2014MNRAS.442..647D

We use MERLIN, VLA and VLBA observations of Galactic H i absorption towards 3C 138 to estimate the structure function of the H i opacity fluctuations at au scales. Using Monte Carlo simulations, we show that there is likely to be a significant bias in the estimated structure function at signal-to-noise ratios characteristic of our observations, if the structure function is constructed in the manner most commonly used in the literature. We develop a new estimator that is free from this bias and use it to estimate the true underlying structure function slope on length-scales ranging from 5 to 40 au. From a power-law fit to the structure function, we derive a slope of 0.81^+0.14_-0.13, i.e. similar to the value observed at parsec scales. The estimated upper limit for the amplitude of the structure function is also consistent with the measurements carried out at parsec scales. Our measurements are hence consistent with the H i opacity fluctuation in the Galaxy being characterized by a power-law structure function over length-scales that span six orders of magnitude. This result implies that the dissipation scale has to be smaller than a few au if the fluctuations are produced by turbulence. This inferred smaller dissipation scale implies that the dissipation occurs either in (i) regions with densities ≳ 10³/cm3 (i.e. similar to that inferred for `tiny scale' atomic clouds or (ii) regions with a mix of ionized and atomic gas (i.e. the observed structure in the atomic gas has a magnetohydrodynamic origin).