Paper TF+PS-ThM5
Filling High Aspect Ratio Features: A Ballistic Transport Model

Thursday, November 13, 2014, 9:20 am, Room 307

The ability to fill a high aspect ratio feature with a thin film material enables the fabrication of many nanoscale devices. Examples include shallow trench isolation or inter-metal dielectric in microelectronics. One approach is to use chemical vapor deposition under conformal coating conditions. However, as film builds up on the sidewalls the width of the feature shrinks and the aspect ratio increases sharply. This often results in incomplete filling, leaving a narrow void or ‘seam’ of low-density material along the central axis. One solution is to taper the feature into a ‘V’ shape, such that uniform deposition causes the apex of the V to move smoothly upwards.

To achieve complete filling, the flux of deposition precursor down the axis of the feature must be sufficient to maintain a uniform growth rate. Precursor transport is typically modeled using the diffusion equation under quasi-static conditions. We show that for high aspect ratio features, the diffusion equation significantly under-estimates the flux of material that is deposited deep in the feature. This occurs because the diffusion formalism assumes a mean transport distance between collisions that is proportional to the feature size. However, in molecular flow some of the transport events occur at glancing angles to the feature sidewall and afford very long flight paths. These events move precursor species to the bottom of the feature, an effect that enhances filling.

We have developed a ballistic transport model based on computing the emission/capture probability between all points on the surface and coding the result as a matrix. Species transport from a starting distribution is found by matrix multiplication to afford the distribution of final positions. We first show how the results of this model compare with the diffusion formalism: the bulk of the transport is similar, but the ballistic model predicts a ‘tail’ of long-range events. We then simulate the filling of V-shaped features as a function of the apex angle and sticking coefficient. The result is a prediction of regimes that can afford complete filling.

Finally, we consider the effect of growth rate saturation under high precursor flux, an effect that is physically significant and vastly improves conformal growth. We derive from the ballistic model the total flux arriving at each position, and self-consistently calculate the effective sticking probability. We simulate the coating profiles on rectangular and V-shaped features and determine that rate-saturated growth conditions, in combination with long-range precursor transport, greatly expand the regime that affords complete filling.