VLBI properties of compact interplanetary scintillators detected by the Murchison Widefield Array

Scintillation is a phenomenon caused by changes in the phase and/or amplitude of radio waves due to the scattering effects of the turbulent neutral and ionised media lying between a radio source and the Earth (Thompson et al., 2017). There are three main types of scintillations, namely ionospheric, interplanetary and interstellar, depending on the distance of the scattering medium from the observer and the size of the plasma structure. It can be readily shown that scintillation is weak at high frequencies and for nearby scattering media. Interplanetary scintillation (IPS) was discovered by Clarke (1964) who identified compact ($<$2 arcsec) radio sources with rapid ($\sim$1 s time-scale) and random flux density fluctuations caused by turbulent solar wind plasma. IPS has been used as an astrophysical tool to infer the sub-arcsecond scale structures of radio sources (Hewish et al., 1964), without the need for high-resolution imaging observations of each source separately. Observations of IPS sources can also be used to probe the characteristics of the intervening plasma and thereby the corresponding solar activity (e.g., Salpeter, 1967; Cohen et al., 1967; Ekers & Little, 1971; Coles, 1978; Jackson et al., 1998). Purvis et al. (1987) published a catalogue of 1789 IPS sources at 81.5 MHz using the Cambridge IPS array.

Large synthesis arrays of multiple elements such as the Murchison Widefield Array (MWA; Tingay et al., 2013) have sufficient instantaneous (u,v) coverage and temporal resolution to allow rapid measurements of a large number of IPS sources throughout the entire field of view. Recently, Morgan et al. (2018) and Chhetri et al. (2018a); Chhetri et al. (2018b) detected sub-arcsecond components within many hundreds of radio sources using the $30\degr\times 30\degr$ MWA field-of-view at $\sim$160 MHz, with just 5 min of data and a time resolution of 0.5 s. Interestingly, they found that those sources displaying strong scintillation are mainly peaked or inverted spectrum sources, analogous to the GHz-peaked spectrum sources (GPS, O’Dea, 1998; O’Dea & Saikia, 2021).  Inverted-spectrum or compact steep-spectrum sources are morphologically associated with radio-loud active galactic nuclei (AGN) with sub-kpc to several kpc scale structures, which are thought to be young radio AGN in their early evolutionary stages or aged AGN frustrated by the surrounding dense interstellar medium (ISM) (e.g., Kunert-Bajraszewska et al., 2010; An & Baan, 2012; Sadler, 2016; Callingham et al., 2017; O’Dea & Saikia, 2021).

In terms of its timescale, mechanism, and frequency dependence, the IPS phenomenon is very different from intra-day variability (IDV), which is due to a combination of intrinsic variability of the radio source itself and extrinsic variability caused by the interstellar medium. IDV is evident at much higher frequencies than IPS. Moreover, the timescale of IDV ranges from hours to days, much larger than the IPS timescale of a few seconds (Bignall et al., 2015). IDV has been observed in many flat-spectrum AGNs (Quirrenbach et al., 1992; Bignall et al., 2003; Liu et al., 2012) with the associated angular size smaller than a few tens of micro-arcseconds according to the scintillation criteria of the first Fresnel zone at $\gtrsim 5$ GHz frequencies for typical scattering screen distances in case of the interstellar scintillation. The IPS sources, on the other hand, are expected to have sub-arcsecond size (e.g., Hewish et al., 1964; Cohen et al., 1967; Morgan et al., 2019). IDV has been well-studied in the literature, especially in the Micro-Arcsecond Scintillation-Induced Variability (MASIV) survey using Very Large Array (Krichbaum et al., 2002; Lovell et al., 2003; Jauncey et al., 2007; Lovell et al., 2008; Koay et al., 2011, 2018). For some sources, very long baseline interferometry (VLBI) observational studies (e.g., Ojha et al., 2004a, b, 2006; Said et al., 2021) were conducted to confirm their compactness and to understand how the interstellar scintillation depends on the source compactness. However, no such high-resolution imaging studies of IPS sources have been carried out to understand their behaviour on VLBI scales.

