Going Wide and Deep with Roman: The z~6-9 UV luminosity function in a Roman Deep Field

Abstract

We present a trade study of possible ultra-deep surveys with the Nancy Grace Roman Space Telescope, optimizing the depth-area-filter parameter space for high-redshift galaxy science. Using a mock galaxy catalog derived from a 2 sq. degree lightcone created using the Santa Cruz semi-analytic model and populated with over 7.6 million galaxies at 0<z<10 with M_UV < -15, with realistic clustering and synthetic photometry, we evaluate sixteen 500-hour survey configurations spanning 0.28-2 sq. degrees and four filter combinations. We demonstrate that even a single Roman pointing dramatically reduces cosmic variance compared to HST-like observations, more faithfully recovering the true UV luminosity function. For each survey configuration, we explore photometric redshift recovery, sample contamination, and measurements of the rest-UV luminosity function and non-ionizing UV luminosity density across four redshift bins at z~6-9. We find that inclusion of the R062 filter is essential for studies at z~5-6, reducing sample contamination from nearly 100% to negligible levels and recovering the bright end of the luminosity function. The F184 filter improves galaxy recovery at z>9 and is critical for stellar contamination removal at all redshifts. Based on these results, we recommend that a Roman ultra-deep survey cover at least two Roman pointings (0.56 sq. degrees) with all six filters (R062, Z087, Y106, J129, H158, F184), reducing uncertainties on the rest-UV luminosity density by factors of 2-4 relative to the deepest existing JWST programs. Building off of the Deep Tier of the High Latitude Time Domain Survey to add depth and filter coverage to existing (or planned) observations is an excellent option.

Figures (2)

• Figure 1: An illustration of the constraining power of a Roman Ultra Deep Field. In the left panel, we show the positions of sources in the full 2 degree$^2$ catalog down to the depth of the HUDF ($m_{\mathrm{F160W}} \leq 29.5$) and in the redshift range $6.5 < z < 7.5$. The middle panel zooms in on a region that is $\sim0\hbox{$.\!\!^\circ$}85 \times 0\hbox{$.\!\!^\circ$}475$. The sources that would be observed in two HST-sized pointings are shaded orange, while the sources observed by a single Roman pointing are in purple. In the right panels, we show the luminosity functions as measured in each of the pointings. The two HST/WFC3 pointings cover regions of relative over- and underdensity, and as a result over- and underestimate the underlying luminosity function. However, the Roman-sized pointing is large enough to marginalize over the local density variations, and the resulting measurement of the luminosity function recovers the true value.
• Figure 2: The unscattered UV luminosity function measured in the redshift range $6.7 < z < 7.7$ for 150 randomly-selected single HUDFs ($m_{\mathrm{F160W}} \leq 29.5$, light orange), 150 HST deep field surveys ('HFF+', eight pointings with $m_{\mathrm{F160W}} \leq 29.0$ plus one with $m_{\mathrm{F160W}} \leq 29.5$, darker orange), and four Roman fields ($m_{\mathrm{F160W}} \leq 29.0$ purple). Left: The shaded regions show the 68% range of the binned number densities for both HST surveys and the 100% range for the four Roman fields. The binned number densities measured in the HFF (orange circles) and Roman field (purple squares) are plotted slightly offset from their magnitude bin centers for clarity. The fit to the full 2 degree$^2$ lightcone ($m_{\mathrm{F160W}} \leq 30.0$) is shown as the white curve and is taken as "truth." Right: We show the distribution of Schechter function parameters fit to each set of binned number densities, including $\phi^*$ (top left), $M^*$ (top right) and $\alpha$ (bottom left). Finally, we show the cumulative number of bright ($M_{1500} < -21$) galaxies selected in each pointing or set of pointings. In these panels, the Roman, 150 single HUDFs, and 150 sets of HFF+ pointings are shown as the filled purple hatched, light orange, and dark orange histograms, respectively. The value from the fit to the full lightcone is plotted as a vertical white line. The distributions are normalized by their maximum value for easy comparison. The rest-UV luminosity function measured in either a single HUDF-sized survey or a survey consisting of nine HST-sized pointings can vary significantly from the "truth," resulting in a range of Schechter function fits (including faint end slope measurements spanning $>1$) and calculated number of galaxies at the bright end. In a similar amount of time, a Roman survey could reach the depth of the HUDF while covering an area large enough to marginalize over density variations and recover the "true" luminosity function.