Visibility stacking in the quest for SNIa radio emission

We describe the process of stacking radio interferometry visibilities to form a deep composite image and its application to the observation of transient phenomena. We apply “visibility stacking” to 46 archival Very Large Array observations of nearby type Ia supernovae (SNeIa). This new approach provides an upper limit on the SNIa ensemble peak radio luminosity of 1.2x10^{25}erg/s/Hz at 5GHz, which is 5-10 times lower than previously measured. This luminosity implies an upper limit on the average companion stellar wind mass loss rate of 1.3x10^{-7}M_o/yr. This mass loss rate is consistent with the double degenerate scenario for SNeIa and rules out intermediate and high mass companions in the single degenerate scenario. In the era of time domain astronomy, techniques such as visibility stacking will be important in extracting the maximum amount of information from observations of populations of short lived events.

💡 Research Summary

This paper introduces a novel technique—visibility stacking—to combine radio interferometric data at the raw visibility level rather than after imaging, thereby overcoming the limitations imposed by incomplete (u,v) coverage and complex noise statistics inherent in traditional image stacking. The authors apply this method to a comprehensive set of archival Very Large Array (VLA) observations of nearby Type Ia supernovae (SNe Ia) in order to place the most stringent radio luminosity upper limits to date on this class of transients.

Motivation and Context
Type Ia supernovae are critical cosmological distance indicators, yet their progenitor systems remain debated between the single‑degenerate (SD) scenario, involving a white dwarf accreting from a non‑degenerate companion, and the double‑degenerate (DD) scenario, where two white dwarfs merge. The density of circumstellar material (CSM) surrounding the explosion directly reflects the mass‑loss history of the progenitor system; interaction of the SN ejecta with this CSM produces synchrotron radio emission. Detecting—or constraining—this radio emission therefore provides a powerful diagnostic of the progenitor’s mass‑loss rate (Ṁ). Prior radio searches have yielded non‑detections with relatively high upper limits, insufficient to discriminate between SD and DD models.

Visibility Stacking Methodology
The authors first review the statistical basis of stacking: for N observations each of duration t and rms σ, a simple inverse‑variance weighting yields an rms improvement of σ/√N, equivalent to a single observation of total time N·t. However, in radio interferometry the incomplete and non‑uniform sampling of the (u,v) plane introduces spatial filtering that can bias image‑domain stacking. By stacking directly in the visibility domain, one can combine complementary (u,v) coverages, improve sampling, and retain the full sensitivity gain. The procedure involves (1) calibrating each dataset, (2) subtracting contaminating background sources from the visibilities, (3) shifting the phase centre of each visibility set to the precise coordinates of the target SN, and (4) concatenating all visibilities before imaging. Weighting is naturally applied during imaging, preserving the optimal sensitivity.

Data Selection and Reduction
The study draws on 46 usable VLA observations of 27 nearby (z < 0.1) SNe Ia collected by Panagia et al. (2006) between 1981 and 2003. The observations span all VLA configurations (A, B, C, D, and hybrids) and multiple frequencies; the authors restrict the analysis to the 6 cm (≈5 GHz) band, which accounts for 75 % of the total data. After discarding observations with missing data, calibration failures, gain errors, or severe field confusion (e.g., those containing the bright source Centaurus A), the final sample comprises 46 epochs, each providing one or two 50 MHz sub‑bands (4.88 GHz and 4.835 GHz). Calibration employed the standard primary calibrator 3C286, with flux stability better than 1 %.

Stack Construction and Sensitivity
Three stacked images are produced: (i) “All” – using every usable epoch, (ii) “Early” – only epochs with SN age ≤ 100 days, and (iii) “Late” – epochs with age ≥ 100 days. The effective integration time for the “All” stack corresponds to 42.1 hours of 50 MHz bandwidth, yielding a 3σ rms of 13.3 µJy beam⁻¹. The “Early” and “Late” stacks achieve 3σ rms of 16.7 µJy beam⁻¹ and 19.2 µJy beam⁻¹, respectively. No emission is detected at the SN positions in any stack, establishing a 3σ upper limit on the ensemble peak radio luminosity of Lν ≤ 1.2 × 10²⁵ erg s⁻¹ Hz⁻¹ at 5 GHz. This limit is 5–10 times lower than previous ensemble constraints.

Conversion to Mass‑Loss Rate
To translate the luminosity limit into a constraint on the progenitor wind, the authors adopt the standard synchrotron model used for core‑collapse SNe (Chevalier 1998) with the scaling Lν ∝ (Ṁ/w) t⁻¹ e⁻τ, where w is the wind velocity. Using the dimensionless constant Λ = 1285 ± 245 derived by Panagia et al. (2006) for SNe Ia, assuming a wind speed w = 10 km s⁻¹ (characteristic of a post‑main‑sequence companion), and setting the optical depth τ = 1 (the epoch of peak radio emission), the authors infer an ensemble average mass‑loss rate of Ṁ ≤ 1.3 × 10⁻⁷ M⊙ yr⁻¹.

Scientific Implications
The derived Ṁ limit is an order of magnitude below the rates expected from a red‑giant or sub‑giant companion in the SD scenario (Ṁ ≈ 10⁻⁶–10⁻⁵ M⊙ yr⁻¹). Consequently, the results strongly disfavour SD models that invoke intermediate‑ or high‑mass companions, while remaining fully consistent with the DD scenario, which predicts negligible CSM and thus negligible radio emission. This statistical non‑detection therefore adds weight to the hypothesis that a substantial fraction of SNe Ia arise from double white‑dwarf mergers.

Methodological Significance and Future Prospects
Beyond the astrophysical conclusions, the paper demonstrates the power of visibility stacking as a general tool for time‑domain radio astronomy. By improving (u,v) coverage and achieving deep sensitivities without sacrificing spatial information, the technique is well suited to upcoming large‑scale surveys with ASKAP, MeerKAT, and the Square Kilometre Array, where massive numbers of short‑lived transients will be observed. Visibility stacking can be extended to other transient classes (e.g., gamma‑ray burst afterglows, fast radio bursts) and to multi‑epoch monitoring campaigns, enabling population‑level constraints that would be impossible from individual detections alone.

In summary, the authors have set the most stringent radio luminosity and mass‑loss limits for Type Ia supernovae to date by pioneering a visibility‑stacking approach, thereby providing compelling evidence in favor of the double‑degenerate progenitor channel and establishing a methodological framework that will be invaluable for the rapidly expanding field of time‑domain radio astronomy.