[Paper Review] The LOFAR Two Meter Sky Survey: Deep Fields, I -- Direction-dependent calibration and imaging

C. Tasse, T. W. Shimwell+30 more|Durham Research Online (Durham University)|Nov 16, 2020

Radio Astronomy Observations and TechnologyPhysics and Astronomy77 references40 citations

TL;DR

The paper presents a direction-dependent calibration and imaging pipeline (ddf-pipeline-v2) for LOFAR High Band Antenna data to produce deep, high-fidelity, thermal-noise-limited images of Boötes and Lockman Hole fields at ~150 MHz, achieving unprecedented low noise levels over deep integrations.

ABSTRACT

The Low Frequency Array (LOFAR) is an ideal instrument to conduct deep extragalactic surveys. It has a large field of view and is sensitive to large scale and compact emission. It is, however, very challenging to synthesize thermal noise limited maps at full resolution, mainly because of the complexity of the low-frequency sky and the direction dependent effects (phased array beams and ionosphere). In this first paper of a series we present a new calibration and imaging pipeline that aims at producing high fidelity, high dynamic range images with LOFAR High Band Antenna data, while being computationally efficient and robust against the absorption of unmodeled radio emission. We apply this calibration and imaging strategy to synthesize deep images of the Bootes and LH fields at 150 MHz, totaling $\sim80$ and $\sim100$ hours of integration respectively and reaching unprecedented noise levels at these low frequencies of $\lesssim30$ and $\lesssim23$ $μ$Jy/beam in the inner $\sim3$ deg$^2$. This approach is also being used to reduce the LoTSS-wide data for the second data release.

Motivation & Objective

• Motivate the need for deep LOFAR extragalactic surveys at low frequencies and explain direction-dependent calibration challenges.
• Describe the mathematical and algorithmic framework for third-generation calibration and imaging (RIME, ddC-rime, ddI-rime).
• Introduce and validate the ddf-pipeline-v2 to produce high dynamic range, thermal-noise-limited images.
• Apply the pipeline to Boötes and Lockman Hole fields and report resulting image quality and robustness improvements.

Proposed method

• Present Radio Interferometry Measurement Equation (RIME) formalism with direction-independent (G) and direction-dependent (J) Jones/Mueller matrices.
• Describe two-step calibration and imaging (ddC-rime and ddI-rime) and the concept of dd-self-calibration.
• Use a faceted, Jones-based approach with multi-directional solvers (kMS, DDFacet, and killMS) embedded in the ddf-pipeline-v2.
• Implement dynamic range improvement strategies to handle bright source artefacts and unmodeled flux absorption.
• Iteratively solve for sky model x and Jones terms with self-calibration in a cycle, incorporating clustering, bootstrapping, and smoothing steps.
• Demonstrate improved stability and robustness over the previous ddf-pipeline-v1.

Experimental results

Research questions

• RQ1 How can third-generation calibration and imaging (direction-dependent) be integrated to produce thermal-noise-limited LOFAR images at ~150 MHz?
• RQ2 What are the benefits and limitations of a facet-based, Jones-based dd-calibration approach for deep LOFAR fields?
• RQ3 Can the ddf-pipeline-v2 deliver robust imaging with reduced artefacts and unmodeled flux absorption compared to earlier pipelines?
• RQ4 What are the noise levels and dynamic range achievable in deep fields like Boötes and Lockman Hole with long integrations?
• RQ5 How do direction-dependent effects (ionosphere, beam shapes) impact high-fidelity sky reconstruction at low frequencies?

Key findings

• Achieved noise levels of ≲30 μJy/beam and ≲23 μJy/beam in Boötes and Lockman Hole inner ~3 deg^2 with ~80–100 hours of integration.
• Demonstrated that a direction-dependent calibration and imaging pipeline can produce thermal-noise-limited maps from LOFAR HBA data.
• Showed improved robustness against artifacts around bright sources and reduced unmodeled flux absorption using the ddf-pipeline-v2.
• Compared to ddf-pipeline-v1, dd-calibration and imaging with kMS/ DDFacet in a unified pipeline yields better image quality and reduced artefacts (as per Sec. 3).
• Provided a scalable, I/O-efficient workflow that handles direction-dependent effects across multiple facets and frequency-time domains.