Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userA multiple-beam CLEAN for imaging intra-day variable radio sources
219
0
0.0
(
0
)
Ask ChatGPT about the research
No Arabic abstract
The CLEAN algorithm, widely used in radio interferometry for the deconvolution of radio images, performs well only if the raw radio image (dirty image) is, to good approximation, a simple convolution between the instrumental point-spread function (dirty beam) and the true distribution of emission across the sky. An important case in which this approximation breaks down is during frequency synthesis if the observing bandwidth is wide enough for variations in the spectrum of the sky to become significant. The convolution assumption also breaks down, in any situation but snapshot observations, if sources in the field vary significantly in flux density over the duration of the observation. Such time-variation can even be instrumental in nature, for example due to jitter or rotation of the primary beam pattern on the sky during an observation. An algorithm already exists for dealing with the spectral variation encountered in wide-band frequency synthesis interferometry. This algorithm is an extension of CLEAN in which, at each iteration, a set of N `dirty beams are fitted and subtracted in parallel, instead of just a single dirty beam as in standard CLEAN. In the wide-band algorithm the beams are obtained by expanding a nominal source spectrum in a Taylor series, each term of the series generating one of the beams. In the present paper this algorithm is extended to images which contain sources which vary over both frequency and time. Different expansion schemes (or bases) on the time and frequency axes are compared, and issues such as Gibbs ringing and non-orthogonality are discussed. It is shown that practical considerations make it often desirable to orthogonalize the set of beams before commencing the cleaning. This is easily accomplished via a Gram-Schmidt technique.
rate research
Read More
With the arrival of a number of wide-field snapshot image-plane radio transient surveys, there will be a huge influx of images in the coming years making it impossible to manually analyse the datasets. Automated pipelines to process the information stored in the images are being developed, such as the LOFAR Transients Pipeline, outputting light curves and various transient parameters. These pipelines have a number of tuneable parameters that require training to meet the survey requirements. This paper utilises both observed and simulated datasets to demonstrate different machine learning strategies that can be used to train these parameters. The datasets used are from LOFAR observations and we process the data using the LOFAR Transients Pipeline; however the strategies developed are applicable to any light curve datasets at different frequencies and can be adapted to different automated pipelines. These machine learning strategies are publicly available as Python tools that can be downloaded and adapted to different datasets (https://github.com/AntoniaR/TraP_ML_tools).
We carried out a pilot campaign of radio and optical band intra-day variability (IDV) observations of five blazars (3C66A, S5 0716+714, OJ287, B0925+504, and BL Lacertae) on December 18--21, 2015 by using the radio telescope in Effelsberg (Germany) and several optical telescopes in Asia, Europe, and America. After calibration, the light curves from both 5 GHz radio band and the optical R band were obtained, although the data were not smoothly sampled over the sampling period of about four days. We tentatively analyse the amplitudes and time scales of the variabilities, and any possible periodicity. The blazars vary significantly in the radio (except 3C66A and BL Lacertae with only marginal variations) and optical bands on intra- and inter-day time scales, and the source B0925+504 exhibits a strong quasi-periodic radio variability. No significant correlation between the radio- and optical-band variability appears in the five sources, which we attribute to the radio IDV being dominated by interstellar scintillation whereas the optical variability comes from the source itself. However, the radio- and optical-band variations appear to be weakly correlated in some sources and should be investigated based on well-sampled data from future observations.
Active galactic nuclei (AGN) are known for irregular variability on all time scales, down to intra-day variability with relative variations of a few percent within minutes to hours. On such short timescales, unexplored territory, such as the possible existence of a shortest characteristic time scale of activity and the shape of the high frequency end of AGN power spectra, still exists. We present the results of AGN single-dish fast photometry performed with the Korean VLBI Network (KVN). Observations were done in a anti-correlated mode using two antennas, with always at least one antenna pointing at the target. This results in an effective time resolution of less than three minutes. We used all four KVN frequencies, 22, 43, 86, and 129 GHz, in order to trace spectral variability, if any. We were able to derive high-quality light curves for 3C 111, 3C 454.3, and BL Lacertae at 22 and 43 GHz, and for 3C 279 at 86 GHz, between May 2012 and April 2013. We performed a detailed statistical analysis in order to assess the levels of variability and the corresponding upper limits. We found upper limits on flux variability ranging from $sim$1.6% to $sim$7.6%. The upper limits on the derived brightness temperatures exceed the inverse Compton limit by three to six orders of magnitude. From our results, plus comparison with data obtained by the University of Michigan Radio Astronomy Observatory, we conclude that we have not detected source-intrinsic variability which would have to occur at sub-per cent levels.
We present our observations of the optical intra-day variability (IDV) in $gamma$-ray BL Lac object Mrk 501. The observations were run with the 1.02 m and 2.4 m optical telescopes at Yunnan Observatories from 2005 April to 2012 May. The light curve at the $R$ band on 2010 May 15 passes both variability tests (the $F$ test and the ANOVA test). A flare within the light curve on 2010 May 15 has a magnitude change $Delta m = 0.03 pm 0.005_{rm{stat}} pm 0.007_{rm{sys}}$ mag, textbf{a darkening timescale of $tau_{rm{d}}=$ 26.7 minutes}, and an amplitude of IDV $Amp=2.9% pm0.7%$. A decline textbf{described by 11 consecutive flux measurements} within the flare can be fitted linearly with a Pearsons correlation coefficient $r = 0.945$ at the confidence level of $> 99.99%$. Under the assumptions that the IDV is tightly connected to the mass of the black hole, textbf{and that the flare duration, being two times $tau_{rm{d}}$, is representative of the minimum characteristic timescale, we can derive upper bounds to the mass of the black hole}. In the case of the Kerr black hole, the timescale of $Delta t_{rm{min}}^{rm{ob}}=$ 0.89 hours gives $M_{bullet}la 10^{9.20} M_{odot}$, which is consistent with measurements reported in the literature. This agreement indicates that the hypothesis about $M_{bullet}$ and $Delta t_{rm{min}}^{rm{ob}}$ is consistent with the measurements/data.
Ongoing or upcoming surveys such as Gaia, ZTF, or LSST will observe light-curves of billons or more astronomical sources. This presents new challenges for identifying interesting and important types of variability. Collecting a sufficient number of labelled data for training is difficult, however, especially in the early stages of a new survey. Here we develop a single-band light-curve classifier based on deep neural networks, and use transfer learning to address the training data paucity problem by conveying knowledge from one dataset to another. First we train a neural network on 16 variability features extracted from the light-curves of OGLE and EROS-2 variables. We then optimize this model using a small set (e.g. 5%) of periodic variable light-curves from the ASAS dataset in order to transfer knowledge inferred from OGLE/EROS-2 to a new ASAS classifier. With this we achieve good classification results on ASAS, thereby showing that knowledge can be successfully transferred between datasets. We demonstrate similar transfer learning using Hipparcos and ASAS-SN data. We therefore find that it is not necessary to train a neural network from scratch for every new survey, but rather that transfer learning can be used even when only a small set of labelled data is available in the new survey.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
