Measurements show that the vicinity of powered ion cyclotron range of frequency (ICRF) antennae is biased positively with respect to its environment [J. Gunn et al., Proc. 22nd IAEA Fusion Energy Conference, Geneva 2008, EX/P6-32]. This is attributed to radio-frequency (rf) sheath rectification. The radial penetration of these direct current (dc) potentials from ICRF launchers into the tokamak scrape-off layer (SOL) determines the power deposition on the walls and especially on the antenna structure, which is a key point for long time clean discharges. Within independent flux tube models of rf sheath rectification the radial penetration of dc potentials is determined by the skin depth x0=c/ωpe for the slow wave. When self-consistent exchange of transverse rf current is allowed between neighboring flux tubes, such a structure can be broadened radially up to a characteristic transverse length L. Broadening arises as soon as L>x0. A linear modeling of the process gives a first evaluation of the theoretical length L=(L∥ρci/2)1/2. Within the “flute assumption,” it scales with the length L∥ of open flux tubes and the ion Larmor radius Ωci. This trend has been confirmed by nonlinear fluid simulations using the SEM code taking into account nonlinearities of the sheath dynamics. Parametric regimes are outlined where broadening or nonlinearity arise. Langmuir probe measurements on Tore Supra suggest that the observed broadening is lower than predicted by the code. This suggests that actual rf current exchanges probably do not occur over the whole length of magnetic field lines but only on a fraction of it. This “effective parallel magnetic connection length” L∥eff is estimated from the measurements. The model is finally applied to several potential maps generated by an ITER antenna, with different plasma parameters depending on possible SOL scenarios in ITER, and “reasonable assumptions” about L∥eff. It comes out that L ranges between 1 and 10 cm depending on local L∥eff and on typical ITER plasma parameters.