Extended Data Fig. 4: Resilience to systematic signals.

Top-left: Plot of the recovered 21-cm signal in purple, compared to the true inserted 21-cm signal in green, for simulated data sets of a log spiral and hexagonal dipole antenna. Each data set consisted of three time bins, 20 minutes apart. The lower plots show the results of fitting an integration of the three bins to a single foreground model and the upper plots show the results of fitting the separate bins jointly to corresponding models in a single fit. The rightmost plots show the results of fitting the data sets from both antenna simultaneously in the same fit. Top-right: Plot of the optimum numbers of foreground regions, determined using the Bayesian evidence, for the model fits shown in the top-left plot. Bottom-left: Plot showing a run of the pipeline where the antenna model included the presence of the finite 20x20 m metallic ground plane underneath the spiral antenna. This plot shows that the chromaticity introduced by reflections at the edge of the REACH ground plane, if properly modelled, would not severely affect the ability of the pipeline to recover the cosmological signal. Bottom-right: Plot showing the result of running the data pipeline when a sinusoidal systematic arising from the presence of the 6 m cable connecting the spiral antenna feed point to the receiver has been introduced in the data. The additive systematic signal is shown as a black-solid line in this plot. In the simulated analysis we included a sinusoidal model to fit for this systematic signal simultaneously with the foregrounds and the cosmological signal. This result shows that a detection of the true signal could be achieved in this case.