Filling high aspect ratio trenches by superconformal chemical vapor deposition: Predictive modeling and experiment

Complete filling of a deep recessed structure with a second material is a challenge in many areas of nanotechnology fabrication. A newly discovered superconformal coating method, applicable in chemical vapor deposition systems that utilize a precursor in combination with a co-reactant, can solve this problem. However, filling is a dynamic process in which the trench progressively narrows and the aspect ratio (AR) increases. This reduces species diffusion within the trench and may drive the component partial pressures out of the regime for superconformal coating. We therefore derive two theoretical models that can predict the possibility for filling. First, we recast the diffusion-reaction equation for the case of a sidewall with variable taper angle. This affords a definition of effective AR, which is larger than the nominal AR due to the reduced species transport. We then derive the coating profile, both for superconformal and for conformal coating. The critical (most difficult) step in the filling process occurs when the sidewalls merge at the bottom of the trench to form the V shape. Experimentally, for the Mg(DMADB)₂/H₂O system and a starting AR = 9, this model predicts that complete filling will not be possible, whereas experimentally we do obtain complete filling. We then hypothesize that glancing-angle, long-range transport of species may be responsible for the better than predicted filling. To account for the variable range of species transport, we construct a ballistic transport model. This incorporates the incident flux from outside the structure, cosine law re-emission from surfaces, and line-of-sight transport between internal surfaces. We cast the transport probability between all positions within the trench into a matrix that represents the redistribution of flux after one cycle of collisions. Matrix manipulation then affords a computationally efficient means to determine the steady-state flux distribution and growth rate for a given taper angle. The ballistic transport model predicts a deeper position for the peak of the super-conformal growth rate than the diffusion-reaction model, and successfully explains the observation of complete filling. These models can be used to predict the behavior of any system given a small set of kinetic coefficients to describe the growth rate.