The Sun is the source with the highest flux density at GHz frequencies and has a large angular size ( ≥ 3 2 ′ ) . The full-width half-maximum of the primary beam of MeerKAT is about 2 ° and 1.2 ° at the central frequency of UHF- and L-band, respectively (de Villiers, 2023). Hence, the Sun fills a significant portion of the primary beam of MeerKAT at UHF and L-band. The Sun fills the entire beam at the higher frequency part of S-bands. This results in a significantly higher beam-integrated power compared to typical astronomical sources, necessitating strong attenuation to ensure that the signal chain of the telescope remains within its optimal regime. However, this creates a calibration challenge. The quiet Sun has a flux density of a few hundred Solar Flux Unit (sfu) [1 sfu = 1 0 4 Jansky (Jy)], increasing with frequency due to its thermal nature and reaching several thousand sfu due to non-thermal emissions during the presence of solar activity. In contrast, all bright astronomical calibrators (Perley and Butler, 2017) as well as the A-team sources (de Gasperin et al., 2020) have negative spectral indices, with flux densities that decrease with frequency. As a result, they cannot be observed with the same attenuation settings as the Sun. This makes traditional flux density calibration using astronomical calibrators infeasible for solar observations with MeerKAT, necessitating an independent method to characterize the spectral response of the attenuators. To prepare MeerKAT for solar observations, we have performed several engineering tests between late 2022 and late 2023 under the project ID: EXT-20221114-PK-01.

In general, the low-noise amplifier (LNA) – the first component in the signal chain of a radio telescope–is designed to have a linear response over a wide dynamic range, enabling it to accommodate strong signals such as those from the Sun. However, downstream sub-systems, including those of MeerKAT, have more limited dynamic ranges, and the default configuration is optimized for observing faint astronomical sources. To manage strong solar signals, MeerKAT employs attenuators within the Radio Frequency Conditioning Unit (RFCU). This is a room-temperature subsystem located just before the analog-to-digital converter (ADC) in the signal chain. These attenuators offer 0–63 dB attenuation in 1 dB steps. To calibrate this attenuation, we have used a built-in noise diode. The built-in noise diode injects noise with a temperature approximately equal to system temperature on cold sky, leading to an increase in power by approximately 3 dB. Hence, measuring the change corresponding to the power injected by the noise diode when using attenuators allows us to measure the effective attenuation.

We estimated the additional signal attenuation required for MeerKAT solar observations to maintain the input power to the ADCs near the nominal level, based on source flux density and the band-averaged System Equivalent Flux Density (SEFD). Under cold-sky conditions, MeerKAT sets attenuation to align ADC input power to the nominal − 29  dBFS (dBFS: Decibels relative to full scale; 0 dBFS is the digital maximum), with − 13  dBFS as the upper operational limit. Using a band-averaged SEFD (MeerKAT SEFD specifications) and typical mean solar flux densities of 100 sfu (UHF) and 200 sfu (L-band), the required additional attenuation is given by S att,dB S sun = 10 ⁡ log 10 S sun + S E F D S E F D , (1) yielding ∼ 32 dB and ∼ 35 dB for UHF- and L-bands, respectively. These estimates were validated with solar test observations on 11 November 2022 (UHF) and 11 January 2023 (L-band).

These estimates are based on quiet solar flux density and are designed to set the ADC power to the nominal power level, which is the minimum input power level required for optimal operation of the ADC. Hence, this value of S att provides the maximum possible headroom needed to accommodate the increased flux density during solar flares. While the estimates of solar flux density mentioned above are representative, the disc-integrated quiet Sun solar flux density can vary on timescales of a day. To account for this time variability, we developed a flexible system that automatically obtains the quiet solar flux density ( S sun ) from the previous day measurements from the Learmonth Solar Radio Observatory; alternatively, this information can also be provided by the user. This system estimates and applies the required additional attenuation ( S att,dB ) following Equation 1.

Since attenuators introduce an additional element into the signal path, it is essential to assess their impact on spectral properties, as well as their influence on visibility amplitudes and phases. To achieve this, we analyze the variations in visibility amplitudes and phases for various attenuation levels, while ensuring that the ADC power stays within its optimal operating range. The latter is needed to ensure that the signal-to-noise ratio (SNR) of visibility does not change significantly.

