Zurueck zur Startseite

  Current Research Highlight :: May 2014 all highlights

Resolving the radio sky

Radio astronomers obtain extremely high resolution sky images by using interferometers, instruments where several single radio telescopes are linked together. However, optimal data analysis procedures for such an instrument are significantly more involved than for a single telescope. Scientists from the Max Planck Institute for Astrophysics have now developed the algorithm RESOLVE which solves a number of outstanding problems in radio imaging.

Fig. 1: The Very Large Array (VLA) is a collection of 27 radio antennas located on the plains of San Augustin near Socorro, New Mexico, each with a dish 25 meters in diameter and weighing more than 200 tons. The data from all antennas can be combined electronically so that the array effectively functions as one giant antenna.
Image courtesy of NRAO/AUI

Fig. 2: This is a false colour image of the region surrounding the W28 supernova remnant with the radio emission detected with the VLA shown in blue. The more compact objects north and south of W28 are regions of ionized hydrogen not directly related to the remnant. The new image reconstruction method will make it much easier to reconstruct interferometry images of such extended sources.
Credit: NRAO/AUI/NSF and Brogan et al.

Fig. 3: A simulated observation of a galaxy cluster with the VLA. The image on the top left shows the (real) input signal. The top right image shows the reconstruction using the RESOLVE algorithm. The image on the bottom left shows a reconstruction with a standard algorigthm (CLEAN) while the image on the bottom right gives the relative uncertainty of the reconstruction, note the different scale on this map.

Using radio interferometers, scientists look into the deepest depths of the Universe. These instruments deliver high-resolution images of many different celestial objects, ranging from the Sun, over pulsars, and the interstellar gas in the Milky Way, to distant sources such as radio galaxies or quasars. The high-resolution radio images of such objects often reveal their complex and extended structure.

Indeed, most of the radio emission from celestial sources originates in extended cosmic plasma clouds, glowing only faintly to the observer on Earth. In consequence, such extended regions of emission are difficult to detect, since they have to be separated from unwanted interferences as e.g. electronic noise from terrestrial technical equipment or atmospheric effects.

Furthermore, imaging in radio interferometry is inherently more complicated as for a single telescope. This is because an interferometer does not detect the celestial sources directly, instead the signals from different detectors are electronically superimposed. To reconstruct the original signal from the data, a so called Fourier transformation needs to be applied, usually implying complex calculations on the computer. Unfortunately, standard imaging methods have the drawback that they often only produce unreliable results for weak and extended emission. Moreover, due to the complex nature of the interferometric observation, in general an estimation of the measurement uncertainty was unreliable so far as well.

In two recent publications, the new imaging algorithm RESOLVE ("Radio Extended Sources Lognormal Deconvolution Estimator") is presented to solve exactly these problems of current methods. RESOLVE employs a statistical approach, estimating the most probable image reconstruction compatible with the measured data. In this process, the algorithm uses the vague prior knowledge of the observer on the source — namely that it is an extended object — to differentiate between likely and unlikely reconstructions. To this end, RESOLVE assumes that the radio intensity does not change abruptly from one place to the next, but instead that the source is comprised of statistically similar structures, connected over several pixels, and not necessarily exactly known prior to an observation. Mathematically, this is expressed by a so-called spatial correlation function, unknown at the beginning of the reconstruction process.

RESOLVE can roughly be divided into two major steps. In the first part, the statistically most probable image reconstruction, compatible with an extended source, is estimated. In this step, the spatial correlation function is assumed to be known by the algorithm and thus influences the reconstruction process. In the second part, the correlation function is estimated using the intermediate image reconstruction obtained in the first step. RESOLVE iterates this two-step process until a statistically optimal reconstruction has been obtained. Finally, from the last reconstruction, a map of the measurement uncertainty is calculated.

This procedure can be extended to observations at different wavelengths. For this, in addition, the spectral dependence of the radio emission in every pixel is estimated using a very similar method as just described.

Simulated reconstructions using RESOLVE show that from high quality interferometric data, it is indeed possible to computationally reverse the complicated measurement process of the interferometer with high precision and to estimate the structure of an extended radio source with high precision. In addition, the noise is removed from the measured signal and a measurement uncertainty is estimated during this process. Possible areas of application in observational radio astronomy range from single objects in the Milky Way like e.g. remnants of exploding stars, to distant radio galaxies and large galaxy clusters. The new image reconstructions will allow for a deeper and better resolved view into the radio sky.

Henrik Junklewitz, Michael Bell and Torsten Enßlin


Henrik Junklewitz, Michael Bell, Marco Selig and Torsten Enßlin, "RESOLVE: A new algorithm for aperture synthesis imaging in radio astronomy", submitted to A&A

Henrik Junklewitz, Michael Bell and Torsten Enßlin, "A new approach to multi-frequency imaging in radio interferometry", submitted to A&A

drucken.gif print version topPfeil.gif Top
© 2003—2017, Max-Planck-Gesellschaft, München
last modified: 2014-5-5