THE SCATTER-BROADENED IMAGE OF CYGNUS-X-3

We have made observations of the highly scatter-broadened image of Cyg nus X-3, using MERLIN at both 408 and 1692 MHz. At each frequency the shape of the image is close to that of an elliptical Gaussian, but the ellipticity and the position angle are frequency-dependent. We interp ret the variable elongation of the scattering disc as being due to sca le-dependent anisotropy in the scattering medium arising from variatio ns in the direction of the projected interstellar magnetic field. From the scale on which the anisotropy changes we derive a value for the o uter scale of turbulent density fluctuations, r(out) is similar to few X 10(11) km approximately 0.01 pc. Cyg X-3 is the third source for wh ich the scattering disc has been shown to be markedly elliptical for a t least one wavelength. This suggests that anisotropy in the turbulent ISM may be a widespread phenomenon, and emphasizes the importance of extending scattering theory to include its effects on observations. At both frequencies our interferometer data show an excess of correlated flux density at large baselines compared with that expected from a pu re Gaussian image profile. This excess visibility power is due to weak image granulation and was predicted for a power-law spectrum of turbu lence by Goodman & Narayan. From the amplitude of the excess visibilit y power we constrain the exponent in the power-law spectrum to be < 3. 83. This is consistent with several other indications for a 'shallow' spectrum, i.e. an exponent below 4. In the rest of the analysis we hav e assumed an exponent of 11/3, which corresponds to a Kolmogorov spect rum. We employ three different methods to estimate the inner scale r(i n) of the density power spectrum in the scattering screen in front of Cyg X-3: (i) from the wavelength-dependence of scatter-broadening; (ii ) from the visibility as a function of those baselines where the sourc e is moderately resolved; and (iii) from the excess power in the visib ility on baselines where the source is highly resolved. From a combina tion of the results from methods (ii) and (iii) we cautiously constrai n r(in) for this line of sight to lie in the range 10(2) < r(in) < 10( 5.5) km. Future MERLIN and VLBI measurements should be able to place b etter constraints on the spectrum of density fluctuations, using the a nalysis methods we present here. However, in view of the subtlety of t he effects being sought, obtaining data of the requisite quality will always present a significant observational challenge.