As discussed in Sections 2.1, 2.3, standard calibrators including the A-team sources cannot be observed with the same attenuation as used for the Sun, and the applied S att,dB exhibits a non-flat spectral response that must be characterized for absolute flux calibration. This is achieved using the built-in noise diode. Located after the LNA and before the attenuator in the signal chain, they inject a power of known strength, similar to the system temperature, T sys , in both UHF and L-band receivers. During calibrator scans without the additional attenuation, the noise diode-induced power change is d cal ( ν ) = P cal,on ( ν ) − P cal,off ( ν ) , where ν is the frequency and the subscripts on and off refer to the observed power with the noise diode switched on and off, respectively. During solar scans with additional attenuation, it is d sun ( ν ) = P sun,on ( ν ) − P sun,off ( ν ) . Hence, the ratio d sun ( ν ) / d cal ( ν ) captures the spectral variation in the attenuation, enabling calibration of the attenuator response and determination of the absolute flux density of the solar observations. The presence of additional attenuation S att,dB reduces the effective noise diode power during solar observations, making it difficult to achieve sufficient SNR, even when the noise power is similar to the T sys . As discussed in Section 4.2, this necessitates longer integration over time and/or frequency. The value of the noise diode being used for MeerKAT is close to the minimum needed for calibrating the attenuator response for solar observations. Reducing them below their current levels will require too long integrations, which is likely to make their use for solar flux density calibration impractical.

In preparation for solar observations with MeerKAT, we conducted tests in the engineering mode and arrived at the following observing procedure for routine solar observations.

1. Flux calibrator scan: Observe a standard MeerKAT flux/bandpass calibrator (e.g., J1939-6342 or J0408-6545 (MeerKAT flux and bandpass calibrators) with nominal attenuation.

We note that once the functionality required for inserting appropriate attenuation in the signal path and for toggling the noise diode on alternate correlator dumps is implemented in the MeerKAT Observation Planning Tool, all essential requirements for enabling a solar observing mode will have been met.

When the telescope points near the Sun, the system temperature can increase significantly. For the MeerKAT beam, this minimum angular distance, D sun,min , is approximately 7 ° in UHF and 4.5 ° in L-bands (MeerKAT Solar Avoidance Zone). Therefore, calibrators or astronomical targets should be observed at angular distances greater than D sun,min . Additionally, solar wind turbulence can introduce phase errors and scatter broadening in calibrator observations near the Sun. The phase error due to such turbulence can be estimated using (VLA Test Memo 236, Butler 2005; NRAO): R degree ∼ 7 λ cm B km 0.29 ϕ degree 0.71 , (2) where R degree is the minimum angular distance from the Sun (in degrees), λ cm the wavelength (in cm), B km the baseline (in km), and ϕ degree the allowable phase error (in degrees). For MeerKAT, assuming ϕ degree = 10 ° , Equation 2 yields R degree ∼ 15 ° (UHF) and 10 ° (L-band), as detailed in Table 1.

The Sun, being a non-sidereal source, its Equatorial coordinates (RA-Decl.) drift across the sky at an average rate of ∼ 1 ° per day ( ∼ 2 . 5 ′ ′ per minute), giving rise to a uniform shift in the coordinates of solar features. In radio interferometry, delay-tracking at the correlator compensates for geometric delays between received signals at different antennas as the source moves. While sidereal sources are tracked at a fixed equatorial coordinate, solar observations with a telescope require tracking the solar center. At MeerKAT, the correlator delay center is continuously updated at a kHz rate using a linear model, with model parameters–delay and delay-rate–estimated and refreshed every 5 s. This approach ensures accurate correlation and prevents decorrelation due to the solar apparent motion.

In addition to the non-sidereal motion, the Sun also exhibits differential rotation–the rotation period at its equator is ∼ 25 days and that at polar regions is ∼ 34 days (Mancuso et al., 2020). This causes solar features at different latitudes to move in the plane of the sky at varying rates, with the maximum projected motion occurring near the solar disk center. The maximum differential motion is, θ diff,rot ≈ 5 . 0 ′ ′ / h o u r . To limit smearing arising due to differential rotation, even after correcting for the overall non-sidereal motion of the Sun, integration times should not exceed ∼ 130 min in the UHF band and ∼ 50 min in the L-band.