In this study, we present an analysis of GHz-frequency VLBI data of a sample of MWA-detected IPS sources in an attempt to understand how their pc-scale properties, such as morphology, radio spectrum, brightness temperature, and variability are related to the scintillation properties inferred from the MWA observations.

We used the IPS catalogue compiled by Chhetri et al. (2018a); Chhetri et al. (2018b). The observation was made on 2016 May 1 at 02:25:20 UTC and lasted 5 minutes. The pointing centre was at RA = 00^h49^m02^s, Dec = $-$19$\degr$58$\arcmin$54$\arcsec$. Of the 2550 continuum sources detected in the 900 deg² MWA field-of-view, 247 IPS sources showing rapid fluctuations were identified with a signal-to-noise ratio of $\geq 5$. After removing three gravitational lensing sources (J013212$-$065232, J014013$-$095654 and J015741$-$104345) identified from the literature, the IPS sample contains 244 AGN. We searched these sources in the available VLBI archive, mainly the VLBA Calibrator Surveys¹¹1http://astrogeo.org/ maintained by L. Petrov. (Beasley et al., 2002; Fomalont et al., 2003; Petrov et al., 2005, 2006; Kovalev et al., 2007; Petrov et al., 2008), and found 63 sources after removing two sources (J004839$-$294718 and J005429$-$235127) with poor quality VLBI data within a search radius of $\sim$1 arcmin. The resolution of MWA observation was $\sim$2 arcmin (Wayth et al., 2015), so the search radius is approximately a half of the MWA beam size. The cross-matching of the sources was further confirmed by examining the corresponding NRAO VLA Sky Survey (NVSS; Condon et al., 1998) and Very Large Array Sky Survey (VLASS; Lacy et al., 2020; Gordon et al., 2020) images. Therefore, the parent sample size of the IPS sources is 244 out of which the VLBI-selected sample size is 63.

Although these VLBI observations were mostly made in snapshot mode at 2, 5 and 8 GHz, they are capable of detecting bright compact sources above $\sim$15 mJy at GHz frequencies and can be used to check whether these sources have compact structure on mas scales. The VLBI survey campaigns cover the sky zone of the IPS observation made by Morgan et al. (2018) and Chhetri et al. (2018a). However, for sources in the southern sky, the sensitivity of VLBA observations decreases due to the high temperature of the antenna and the unavailability of some high-latitude telescopes. Therefore, even if all sources in the parent sample are at $\delta>-40^{\circ}$, the lower sensitivity of the VLBA snapshot observations may result in some of the sources not being detected. A total of 484 sources are found in VLBI archive within the 900 deg² MWA field-of-view, suggesting that a large fraction of these sources are not detected by the MWA during the time span (i.e., 5 min). Continuous IPS observations on longer time scales can help determine the fraction of VLBI sources in the IPS sample and the physical reasons (e.g., sensitivity, variability) behind this fraction.

These VLBI sources were well distributed over a variety of pc-scale morphological classes (i.e., compact core, core + one sided jet, core + two sided jets, and core + extended structure) and a variety of overall spectrum shapes (i.e., simple power-law spectrum, inverted/convex spectrum, and flat spectrum); see Sections 4 and 5 for details. As mentioned above, although the VLBI sample in this paper is not complete and inclined to bright, flat-spectrum sources, the widely distribution of the radio properties still allow for an exploratory study. The basic properties of the IPS–VLBI sample are presented in Table 1, which include the source coordinates, redshifts (obtained from the NASA/IPAC Extragalactic Database, NED, if available), MWA flux densities at 162 MHz ($S_{\rm MWA}^{\rm 162MHz}$), spectral index $\alpha$ from MWA GLEAM ($\alpha_{\rm GLEAM}$), the NVSS 1.4-GHz flux densities ($S_{\rm NVSS}^{\rm 1.4GHz}$), the VLASS 3-GHz flux densities ($S_{\rm VLASS}^{\rm 3GHz}$), and normalised scintillation indices (NSI; Chhetri et al., 2018a). The scintillation index of an IPS source is defined as the ratio of the root mean square (rms) of the flux density fluctuations of the source to its mean flux density, and varies with the source elongation from the Sun (e.g. Manoharan, 1993). The normalised scintillation index is defined as the ratio of the observed scintillation index and the scintillation index predicted by the solar elongation of the source, and is close to unity for sources dominated by a single sub-arcsec compact component (Chhetri et al., 2018a).

