Use of Time Dependent Data in Bayesian Global 21cm Foreground and Signal Modelling

Detecting the faint, sky-averaged 21cm signal from the Cosmic Dawn and the Epoch of Reionisation is a primary objective for modern cosmology, but it is an immense observational challenge. In our new paper, “Use of Time Dependent Data in Bayesian Global 21cm Foreground and Signal Modelling,” we present a novel data analysis strategy to significantly improve our ability to extract this cosmological signal from overwhelming instrumental and astrophysical noise.

The Challenge: Disentangling Signal from Systematics

The core difficulty in global 21cm experiments is that the cosmological signal is swamped by astrophysical foregrounds—predominantly synchrotron and free-free emission from our own galaxy—which are several orders of magnitude brighter. While these foregrounds are spectrally smooth, the problem is compounded by the instrument’s chromaticity: the antenna’s directional sensitivity (or beam) changes with frequency. This effect couples the bright, spatially varying foregrounds into the data, creating complex, non-smooth distortions that can easily mimic or obscure the true 21cm signal.

A New Approach: Leveraging Time-Variation

This work, led by [Dominic Anstey] in collaboration with Eloy de Lera Acedo and Will Handley, introduces a powerful Bayesian framework to tackle this issue. Instead of averaging out the time-variation of the sky as seen by the telescope, our method leverages it as a source of information. The key insight is that while the observed sky changes due to the Earth’s rotation, the underlying physical properties of the foregrounds are static.

By performing a single, joint Bayesian analysis on data from multiple distinct observation times, our model can more effectively disentangle the constant sky emission from the time-varying instrumental response. This approach builds upon our group’s physically-motivated foreground modelling framework, presented in 10.1093/mnras/stab1765, which parameterises the sky based on its intrinsic spectral index.

Key Findings and Impact

We tested this “time-separated” fitting technique against the traditional method of fitting a “time-integrated” (averaged) dataset. Our findings, derived from detailed simulations, highlight several key advantages:

• Improved Signal Recovery: For antennas with significant and complex chromaticity (like a hexagonal dipole), the time-separated method yields a dramatic improvement in the accuracy and precision of the recovered 21cm signal.
• Enhanced Foreground Modelling: The technique also produces a far more accurate model of the foregrounds themselves, improving their reconstruction by a factor of ~2-3. This demonstrates that the method is not just subtracting systematics but is gaining a better physical understanding of the sky.
• Instrument-Dependent Gains: The benefit for signal recovery was most pronounced for the highly chromatic hexagonal dipole. For a less chromatic log-spiral antenna, the improvement was more modest, suggesting the technique is especially powerful for mitigating complex instrumental effects. This result is consistent with work from other groups which also found benefits in using time-variable data, as noted in analyses like 10.3847/1538-4357/ab9b2a.
• Multi-Antenna Synergy: The paper proposes a powerful extension: simultaneously fitting data from different antennas. This approach leverages the distinct chromatic responses of each instrument to break degeneracies even more effectively, leading to superior signal and foreground recovery than using time-variation from a single antenna alone.

In conclusion, this research provides a robust and physically-grounded pathway for global 21cm experiments to overcome the immense challenge of foreground subtraction. By treating time-variation as a source of information, our method significantly enhances our ability to detect the faint whisper of the Cosmic Dawn and, in parallel, provides a more detailed map of the radio sky.

Content generated by gemini-2.5-pro using this prompt.

Image generated by imagen-3.0-generate-002 using this prompt.