We note that differential solar rotation breaks the “rigid-sky” assumption of radio interferometric imaging, making corrections in the visibility domain or during the imaging and deconvolution non-trivial. Current tools like Common Astronomy Software Application (CASA; The CASA Team et al., 2022) and W-Stacking CLEAN (WSClean; Offringa et al., 2014) do not support such corrections. This limitation will become more critical for the SKAO, with its higher spatial resolution, where uncorrected differential rotation may smear fine-scale features even when integrating over short times. For example, maximum integration time should be less than ∼ 5 min at 1 GHz to avoid smearing due to differential rotation. A dedicated imaging algorithm to address this is under development and will be presented in a forthcoming publication.

The Measurement Set (MS) is partitioned by scans and converted into multi-MS format for parallel processing using Dask. Flagging, calibration, and application of gain solutions are performed in parallel across scans. Persistent RFI and faulty antennas are flagged in all calibrator and solar scans. Automated RFI flagging using flagdata task of CASA in tfcrop mode is applied to flux and phase calibrators but skipped for solar scans due to the intrinsic variability of solar emission. Bandpass calibrator models (MeerKAT flux density and bandpass calibrator models) are used, and only scans without noise-diode firings are selected to derive delay, bandpass, and gain solutions using gaincal and bandpass, limited to baselines > 200 λ to avoid contamination from large angular scale quiet Sun emission. Post-calibration flagging is applied to residuals using rflag, followed by a final calibration round. However, bright compact solar features may still contaminate the longer baselines used. To assess this, we shift the phase center of the calibrator scans to the Sun and generate a dirty image using baselines > 200 λ . If the resulting contamination level, quantified as Δ I / I , exceeds 2% (corresponding to a tolerable gain error of 1%), we perform direction-dependent calibration to subtract the solar contribution and repeat the calibration iteration.

Absolute flux density calibration of solar observations is performed in two steps. First, the instrumental bandpass is calibrated without applying S att , using bandpass calibrator scans as described in Section 4.1. Next, the spectral response of S att is calibrated using the noise diode by measuring the change in auto-correlation power between the diode-on and diode-off states in both calibrator ( d cal ) and solar ( d sun ) scans. For calibrator scans without S att , the diode induces a significant power increase, leading to an estimated ∼ 2% variation in gain due to the departure of the system from linearity. In contrast, the effect is negligible for solar scans with S att due to the fractionally much smaller power increase. To allow us to correct for this non-linearity, bandpass solutions are derived separately for the diode-on and diode-off states using the calibrator scans and applied accordingly.

While power variations due to the noise diode are easily detectable in calibrator scans without averaging, this is not the case for solar scans due to the suppression by S att . Individual d sun estimates are noisy, but their SNR improves with integration time, until they get limited by intrinsic solar variability. To determine the optimal averaging time, we evaluated the standard deviation of the d sun spectrum as a function of integration time. As shown in the left panel of Figure 4, the standard deviation saturates beyond 15 min. Therefore, we adopt 15 min as the optimal integration interval for estimating d sun from solar scans. The middle and right panels of Figure 4 show the estimated S att spectra for a UHF-band observation with 32 dB attenuation in both polarizations. The spectra exhibit intrinsic frequency dependence but no significant scan-to-scan variation beyond noise. To avoid using noisy per-channel estimates and increasing SNR, we fit a cubic spline to the scan-averaged spectra and use it to scale the bandpass solutions applied to the solar scans. The fitted spectra are shown by the solid black lines in the middle and right panels of Figure 4.

Solar radio emission exhibits strong spectral and temporal variability, making the solar sky model inherently time-dependent. The time and frequency scale of the variation depends on solar activity and can range from a few seconds to several minutes and a few kHz to several MHz. To address this dynamic temporal and spectral variability scale, the spectral and temporal axes are adaptively divided into chunks such that deviations from the mean remain below certain thresholds (default is 10% for frequency, 1% for time). This ensures that variability is preserved while enabling computationally efficient self-calibration. A spectral chunk from the lower part of the band is selected to maximize surface brightness sensitivity and improve modeling on shorter baselines. The self-calibration procedure follows the convergence criteria in Kansabanik et al. (2022), Kansabanik et al. (2023), starting with phase-only calibration and advancing to joint amplitude-phase calibration upon convergence. CLEAN thresholding is progressively reduced, and the process is stopped when no further improvement in image dynamic range is observed. Convergence is defined as a relative change in dynamic range below a user-defined threshold ε over three iterations, with a maximum iteration cap to prevent oscillatory behavior for small ε .

