Characterizing the Roman Grism Redshift Efficiency of Type Ia Supernova Host Galaxies for the High-Latitude Time-Domain Survey

TL;DR

This study evaluates the utility of the Roman grism for obtaining redshifts of Type Ia SN host galaxies within the High-Latitude Time-Domain Survey, supplementing prism and ground-based redshifts. It uses Grizli-based 2D grism simulations on OpenUniverse images to measure host-galaxy redshift recovery, finding $m_{F158}$-dependent efficiencies with 50% at $m_{F158}=20.61$ and 90% at $m_{F158}=19.27$, then propagates these efficiencies into SNANA/SALT3-NIR catalog simulations to predict ~6800 SNe with redshifts when combining grism, prism, and ground-based channels. The work also investigates systematics from grism-efficiency modeling, exploring dependencies on host stellar mass and color, and quantifies their impact on $w_0$ and $w_a$, with mass-dependent variants producing the largest shifts. Furthermore, it analyzes how redshift-source assumptions influence the dark energy FoM across survey configurations, underscoring that optimization decisions depend on whether spectroscopic or photometric redshifts are used. Overall, the paper provides an optimistic upper bound on the grism’s contribution to Roman SN cosmology, highlights key systematic sensitivities, and suggests directions for refined modeling and survey design.

Abstract

The High-Latitude Time-Domain Survey (HLTDS) for the Nancy Grace Roman Space Telescope (Roman) will discover thousands of high redshift Type Ia supernovae (SNeIa) to make generation-defining cosmological constraints on dark energy. To construct the Roman SN Hubble diagram, a strategy to obtain redshifts must be determined. While the nominal HLTDS will use only the Roman prism, in this work we consider the utility of the Roman grism observations from overlap with the High-Latitude Wide-Area Survey for SNIa cosmology. We determine a galaxy grism redshift recovery rate by simulating dispersed grism images and measuring redshifts with the Grizli software, obtaining an $H$-band 50% redshift recovery at magnitude 20.61 and 90% recovery at magnitude 19.27. To estimate the total number of spectroscopic redshifts expected for Roman SN cosmology, we also consider a Roman prism SN redshift efficiency and a ground-based telescope redshift efficiency for host-galaxies. We apply these redshift efficiencies to SNIa catalog level simulations and predict that $\sim$6800 SNe will have a SN or host spectroscopic redshift. Second, we evaluate the size of potential systematics related to modeling the grism redshift efficiency by considering the impact of additional dependencies on stellar mass and host galaxy color. We estimate the largest potential size of this systematic to be 0.0066$\pm$0.002 and -0.0266$\pm$0.007, roughly 42.9 and 49.6% of the statistical uncertainty for $w_0$ and $w_a$ respectively. Lastly, we consider the effects of assuming different redshift sources on the HLTDS survey strategy optimization by measuring relative changes to the dark energy Figure of Merit.

Figures (9)

• Figure 1: Spectra extraction and template fitting examples for a quiescent, red galaxy (left panel) and an emission line galaxy (right panel). The blue solid lines show the optimally extracted spectra and the red solid line shows the Grizli best-fit template. The true redshift and measured redshift are shown in each panel. The 4000Å break is marked with dashed grey in the left panel, and strong emission lines are labeled in the right panel. The flux unit is in $10^{-19}$ erg/s/cm$^2$/Å.
• Figure 2: Top: Redshift efficiency from the grism image simulations as a function of $H$-band magnitude, with a logistic regression curve fitted to the data. The left y-axis indicates the fraction of redshift completeness (i.e., redshift efficiency). The background gray histogram shows the overall magnitude distribution of galaxies for which redshifts are successfully estimated, with the number of galaxies on the right y-axis. The 90% and 50% efficiencies are marked with dash-dotted and dashed lines, respectively. Middle: Redshift efficiency as a function of redshift, with a background gray histogram of galaxy redshifts. No fit is provided as the curve is illustrative and not used as an input to any SN simulations. Bottom: Heat map of redshift completeness in $H$-band magnitude and redshift space.
• Figure 3: Left: Post light-curve fit SN redshift distributions for simulations applying a grism efficiency (blue), a ground-based efficiency (orange), a prism efficiency (green), and all three together (black). Right: Overall efficiency for SNe, defined as the number of SNe post light-curve fit for a simulation given a particular redshift efficiency divided by the number of SNe post light-curve fit for a simulation with no redshift efficiency. Note that the prism efficiency is scaled down by 0.25 as described in Section \ref{['sec:prismeff']}.
• Figure 4: Top: Same as top of Figure \ref{['fig:zeff']} but for galaxies with logMass $<10$ in blue and galaxies with logMass $\geq10$ in magenta. Bottom: Same as center of Figure \ref{['fig:zeff']} but for galaxies with logMass $<10$ in blue and galaxies with logMass $\geq10$ in magenta.
• Figure 5: Top: Same as top of Figure \ref{['fig:zeff']} but for galaxies with g-r color $<0.95$ in blue and galaxies with g-r color $\geq0.95$ in red. Bottom: Same as center of Figure \ref{['fig:zeff_color']} but for galaxies with g-r color $<0.95$ in blue and galaxies with g-r color $\geq0.95$ in red.