Figure 1 shows the distributions of $S_{\rm MWA}^{\rm 162MHz}$ and $S_{\rm VLASS}^{\rm 3GHz}$ for the 63 sources and their comparison with the parent sample data. These histograms indicate that for most sources, $S_{\rm MWA}^{\rm 162MHz}<6$ Jy and $S_{\rm VLASS}^{\rm 3GHz}<0.6$ Jy. The flux density distribution of the IPS-VLBI sample at low frequency (162 MHz) is consistent with that of the parent sample. However, the peak of the flux density distribution of the selected sample at 3 GHz is shifted toward higher flux densities compared to the parent sample. This is expected as the VLBA Calibrator Surveys were designed to detect compact sources that are brighter than $\sim$100 mJy at 1.4 GHz at arcsec scales if the spectral index $\geq-0.5$, otherwise brighter than $\sim$200 mJy. However, for the southern sky, the flat-spectrum sources with 1.4 GHz flux density $\geq$200 mJy at arcsec scales were mainly observed using the VLBA in snapshot mode. In the parent sample, 81 IPS sources are fainter than 100 mJy at 3 GHz, and 152 sources are fainter than 200 mJy. This indicates at least one-third of sources the parent sample are not observed by the VLBA because of this flux density limitation.

Figure 2 shows $\alpha_{\rm GLEAM}$, $\alpha_{\rm GHz}$ and NSI values for the VLBI-detected sample and the parent sample. The estimated spectral indices in the GLEAM bands range from $-1.1$ to $1.0$. The histogram of $\alpha_{\rm GLEAM}$ of the parent sample shows a peak around $\alpha=-0.7$, a typical value of optically thin steep spectrum; while there is a secondary peak around $\alpha=-0.4$. The histogram of $\alpha_{\rm GHz}$ of the parent sample shows a narrow peak around $\alpha=-1.0$, steeper than the peak of the $\alpha_{\rm GLEAM}$ histogram. $\alpha_{\rm GLEAM}$ of both the VLBI-detected and parent samples shows a relatively flatter distribution, but the relative proportion of VLBI-detected flat-spectrum sources in the parent sample is higher than that of steep-spectrum sources, which is consistent with the bias of VLBI observations towards flat-spectrum radio-loud AGN. The NSI values of the VLBI-detected sample sources are evenly distributed over the NSI range of 0.1–1.7, and similarly, the parent sample shows a flat distribution of NSI values between 0.2 and 1.4. However, the available 63 VLBI sources show a weak tendency to cluster towards higher NSI values compared to the parent sample; this of course needs to be confirmed with more data. Among the 63 VLBI sources, 31 are strongly scintillating sources with NSI $\geq 0.9$, 25 moderately scintillating sources with $0.9>$ NSI $\geq 0.4$ and 7 weakly scintillating sources with $0.4>$ NSI $\geq 0.1$.