At GHz frequencies, solar radio emission exhibits structure across a broad range of angular scales—from arcseconds to the full solar disc—often with significant complexity. The imaging pipeline supports user-defined baseline selection and weighting strategies, and by default adopts Briggs weighting (Briggs, 1995) with a robust parameter of 0.0 to achieve a balance between resolution and sensitivity. Multiscale deconvolution is employed with frequency-dependent multiscale parameters. These choices, detailed in Supplementary Appendix S9.1, are made to avoid deconvolution artifacts.

As the Sun is an extended source, its observed emission must be corrected for the direction-dependent primary beam response. We apply image-based primary beam correction using the array-averaged MeerKAT beam model from holography measurements (de Villiers and Cotton, 2022; de Villiers, 2023). The beam is described by the Jones matrix P ( l , m , ν ) in direction cosines ( l , m ) from the boresight of the telescope. At MeerKAT, H and V correspond to Y and X polarizations (MeerKAT polarization convention) as per IAU convention (Commission 40: Radio astronomy, 1973) used in common softwares like with CASA and WSClean. H and V polarizations are appropriately labeled in the IAU convention in the measurement set using katdal software package. Hence, appropriate changes are also made in P ( l , m , ν ) to be consistent with the IAU convention.

To apply correction, the beam is first mapped from ( l , m ) to equatorial coordinates of the image, then rotated by the parallactic angle χ . The sky-frame beam matrix is: P sky = P l , m R χ (3) where the parallactic rotation matrix is: R χ = c o s χ − s i n χ s i n χ c o s χ (4)

The frequency-averaged Stokes I beam is computed as: P I = ∑ ν = ν 0 ν 1 1 2 ∑ i , j = 0 i , j = 1 | P sky i , j | 2 ν 1 − ν 0 (5) where ν 0 and ν 1 represent the start and end frequencies of the image. The Stokes I image corrected for the primary beam is obtained by dividing by P I obtained using Equations 3–5. We note that while UHF and L-band have overlapping frequencies, the primary beam should be derived from the appropriate band, as they use different feeds, resulting in distinct beam responses even at the same frequency.

EUV spectral lines (Khan and Nagaraju, 2022) and soft X-ray observations are widely used to probe the thermal structure of the solar atmosphere. Slit-based spectrographs like EUV Imaging Spectrometer onboard Hinode (Hinode/EIS; Culhane et al., 2007) and Coronal Diagnostic Spectrometer onboard Solar and Heliospheric Observatory (SOHO/CDS; Del Zanna et al., 2001) offer good spectral resolution and broad temperature coverage (log T ∼ 4.9 − 6.5 ), primarily focused on coronal plasma. The Interface Region Imaging Spectrograph (IRIS; DePontieu et al., 2014) provides better spectral and spatial resolution for chromosphere and transition region (TR) plasma. Although these instruments can probe all the layers of the solar atmosphere, their limited field of view (FoV) restricts them to studying local dynamics.

In contrast, full-disk EUV imagers such as SDO/AIA, Solar Ultraviolet Imager onboard Geostationary Operational Environmental Satellite (GOES/SUVI; Darnel et al., 2022) and mid- and near-UV imagers like the Solar Ultraviolet Imaging Telescope onboard Aditya-L1 (SUIT/Aditya-L1; Tripathi et al., 2025), provide high-cadence observations in multiple broad UV channels. While EUV-imagers nominally span log T ∼ 3.7 − 8.0 , their temperature sensitivity for chromospheric and TR plasma is limited, and DEMs can be reliably estimated primarily in the log T ∼ 5.0 − 8.0 range using optically thin lines (Hannah and Kontar, 2012; Cheung et al., 2015).

Radio observations directly measure the free–free continuum emission from all of the plasma along the line of sight (LoS), providing sensitivity to the total emission measure across a broad temperature range. Hence, high-fidelity spectroscopic and snapshot imaging observation using MeerKAT offers a complementary tool to probe the full-disk TR and coronal plasma dynamics, filling the gap between slit-based spectrographs and EUV imagers.

