Chromaticity-Optimized Antenna Design and Bayesian Foreground Validation for the CANTAR Global 21 cm Experiment

TL;DR

This work develops an integrated framework to advance global 21 cm experiments by optimizing antenna chromaticity, validating foreground and ionospheric models through Bayesian posterior predictive checks, and evaluating observing sites. Using PSO-optimized antennas and beam–sky convolution with Haslam-based skies, the authors demonstrate substantial chromaticity reductions (down to $\lesssim$0.02 MHz$^{-1}$) and identify mid-latitude sites as favorable for foreground suppression, while Antarctic deployments provide calibration opportunities. A key finding is that, when tested against publicly released EDGES data, physically motivated foreground/ionospheric models are statistically consistent only without an embedded 21 cm absorption feature; models including absorption fail to produce realizations consistent with the data under reasonable noise assumptions. The authors also show that residual structure can be attributed to modest impedance-mismatch ripples that can be modeled with a two-component, exponentially tapered log-periodic template, reducing post-foreground residuals significantly, and they generate a statistically validated ensemble of foregrounds for beam–sky simulations. Overall, the paper supports a two-phase CANTAR strategy—Antarctic calibration/validation followed by mid-latitude science—coupled with a robust, statistically validated foreground framework for robust global 21 cm signal detection.

Abstract

Detecting the global 21 cm signal from the epoch of reionization remains a major observational challenge due to bright foregrounds and instrumental systematics. As part of the Colombian Antarctic Telescopes for 21 cm Absorption during Reionization (CANTAR) initiative, we present a simulation and analysis framework to evaluate antenna chromaticity, optimize instrument design, and assess site suitability for global 21 cm experiments. Using frequency-dependent beam models and Haslam-based sky maps, we compute dynamic spectra for the EDGES blade dipole and a set of dipole and novel monopole antennas optimized via particle swarm optimization. The optimized designs exhibit improved spectral smoothness compared to EDGES, particularly in the 70-120 MHz range. We also evaluate latitude-dependent sky brightness and identify mid-latitude sites (-40° to +5°) as optimal for foreground suppression. We apply Bayesian inference together with posterior predictive model validation to the publicly released EDGES data, assessing statistical consistency rather than hypothesis testing or model comparison. We find that physically motivated foreground and ionospheric models are statistically consistent with the data only when a 21 cm absorption feature is excluded. From the validated posterior, we generate a statistically validated ensemble of foreground corrections for use in beam-sky simulations. These results support a two-phase strategy for CANTAR: Antarctic deployments for calibration and testing, and future science operations at mid-latitude sites. Our framework provides a validated path toward robust foreground modeling, antenna design, and systematics control for global 21 cm signal detection.

Figures (22)

• Figure 1: Geometry of the three PSO-optimized monopole blade antennas proposed for site testing, each mounted over a metallic ground plane. All designs share the same values of $W_2$, $H_2$, and $H_3$, but differ in the width of the upper blade section $W_3$, which controls impedance bandwidth. (a) Monopole-1: baseline configuration with $W_2 = W_3$. (b) Monopole-2: wingless variant with $W_3 < W_2$, optimized for narrower high-frequency response. (c) Monopole-3: winged variant with $W_3 > W_2$, designed to enhance low-frequency coupling. These designs were selected to explore how blade asymmetry impacts chromaticity and suitability for foreground modeling in site tests.
• Figure 2: HaslamMap 408 MHz sky map extrapolated to 100 MHz assuming a spectral index $\alpha = 2.5$, in Galactic coordinates. The interpolation to each frequency serves as the foreground model in our beam–sky convolution simulations.
• Figure 3: $S_{11}$ parameter for the three PSO-optimized monopole antenna models. The frequency range of interest for global 21 cm observations is indicated, and all designs achieve reflection coefficients below $-10$ dB across the target band.
• Figure 4: ANSYS-HFSS simulated gain patterns for the proposed monopole blade antennas, with $\theta$ being the altitude, and $\phi$ the azimuth. Top panels: cuts in the $\phi = 90^\circ$ plane. Bottom panels: cuts in the $\phi = 0^\circ$ plane. These slices illustrate the angular response and beam symmetry across the sky for each design.
• Figure 5: Linear gain gradient for the monopole blade antennas. Top panels: $\phi = 90^\circ$ plane. Bottom panels: $\phi = 0^\circ$ plane. The gradients are a measure of the beam chromaticity across frequency and angle for each design.