The calibrated VLBI archival data for the $S$ (2.3 GHz), $C$ (4.3 GHz), $X$ (7.6 and 8.7 GHz) and $U$ (15.3 GHz) bands for the 63 sources were obtained from the Astrogeo database. The calibration had been performed with the VLBI data processing software Pima (Petrov et al., 2011). The calibrated data were further analysed using the Caltech Difmap software package (Shepherd et al., 1994) to deal with any residual phase errors, imaging, and parameter estimation. If the on-source time exceeded 1 minute, any data with significant scattering were averaged in time over 30 seconds. Any bad data points due to the presence of radio frequency interference and/or any antenna not working properly were identified and flagged. The edited visibility data were then imaged and self-calibrated following standard procedures. Final images were generated using self-calibrated data by applying natural weighting. Deconvolution of the images used the CLEAN algorithm (Hogbom & Brouw, 1974). In the final images, the CLEAN procedure was performed down to three times the residual noise. In many cases, the on-source time was less than 1 minute, so CLEAN boxes were carefully used during the processing of such data with poor (u,v)-coverage. It is important to note that the VLBI data with poor (u,v)-coverage caused strong side-lobes and prevent deep CLEANing. The observing and imaging parameters are given in Table 2, which lists the quality of the archival data in the different bands and epochs. Based on the availability of the data, we used single-epoch data at $S$, $C$ and $U$ bands, and multiple-epoch data at $X$ band for this study. As the archive data were observed with different on-source times and at different frequency bandwidths in different epochs, the final images have different sensitivities.

By using the modelfit task in Difmap, we fitted circular Gaussian models to the self-calibrated visibility data until no structure with brightness higher than 5$\sigma$ appears in the residual image and the amplitude radial profile of the model matches the visibility data. We assume the brightest component to be the core and other prominent components as jet components. The integrated flux densities and angular sizes of the source components are given in Table 6. The parameters derived from fitting using circular Gaussian models are consistent with those obtained by using elliptical Gaussian models. When the fitted Gaussian size $d$ of a component was found to be smaller than the minimum resolvable size $d_{\rm min}$ estimated using the Lobanov (2005) relation, we used $d_{\rm min}$ as the upper limit for the component size. The uncertainties in the model fitted parameters were estimated using the relations given in Schinzel et al. (2012), which were based on the approximations introduced by Fomalont (1999) and were modified for the strong sidelobes in VLBI measurements. In addition to the random statistical error, a systematic amplitude calibration error of 5 per cent of the flux density value was also included in the final flux density error.

We have performed a high-resolution imaging study of 63 interplanetary scintillation sources with available VLBI data selected from the total 244 IPS sources detected in the $30\degr\times 30\degr$ MWA field-of-view at 162 MHz (Chhetri et al., 2018a; Chhetri et al., 2018b). The main conclusion of this study is that for the ‘compact core’, ‘core + one-sided jets’ and ‘core + two-sided jets’ sources (excluding the blazar-like sources), there is a positive correlation between the source compactness and the scintillation index, indicating stronger scintillation for more compact sources. A similar positive correlation is found for the sources with simple power law or convex spectrum shapes. The scatter of these correlations might be due to the intrinsic variability and complex structure of the sources.

Observations of IPS sources at meter wavelengths with connected synthesis arrays have provided information on the presence of compact components with sizes smaller than $\sim$0.3 arcsec. Moreover, details of these compact components and how interplanetary scintillation properties vary with compactness can be obtained from VLBI studies of IPS sources. This comprehensive study of the radio properties of IPS sources on mas scales is the first of its kind to use the most extensive archival VLBI database. New VLBI observations of other IPS sources in the sample for which VLBI data are not available would be a necessary complement to improve the statistical confidence of the results presented here. Future IPS survey with the extended MWA Phase II will help to extend this study.

We thank the referee Hayley Bignall for her valuable comments and suggestions which help to improve the quality of the manuscript. SJ and TA thank Rajan Chhetri for providing the catalog of 247 IPS sources used in this study. This work is supported by the SKA grant from the National Key R&D Programme of China (2018YFA0404603), the Chinese Academy of Sciences (CAS, 114231KYSB20170003). SJ is supported by the CAS-PIFI (grant No. 2020PM0057) postdoctoral fellowship. TA is supported by the Youth Innovation Promotion Association of CAS. We are grateful to the Astrogeo team for making their fully calibrated VLBI FITS data publicly available. The NASA/IPAC Extragalactic Database (NED) is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc.

The VLBI datasets underlying this article were derived from the public domain in Astrogeo archive (http://astrogeo.org/). The IPS data can be found in the published literature.