UV Assisted Densification of Perhydropolysilazane (PHPS) Based Spin-on Glass in High Aspect Ratio Gap Fill Structure

Abstract

Listen

Translate article

As device dimensions scale, defect-free, robust gap filling in high aspect ratio structures such as STI and inter-gate spaces becomes increasingly challenging. This is due to very high aspect ratio (AR) of gap fill structure (AR > 7:1), shrinking CDs (<20nm) and nearly vertical sidewalls gate profiles (>88deg). The Spin-On Dielectric such as (SOD) PerHydroPolySilazane (PHPS) is preferred choice as it provides defect-free and structure agnostic gap fill (1). As coated, the PHPS film has a long polymeric chain of hydrogen terminated -Si-N-Si- network with high % of N and terminal H. The film as-coated has very high etch rate in HF based chemistry. The film structure needs to be fully converted into a dense, cross-linked network of SiO2 via silanol condensation reaction during steam anneal carried out at ≤600C. However, as the critical dimension reaches below 20nm, the inability of steam to diffuse to the bottom of high aspect ratio features results in oxide density gradient leaving very weak oxide near the bottom of trench. High temperature anneal (> 600⁰C) is not feasible for inter-gate gap fill due to its impact on junction. Non-thermal approaches are needed to improve the oxide quality inside high aspect ratio features. The use of monochromatic UV radiation for low temperature conversion of PHPS film is well documented (2). Here, broadband UV (200nm-450nm) was explored to assess its effectiveness in improving In-Trench oxide density post steam anneal. The effect of three UV conditions (UV-A, UV-B, UV-C) and modified steam anneal condition on structure and composition of as-coated PHPS film was assessed using FTIR (Fig. 1) and XPS Depth Profiling (Fig. 2). In blanket experiments, the densification effect by UV of as-coated PHPS films was assessed post-steam anneal by measuring through-thickness wet etch rate (Fig. 3). This was done by sequential etching of annealed films in 100:1 dHF chemistry (400s) followed by thickness measurement. The average relative (to thermal oxide) wet etch rate of blanket oxide films exposed to various UV conditions and modified anneal conditions is listed in Table 1. FTIR and XPS depth profiles (Fig. 1) of UV treated PHPS films show significant modulation with UV conditions. The more aggressive the UV conditions, the lower the overall %N. Data also suggests that UV treated films have progressively higher O2 and reduced N at more aggressive UV conditions. This is because the effective breaking of Si-H, N-H and SI-N bonds in UV treated films which loosen the backbone structure, thereby facilitating oxygen/moisture diffusion relative to as-coated film. Low-temperature O2 soak step prior to steam anneal was found to be very effective in achieving denser oxide post-anneal for UV cured films. However, due to limited O2 diffusivity of oxygen in as-coated PHPS film due to denser network for non-UV treated PHPS films, O2 soak prior to steam anneal is more effective only at higher temperature. Optimized UV and anneal conditions from the blanket study were further explored on integrated wafers consiting of 7nm dimension inter-gate gap fill. These structures were filled with PHPS and treated with different UV conditions prior to annealing and planarization. In-Trench wet etch rate was obtained by sequentially measuring In-trench wet etch rate of oxide similar to the blanket film etch study. In each cycle, the wafer was exposed for 60s in 100:1 HF chemistry. Optical Scatterometry was used to measure In Trench oxide recess amount per etch cycle as well as cumulative oxide recess after 5 sequential etch cycles (5X 60s HF (100:1)) (fig. 4). Table 2 displays average In Trench wet etch rate comparison between the UV treated films and control cell post-steam anneal. The blanket wet etch rate data in Table 1 suggest 18% lowering just by modifying anneal for non-UV cell. Insertion of UV between spin coating and anneal helps further reduction in blanket wet etch rates (UV-A > UV-B > UV-C). UV-C condition had 34% reduction in blanket etch rate. However, blanket etch rate trend with UV condition did not quite translate to the In-Trench wet etch rate. While the overall in-trench wet etch performance for UV treated cells was better than control, the UV-A and UV-B conditions had lower in-trench WERR than UV-C condition. These results will be explained in the light of moisture sensitivity of UV exposed PHPS films as a function of UV conditions and its impact on overall wet etch rate post-anneal. References Trivedi et al., J. Vac. Sci. Technol. B, Vol 27, No. 6, pp 3145-3148Prager et al., Chem. Eur. J., 2009, vol. 15, pp 675-683 Figure 1