Column Density Probability Distribution Functions in Turbulent Molecular Clouds: A Comparison between Theory and Observations

Abstract

The one-point statistics of column density distributions of turbulent molecular cloud models are investigated and compared with observations. In agreement with the observations, the number N of pixels with surface density S is distributed exponentially N(S)=exp(-S/S0) in models of driven compressible supersonic turbulence. However, in contrast to the observations, the exponential slope defined by S0 is not universal but instead depends strongly on the adopted rms Mach number and on the smoothing of the data cube. We demonstrate that this problem can be solved if one restricts the analysis of the surface density distribution to subregions with sizes equal to the correlation length of the flow which turns out to be given by the driving scale. In this case, the column density distributions are universal with a slope that is in excellent agreement with the observations and independent of the Mach number or smoothing. The observed molecular clouds therefore are coherent structures with sizes of order their correlation lengths. Turbulence inside these clouds must be driven on the largest scales, if at all. Numerical models of turbulent molecular clouds have to be restricted to cubes with sizes similar to the correlation lengthscale in order to be compared with observations. Our results imply that turbulence is generated on scales that are much larger that the Jeans length. In this case, gravitational collapse cannot be suppressed and star formation should start in molecular clouds within a dynamical timescale after their formation.