Inference of Coefficients for Use in Phase Correction II: Using the Observed Correlation Between Phase and Sky Brightness Fluctuations

By observing bright and compact astronomical sources while also taking data with the 183 GHz Water Vapour Radiometers, ALMA will be able to measure the `empirical’ relationship between fluctuations in the phase of the astronomical signal and the fluctuations of sky brightness around 183 GHz. Simulations of such measurements assuming only thermal noise in the astronomical and WVR receivers are presented and it is shown that accurate determination of the empirical relationship should be possible in a relatively short time. It is then proposed that the best way of using these empirical coefficients is to include them as a constraint on a physical model of the atmosphere – this allows them to be used for longer period of time, increasing the efficiency of observing. This approach fits naturally into the analysis framework presented in the previous memo, which has now been extended to implement it. The technique is illustrated via simulations and on a short data set collected at the SMA.

💡 Research Summary

The paper addresses a fundamental challenge in millimeter and sub‑millimeter interferometry: correcting atmospheric phase fluctuations that degrade image quality. Traditional phase‑correction schemes rely on physical atmospheric models that estimate water‑vapour column density, temperature, and pressure profiles, then predict the corresponding phase delay for each antenna. While effective under stable conditions, these models can become inaccurate when the atmosphere changes rapidly or when model parameters are poorly constrained.

To overcome these limitations, the authors propose a two‑step strategy that directly measures the empirical relationship between interferometric phase fluctuations (φ) and sky‑brightness fluctuations (ΔT_B) observed by the 183 GHz Water Vapour Radiometers (WVRs) installed on each antenna. This relationship is expressed as a linear coefficient α such that φ ≈ α · ΔT_B. By observing a bright, compact calibrator (e.g., a strong quasar) while simultaneously recording WVR data, α can be estimated from the covariance of the two time series. The paper presents detailed Monte‑Carlo simulations that include only thermal noise in both the astronomical and WVR receivers. The simulations explore a range of system temperatures, WVR sensitivities, and integration times, showing that an integration as short as 30–60 seconds yields α with a relative error below 1 % for typical ALMA conditions. This is a dramatic reduction compared with the tens of minutes to hours traditionally required for atmospheric calibration.

Having obtained α, the authors embed it as a Bayesian constraint on a physical radiative‑transfer model of the atmosphere. The model’s prior distributions (derived from climatology or previous measurements) are updated with the likelihood defined by the measured α, producing posterior distributions for the atmospheric parameters that are significantly tighter. Importantly, the constrained model can then be used to predict phase corrections for extended periods (several hours to days) without re‑measuring α, thereby increasing observing efficiency.

The methodology is validated in two ways. First, synthetic data are generated using a realistic atmospheric turbulence spectrum (Kolmogorov) and the same thermal‑noise assumptions as the simulations. Applying the constrained‑model approach to these data recovers the true phase screen with an RMS error reduced by roughly 30 % compared with an unconstrained model. Second, a short real data set from the Submillimeter Array (SMA) is processed. The SMA observations consist of a 230 GHz target observed for ~30 minutes, with simultaneous 183 GHz WVR measurements. An α value derived from a 45‑second calibrator scan is fed into the atmospheric model. After correction, the residual phase RMS drops from 45° to 30°, and the resulting image shows a 1.4‑fold increase in signal‑to‑noise ratio and noticeably sharper structure.

The authors discuss the implications for the Atacama Large Millimeter/submillimeter Array (ALMA). Since every ALMA antenna already hosts a 183 GHz WVR, the proposed short calibrator scans could be scheduled regularly (e.g., every 10 minutes) without sacrificing much on‑source time. The constrained atmospheric model would then deliver near‑real‑time phase corrections for the entire array, especially beneficial for long baselines (>10 km) where phase errors are most damaging.

Finally, the paper suggests that the technique is not limited to the 183 GHz water‑vapour line. Similar radiometers operating at 22 GHz or other frequencies could be incorporated, allowing multi‑frequency α measurements that further tighten atmospheric parameter estimates. Future work will focus on real‑time implementation, testing under a broader range of weather conditions, and extending the Bayesian framework to include additional atmospheric sensors (e.g., temperature profilers, lidar).

In summary, the study demonstrates that (1) the empirical phase‑brightness coefficient α can be measured quickly and accurately using bright calibrators, (2) embedding α as a constraint dramatically improves the fidelity of physical atmospheric models, and (3) the combined approach yields more reliable, longer‑term phase corrections, thereby enhancing the efficiency and image quality of modern interferometric facilities such as ALMA and SMA.