[Paper Review] Molecules with ALMA at Planet-forming Scales (MAPS) II: CLEAN Strategies for Synthesizing Images of Molecular Line Emission in Protoplanetary Disks

Ian Czekala, Ryan A. Loomis+28 more|arXiv (Cornell University)|Sep 13, 2021

Astrophysics and Star Formation Studies62 references121 citations

TL;DR

This paper presents a validated, multi-stage imaging workflow for synthesizing high-fidelity position-position-velocity cube images of molecular line emission in protoplanetary disks using ALMA data. It introduces the 'JvM correction'—a critical flux calibration fix for CLEAN deconvolution that corrects for inaccurate residual map scaling, significantly improving flux accuracy in low signal-to-noise regimes, and provides best-practice recommendations for visibility tapering, CLEAN masking, and robust weighting to ensure reliable, reproducible image cubes across diverse molecular transitions.

ABSTRACT

The Molecules with ALMA at Planet-forming Scales large program (MAPS LP) surveyed the chemical structures of five protoplanetary disks across more than 40 different spectral lines at high angular resolution (0.15" and 0.30" beams for Bands 6 and 3, respectively) and sensitivity (spanning 0.3 - 1.3 mJy/beam and 0.4 - 1.9 mJy/beam for Bands 6 and 3, respectively). In this article, we describe our multi-stage workflow -- built around the CASA tclean image deconvolution procedure -- that we used to generate the core data product of the MAPS LP: the position-position-velocity image cubes for each spectral line. Owing to the expansive nature of the survey, we encountered a range of imaging challenges; some are familiar to the sub-mm protoplanetary disk community, like the benefits of using an accurate CLEAN mask, and others less well-known, like the incorrect default flux scaling of the CLEAN residual map first described in Jorsater & van Moorsel 1995 (the "JvM effect"). We distill lessons learned into recommended workflows for synthesizing image cubes of molecular emission. In particular, we describe how to produce image cubes with accurate fluxes via the "JvM correction," a procedure that is generally applicable to any image synthesized via CLEAN deconvolution but is especially critical for low S/N emission. We further explain how we used visibility tapering to promote a common, fiducial beam size and contextualize the interpretation of signal to noise ratio when detecting molecular emission from protoplanetary disks. This paper is part of the MAPS special issue of the Astrophysical Journal Supplement.

Motivation & Objective

• To address imaging challenges in high-dynamic-range, low-signal-to-noise molecular line observations of protoplanetary disks using ALMA.
• To identify and correct systematic flux errors in CLEAN deconvolution, particularly the 'JvM effect'—an incorrect default residual map scaling that distorts fluxes.
• To establish a standardized, reproducible workflow for generating accurate, high-resolution image cubes from ALMA visibility data across multiple spectral lines.
• To enable reliable scientific interpretation of molecular emission by ensuring consistent beam sizes and correct flux calibration through visibility tapering and robust weighting.
• To provide a reference framework for future ALMA surveys requiring high-fidelity, flux-calibrated image cubes of molecular emission.

Proposed method

• Employed the CASA tclean algorithm as the core deconvolution engine for synthesizing image cubes from calibrated ALMA visibility data.
• Applied multi-scale CLEAN with scales=[0, 5, 15, 25] pixels, where pixel size was ≈1/7th of the beam FWHM, to handle extended and compact emission simultaneously.
• Used Keplerian CLEAN masks matched to 13CO J=2–1 emission for all lines except CO, improving dynamic range and reducing artifacts.
• Iterated the CLEAN process until peak residual emission dropped below 4 × RMS threshold, ensuring convergence.
• Applied Briggs weighting with robust=0.5 for untapered beams and forward-modeled uvtaper to achieve target beam size using the largest robust value ≤0.5.
• Implemented the 'JvM correction' by computing the JvM factor ϵ as the ratio of CLEAN beam volume to dirty beam volume, scaling the residual map by ϵ before combining with the convolved model to produce flux-accurate final images.

Experimental results

Research questions

• RQ1 How can CLEAN deconvolution be optimized to produce accurate, flux-calibrated image cubes of molecular emission in protoplanetary disks with low signal-to-noise?
• RQ2 What is the impact of the 'JvM effect'—the incorrect default scaling of the CLEAN residual map—on flux measurements in ALMA imaging?
• RQ3 How can visibility tapering be used effectively to achieve a common, fiducial beam size across multiple lines and observations while preserving dynamic range?
• RQ4 What are the optimal CLEAN mask strategies for molecular line emission in disks with complex kinematic structures?
• RQ5 How can the imaging workflow be standardized and made reproducible across a large survey like MAPS with over 40 spectral lines?

Key findings

• The 'JvM correction'—a previously undocumented flux scaling error in CLEAN deconvolution—was identified and quantified, with the JvM factor ϵ defined as the ratio of CLEAN beam volume to dirty beam volume.
• Applying the JvM correction significantly improves flux accuracy in low signal-to-noise molecular line emission, especially critical for faint transitions like N2H+ and SO.
• Visibility tapering was successfully used to achieve a common, fiducial beam size across all lines, enabling consistent comparison and analysis across the survey.
• The use of Keplerian CLEAN masks matched to 13CO J=2–1 emission improved dynamic range and reduced artifacts in the final image cubes.
• The multi-stage workflow, including multi-scale CLEAN, robust weighting, and JvM correction, produced image cubes with reliable fluxes and minimal artifacts, forming the core data product of the MAPS LP.
• The recommended workflow is generally applicable to any CLEAN-based imaging of molecular emission and is especially critical for low signal-to-noise data.