Radio antenna design for sky-averaged 21 cm cosmology experiments: the REACH case

J. Cumner, Eloy de Lera Acedo, Dirk I. L. de Villiers, D. Anstey, C. I. Kolitsidas, B. Gurdon, N. Fagnoni, P. Alexander, G. Bernardi, H. T. J. Bevins, S. Carey, J. Cavillot, R. Chiello, C. Craeye, W. Croukamp, J. A. Ely, A. Fialkov, T. Gessey-Jones, Q. Gueuning, Will Handley, R. Hills, A. T. Josaitis, G. Kulkarni, A. Magro, R. Maiolino, P. Daniel Meerburg, Shikhar Mittal, Jonathan R. Pritchard, E. Puchwein, N. Razavi-Ghods, I. L. V. Roque, A. Saxena, K. H. Scheutwinkel, E. Shen, Peter Sims, O. Smirnov, M. Spinelli, K. Zarb-Adami

TL;DR

The REACH (Radio Experiment for the Analysis of Cosmic Hydrogen) team has developed a rigorous, figure-of-merit-based framework to design a wideband dipole antenna for 21 cm cosmology. By prioritizing spectral smoothness and "modelability" over raw performance metrics, they produced a 2.5:1 bandwidth antenna capable of distinguishing the faint signal of the first stars from overwhelming galactic noise.

What Problem Does This Paper Address?

Detecting the "Cosmic Dawn"—the era when the first stars ignited—relies on finding a tiny absorption feature in the sky-averaged 21 cm radio signal. This signal is incredibly faint (~100 mK), while the foreground radiation from our own galaxy is five orders of magnitude brighter (~10,000 K).

The primary technical hurdle is chromaticity: an antenna’s beam shape and impedance change with frequency. If these changes aren't perfectly smooth or perfectly understood, they can create "wiggles" in the data that look exactly like the cosmological signal researchers are searching for. Following the controversial 2018 detection by the EDGES experiment, the REACH project aims for a more robust detection by designing hardware that is specifically optimized for high-fidelity modeling and calibration.

Methods

The authors moved away from traditional "intuition-based" antenna design toward a quantitative, iterative refinement process:

  1. Multi-Objective Optimization: They defined five specific Figures of Merit (FoM) to evaluate designs:
  2. $\sigma$ (Beam Smoothness): How much the beam pattern varies with frequency when coupled with a sky model.
  3. $\alpha$ (Impedance Match): The mean reflection coefficient ($S_{11}$).
  4. $\beta$ (Bandwidth): Ensuring the antenna operates across the critical 50–130 MHz range.
  5. $\gamma$ (Reflection Flatness): Ensuring the $S_{11}$ curve doesn't have sharp, unmodelable features.
  6. $\delta$ (Minimum Frequency): Keeping the operational floor at 50 MHz.
  7. Sigmoid Scaling: To prevent the optimizer from over-improving one metric (like impedance) at the expense of another (like beam smoothness), they used sigmoid functions to "level off" gains once a parameter reached an acceptable threshold.
  8. Mock Detection Pipeline: For the final candidates, the authors didn't just look at antenna plots; they injected simulated 21 cm signals into the antenna models and ran them through a full Bayesian data analysis pipeline to see if the signal could be successfully recovered.
  9. Simulation vs. Reality: They used CST Microwave Studio for EM simulations and validated the results by building a physical prototype and measuring its impedance at Stellenbosch University.

Datasets

  • Sky Models: The team used the Haslam 408 MHz map (both the unfiltered and the "destriped/desourced" DSDS versions) and the GSM(200) map to simulate how the antenna would "see" the galaxy.
  • Spectral Index Maps: They utilized angularly dependent spectral index maps to account for the fact that different parts of the galaxy have different "colors" in the radio spectrum.
  • Simulated Signals: Gaussian profiles representing the expected 21 cm signal were used to test the mock detection pipeline.