Solar eruptions are explosive phenomena that occur in the solar atmosphere, involving the sudden release of vast amounts of energy stored in the magnetic field due to coronal dynamics. These events include solar flares, CMEs, and eruptive prominences, and are manifestations of magnetic reconnection and plasma instabilities in the solar corona. They can accelerate particles to high energies and expel large amounts of magnetized plasma in the form of CMEs into the heliosphere. These energetic particles and CMEs play a crucial role in driving space weather. Understanding the evolution of magnetic fields during their initiation, evolution, and propagation is essential for a deeper understanding of these phenomena. Spectroscopic radio imaging plays a crucial role in providing the magnetic field measurements remotely during these eruptions (Vourlidas et al., 2020; Alissandrakis and Gary, 2021; Carley et al., 2021) as well as providing an estimation of non-thermal particles associated with these processes. Both high-frequency ( > 1 GHz) (Chen et al., 2020; Fleishman et al., 2020) and low-frequency (meter-wavelength) (Bastian et al., 2001; Mondal et al., 2020b; Kansabanik et al., 2023; Kansabanik et al., 2024) observations have demonstrated their capabilities for probing non-thermal particles and measuring magnetic fields at the flare site and CME plasma at higher coronal heights, respectively. However, these eruptions lack observational probes in a crucial region in the lower corona, both in white-light, EUV, and radio wavelengths. Recently, new-generation visible light instruments, PROBA-3 (Shestov, S. V. et al., 2021) and Aditya-L1/VELC (Singh et al., 2025), and SunCET (Mason et al., 2021) in the EUV, have been designed to observe this coronal region (at heliocentric distances ≲ 2 R ⊙ ). The frequency range and high fidelity spectroscopic snapshot imaging capability of MeerKAT make it highly suitable for observing these eruptions between heliocentric distances of ∼ 1 − 2 R ⊙ .

We observed the Sun with MeerKAT from 2024 June 9–11 as part of SSV observations, targeting the active region (AR) NOAA (National Oceanic and Atmospheric Administration) AR 13711 and the reappearance of AR 13664, which was responsible for the May 2024 geomagnetic storm, anticipating continued activity. The top left panel of Figure 9 shows a composite image of a CME event which erupted from the western limb on 10 June 2024, at 09:40 UTC, as generated using JHelioviewer (Müller et al., 2017), with the central image showing GOES/SUVI 195 Å EUV image, and the outer panel showing SOHO/LASCO (Brueckner et al., 1995; Domingo et al., 1995) C2 base-difference coronagraph image. The top right panel shows an image from MeerKAT at 629 MHz with contours overlaid on the SUVI image, with the eruption site marked by the red box. During this observing window, a long-duration X-class flare peaking at 11:07 UTC (blue line in the bottom panel of Figure 9) occurred and remained above M-class level for a large period, from approximately 09:48 to 12:22 UTC (blue shaded region).

Several intense solar radio bursts were observed during this period in MeerKAT total power normalized dynamic spectrum, as shown in the top panel of Figure 10. A prominent, broadband burst around 10:46 UTC is highlighted in the bottom left panel, while several reverse-drifting bursts, typically associated with sunward-traveling electrons, are seen in the bottom right panel. Since dynamic spectra provide only spatially integrated information, distinguishing overlapping emissions requires spatially resolved dynamic spectra (SPREDS; Mohan and Oberoi, 2017). Previously demonstrated at meter wavelengths using the MWA, the high-dynamic-range spectroscopic snapshot capability of MeerKAT at GHz frequencies offers similar potential. A detailed SPREDS-based analysis of these radio bursts is beyond the scope of this work.

Three snapshot radio images at 629 MHz overlaid on SUVI images during the CME eruption are shown in the top panels of Figure 11. The bottom left panel highlights the eruption site. The region from where the spectra have been extracted at multiple timestamps is marked and the spectra are shown in the bottom right panel. A progressive steepening of the spectra with time is observed, suggesting a non-thermal origin of the emission. Spectroscopic snapshot imaging at high spectral and temporal cadence can offer valuable insights into the evolving physical conditions at the eruption site, from pre-eruption to post-eruption. A detailed analysis will be presented in a future study.