Invited Paper PS-TuM3
Surface Reaction Control for BEOL Application

Tuesday, November 1, 2011, 8:40 am, Room 201

Increasingly there are more challenges of controlling the plasma processes for BEOL integration, which include the Cu/low-k interconnects, contact hole etching, etc. The presentation will give an overview of the surface reaction control during plasma processes. The main focus is damage reduction and suppression of process fluctuations.

Reducing the damage to low-k dielectrics caused by plasma exposure is one of the key issues. O₂-based plasma has been widely used for ashing the photoresist on low-k SiOCH. H₂-based plasma is employed for the etching of organic low-k film as well as the ashing of photoresist. The origin of damage generation is classified by ions, radicals, and UV/VUV radiation. It was found that not only the ions but also the synergy of radicals and UV/VUV radiation cause a significant amount of damage in the SiOCH, measured by using the pallet for plasma evaluation (PAPE).¹ Hence, precise control of incident ions, radicals, and UV/VUV radiation is required for controlling the surface reactions.

In addition to the precise control of incident species, the optimization of subsequent processes (wet treatment, annealing, etc.) are very important to obtain sufficient electrical yields and reliabilities. The remaining damaged layer after wet treatment degrades the Cu and SiOCH surface and corrupts the interface between the Cu and barrier metal. For instance, the desorbed H₂O from the SiOCH damaged layer causes the oxidation of barrier metal, which results in a shorter EM lifetime. The surface modification during SiCN and SiN etching by CH_xF_y-based plasma and its impact on the electrical yields will also be discussed.

The suppression of fluctuation is also required in the advanced interconnects. Although statistical prediction is one of the approaches to realize stable processes, there is a limitation of prediction accuracy for purely statistical predictions due to the lack of physical models. Thus, the combination of statistical and physical models for highly accurate prediction has become an emerging trend in mass production. We will demonstrate a novel statistical etch rate prediction model by considering the fluctuation caused by the plasma-wall interactions.²

The developed prediction model is one of the approaches to realize stable processes. Changes in the spatial distribution of reactive species in the etching chamber are, however, very difficult to detect in current mass production tools. Thus, greater progress of the in-situ monitoring tools and prediction methods based on the physical model (simulation) are strongly required in the near future.

1. S. Uchida et al., JAP 103, 073303 (2008).

2. M. Fukasawa et al., JJAP 48, 08HC01 (2009).