Key Insights and Ideas

  • Modelability > Perfection: The authors argue that it is better to have an antenna with a slightly higher reflection coefficient that is easy to simulate accurately than a "perfect" antenna with a complex physical structure that introduces simulation errors.
  • The Ground Plane is Part of the Antenna: The paper highlights how the ground plane (the metal sheet under the antenna) and the balun (the feed structure) significantly impact the beam's chromaticity. Simulating just the antenna "blades" is insufficient for 21 cm cosmology.
  • The "Aha" Moment on Beam Resolution: While high-resolution (narrow beam) antennas might seem better for avoiding the galactic center, they are often more chromatic due to their physical complexity. REACH opted for a wide-beam dipole because its simplicity makes the systematic errors easier to calibrate out.

Key Results

  • Successful Design: The process yielded a hexagonal-bladed dipole with a 2.5:1 frequency bandwidth (roughly 50–135 MHz).
  • Impedance Performance: The design achieved a mean reflection coefficient ($\alpha$) of -10 dB, balancing power transfer with spectral smoothness.
  • Pipeline Validation: In the mock pipeline, the antenna successfully identified injected signals at 80, 100, and 120 MHz with a 2.5:1 odds confidence of rejecting a null (no-signal) hypothesis.
  • Experimental Agreement: Prototype measurements showed that the residual difference between the simulated and measured impedance was generally within 10 $\Omega$, confirming that their computational models are accurate enough for field use.

Limitations and Future Work

  • Computational Cost: The "mock detection pipeline" is extremely slow, taking over 7 days to evaluate a single design, which limits its use to the final stages of the design process.
  • Ground Plane Scaling: Most simulations used a 5m or 10m ground plane for computational efficiency, but the final deployment will require even larger structures to minimize edge reflections.
  • Phase II Antennas: This paper focuses on the dipole; however, REACH Phase I also intends to use a conical log-spiral antenna, which will require a similar rigorous design process to handle its increased physical complexity.

Summary updated

Abstract

Following the reported detection of an absorption profile associated with the 21~cm sky-averaged signal from the Cosmic Dawn by the EDGES experiment in 2018, a number of experiments have been set up to verify this result. This paper discusses the design process used for global 21~cm experiments, focusing specifically on the Radio Experiment for the Analysis of Cosmic Hydrogen (REACH). This experiment will seek to understand and compensate for systematic errors present using detailed modelling and characterization of the instrumentation. There is detailed the quantitative figures of merit and numerical modelling used to assist the design process of the REACH dipole antenna (one of the 2 antenna designs for REACH Phase I). This design process produced a 2.5:1 frequency bandwidth dipole. The aim of this design was to balance spectral smoothness and low impedance reflections with the ability to describe and understand the antenna response to the sky signal to inform the critically important calibration during observation and data analysis.
BibTeX
@article{cumner2021radio,
  author = {J. Cumner and Eloy de Lera Acedo and Dirk I. L. de Villiers and D. Anstey and C. I. Kolitsidas and B. Gurdon and N. Fagnoni and P. Alexander and G. Bernardi and H. T. J. Bevins and S. Carey and J. Cavillot and R. Chiello and C. Craeye and W. Croukamp and J. A. Ely and A. Fialkov and T. Gessey-Jones and Q. Gueuning and Will Handley and R. Hills and A. T. Josaitis and G. Kulkarni and A. Magro and R. Maiolino and P. Daniel Meerburg and Shikhar Mittal and Jonathan R. Pritchard and E. Puchwein and N. Razavi-Ghods and I. L. V. Roque and A. Saxena and K. H. Scheutwinkel and E. Shen and Peter Sims and O. Smirnov and M. Spinelli and K. Zarb-Adami},
  title = {{Radio antenna design for sky-averaged 21 cm cosmology experiments: the REACH case}},
  journal = {Journal of Astronomical Instrumentation},
  year = {2021},
  doi = {10.1142/s2251171722500015},
}
Figure