A model is presented to describe the hole opening in ablative optical recording. The model accounts for changes in geometry of the sensitive film, for adhesion of the film to the substrate, for surface tension gradients in the sensitive film and for evaporation or sublimation at the film-substrate interface. The changes in geometry during hole opening are shown to be responsible for an energy barrier Fact=10-16 to 10-15 J, corresponding to 1.5-6J/cm3 which is much too large to make thermal activation of hole opening possible. Fact is reduced by surface tension gradients in the liquid film and/or evaporation at the filmsubstrate interface. The predictions of the model are compared with laser recording experiments performed with Te-alloy films on various polymer sublayers. Experiments and model show that the threshold energy Eth for hole opening (and Fact) can be reduced by (i) lowering the surface tension and/or polarity of the polymer sublayer, (ii) reducing the dynamic work of adhesion, e.g. by decreasing the crosslink density or the mean molecular weight of the sublayer, and (iii) increasing the concentration of low molecular weight material in the sublayer, which reduces the adhesion of the film and stimulates evaporation at the interface.