Domain walls form at phase transitions which break discrete symmetries. In a cosmological context they often overclose the universe (contrary to observational evidence), although one may prevent this by introducing biases or forcing anisotropic evolution of the walls. In a previous work [Correia {\it et al.}, Phys.Rev.D90, 023521 (2014)] we numerically studied the evolution of various types of biased domain wall networks in the early universe, confirming that anisotropic networks ultimately reach scaling while those with a biased potential or biased initial conditions decay. We also found that the analytic decay law obtained by Hindmarsh was in good agreement with simulations of biased potentials, but not of biased initial conditions, and suggested that the difference was related to the Gaussian approximation underlying the analytic law. Here we extend our previous work in several ways. For the cases of biased potential and biased initial conditions we study in detail the field distributions in the simulations, confirming that the validity (or not) of the Gaussian approximation is the key difference between the two cases. For anisotropic walls we carry out a more extensive set of numerical simulations and compare them to the canonical velocity-dependent one-scale model for domain walls, finding that the model accurately predicts the linear scaling regime after isotropization. Overall, our analysis provides a quantitative description of the cosmological evolution of these networks.

We study the evolution of various types of biased domain wall networks in the early universe. We carry out larger numerical simulations than currently available in the literature and provide a more detailed study of the decay of these networks, in particular by explicitly measuring velocities in the simulations. We also use the larger dynamic range of our simulations to test previously suggested decay laws for these networks, including an ad-hoc phenomenological fit to earlier simulations and a decay law obtained by Hindmarsh through analytic arguments. We find the latter to be in good agreement with simulations in the case of a biased potential, but not in the case of biased initial conditions.