Insight into the removal and reapplication of small inhibitor molecules during area-selective atomic layer deposition of SiO2

As the semiconductor industry progresses towards more complex multilayered devices with ever smaller features, accurately aligning these layers with respect to each other has become the bottleneck in the advancement to smaller transistor nodes. To avoid alignment issues, area-selective ALD can be employed to deposit material in a self-aligned fashion. Previously, we demonstrated area-selective ALD of SiO2 in three-step (i.e. ABC-type) ALD cycles comprising an acetylacetone (Hacac) dose (step A), a bis(diethylamino)silane (BDEAS) precursor dose (step B), and an O2 plasma exposure (step C). In this work, the mechanisms of the removal and reapplication of the inhibitor molecules during area-selective ALD were studied, with the aim of enhancing the selectivity of the process. In situ infrared (IR) spectroscopy shows that the O2 plasma exposure does not completely remove the adsorbed Hacac species (i.e. acac adsorbates) at the end of the cycle. The persisting species were found to contain fragments of Hacac molecules, which hinder subsequent inhibitor adsorption in the next ALD cycle, and thereby contribute to a loss in selectivity. Alternatively, it was found that
an H2 plasma is able to completely remove all acac species from the surface. An improvement in selectivity was achieved by using a four step ALD cycle that includes an H2 plasma step, allowing the nucleation delay to be prolonged from 18 ± 2 to 30 ± 3 ALD cycles. As a result, 2.7 ± 0.3 nm SiO2 can be deposited with a selectivity of 0.9, whereas only 1.6 ± 0.2 nm can be achieved without the H2 plasma step. This work shows that the addition of a dedicated inhibitor removal step before the reapplication of the inhibitors can significantly improve the selectivity.