and observing for the survey will finish in 2024. VLASS will use approximately 5500 hr of time on the Karl G. Jansky Very Large Array (VLA) to cover the whole sky visible to the VLA (decl. > −40°), a total of 33 885 deg2. The data will be taken in three epochs to allow the discovery of variable and transient radio sources. The survey is designed to engage radio astronomy experts, multi-wavelength astronomers, and citizen scientists alike. By utilizing an “on the fly” interferometry mode, the observing overheads are much reduced compared to a conventional pointed survey. In this paper, we present the science case and observational strategy for the survey, and also results from early survey observations.

Abstract:

We present 15-GHz Karl G. Jansky Very Large Array observations of a complete sample of radio galaxies selected at 15.7 GHz from the Tenth Cambridge (10C) survey. 67 out of the 95 sources (71 per cent) are unresolved in the new observations and lower frequency radio observations, placing an upper limit on their angular size of ∼2 arcsec. Thus, compact radio galaxies, or radio galaxies with very faint jets, are the dominant population in the 10C survey. This provides support for the suggestion in our previous work that low-luminosity (⁠L<1025W~Hz−1⁠) radio galaxies are core dominated, although higher resolution observations are required to confirm this directly. The 10C sample of compact, high-frequency selected radio galaxies is a mixture of high-excitation and low-excitation radio galaxies and displays a range of radio spectral shapes, demonstrating that they are a mixed population of objects.

Abstract:

We utilize deep near-infrared survey data from the UltraVISTA fourth data release (DR4) and the VIDEO survey, in combination with overlapping optical and Spitzer data, to search for bright star-forming galaxies at z ≳ 7.5. Using a full photometric redshift fitting analysis applied to the ∼6 deg2 of imaging searched, we find 27 Lyman break galaxies (LBGs), including 20 new sources, with best-fitting photometric redshifts in the range 7.4 < z < 9.1. From this sample, we derive the rest-frame UV luminosity function at z = 8 and z = 9 out to extremely bright UV magnitudes (MUV ≃ −23) for the first time. We find an excess in the number density of bright galaxies in comparison to the typically assumed Schechter functional form derived from fainter samples. Combined with previous studies at lower redshift, our results show that there is little evolution in the number density of very bright (MUV ∼ −23) LBGs between z ≃ 5 and z ≃ 9. The tentative detection of an LBG with best-fitting photometric redshift of z = 10.9 ± 1.0 in our data is consistent with the derived evolution. We show that a double power-law fit with a brightening characteristic magnitude (ΔM*/Δz ≃ −0.5) and a steadily steepening bright-end slope (Δβ/Δz ≃ −0.5) provides a good description of the z > 5 data over a wide range in absolute UV magnitude (−23 < MUV < −17). We postulate that the observed evolution can be explained by a lack of mass quenching at very high redshifts in combination with increasing dust obscuration within the first ∼1Gyr of galaxy evolution.

Abstract:

We present the radio luminosity function (RLF) of optically selected quasars below 1 mJy, constructed by applying a Bayesian-fitting stacking technique to objects well below the nominal radio flux density limit. We test the technique using simulated data, confirming that we can reconstruct the RLF over three orders of magnitude below the typical 5σ detection threshold. We apply our method to 1.4-GHz flux densities from the Faint Images of the Radio Sky at Twenty-Centimeters (FIRST) survey, extracted at the positions of optical quasars from the Sloan Digital Sky Survey over seven redshift bins up to z = 2.15, and measure the RLF down to two orders of magnitude below the FIRST detection threshold. In the lowest redshift bin (0.2 < z < 0.45), we find that our measured RLF agrees well with deeper data from the literature. The RLF for the radio-loud quasars flattens below log10[L1.4/WHz−1]≈25.5 and becomes steeper again below log10[L1.4/WHz−1]≈24.8⁠, where radio-quiet quasars start to emerge. The radio luminosity where radio-quiet quasars emerge coincides with the luminosity where star-forming galaxies are expected to start dominating the radio source counts. This implies that there could be a significant contribution from star formation in the host galaxies, but additional data are required to investigate this further. The higher redshift bins show a similar behaviour to the lowest z bin, implying that the same physical process may be responsible.

Abstract:

Computer simulations are invaluable tools for scientific discovery. However, accurate simulations are often slow to execute, which limits their applicability to extensive parameter exploration, large-scale data analysis, and uncertainty quantification. A promising route to accelerate simulations by building fast emulators with machine learning requires large training datasets, which can be prohibitively expensive to obtain with slow simulations. Here we present a method based on neural architecture search to build accurate emulators even with a limited number of training data. The method successfully accelerates simulations by up to 2 billion times in 10 scientific cases including astrophysics, climate science, biogeochemistry, high energy density physics, fusion energy, and seismology, using the same super-architecture, algorithm, and hyperparameters. Our approach also inherently provides emulator uncertainty estimation, adding further confidence in their use. We anticipate this work will accelerate research involving expensive simulations, allow more extensive parameters exploration, and enable new, previously unfeasible computational discovery.