LENSCLEAN revisited

We discuss the LENSCLEAN algorithm which for a given gravitational lens model fits a source brightness distribution to interferometric radio data in a similar way as standard CLEAN does in the unlensed case. The lens model parameters can then be varied in order to minimize the residuals and determine the best model for the lens mass distribution. Our variant of this method is improved in order to be useful and stable even for high dynamic range systems with nearly degenerated lens model parameters. Our test case B0218+357 is dominated by two bright images but the information needed to constrain the unknown parameters is provided only by the relatively smooth and weak Einstein ring. The new variant of LENSCLEAN is able to fit lens models even in this difficult case. In order to allow the use of general mass models with LENSCLEAN, we develop the new method LENTIL which inverts the lens equation much more reliably than any other method. This high reliability is essential for the use as part of LENSCLEAN. Finally a new method is developed to produce source plane maps of the unlensed source from the best LENSCLEAN brightness models. This method is based on the new concept of ‘dirty beams’ in the source plane. The application to the lens B0218+357 leads to the first useful constraints for the lens position and thus to a result for the Hubble constant. These results are presented in an accompanying Paper II, together with a discussion of classical lens modelling for this system.