An ion cyclotron resonance heating (ICRH) antenna system must launch radio frequency (RF) power with a wavenumber spectrum which maximizes the coupling to the plasma. It should also ensure good absorption while minimizing the wave interaction with the plasma edge. Such interactions lead to impurity release, whose effect has been measured far from the antenna location (Klepper et al. 2013; Wukitch et al. 2017; Perkins et al. 2019) and can involve the entire scrape-off layer. In the normal heating scenario, for which the frequency of the waves launched by the antenna is larger than the ion cyclotron frequency of the majority ions $\omega > \omega _{\textrm {ci},\textrm {maj}}$ , release of impurities due to ICRH can be affected by minimizing the low $|k_{\parallel }| < k_0$ power spectrum components of the antenna. Impurity release can be the result of low central absorption of the waves or power transfer from the fast to the slow wave due to the presence of a confluence in the plasma edge. In ASDEX Upgrade (AUG), a reduction of heavy impurity release by ICRH in the plasma was qualitatively well correlated to the parallel electric field and RF currents flowing around the antenna (Bobkov et al. 2017). In this article, we first show a correlation between the reduction in impurity release by ICRH in AUG and the rejection of the low $|k_{\parallel }| < k_0$ region of the antenna power spectrum. We show that the same correlation holds for results obtained in the Alcator C-Mod tokamak. Finally, using this idea, we reproduce ICRH induced impurity release behaviour in a not yet published experiments of JET, and make predictions for ITER and DEMO.