Nano-Scale Feature Profile Modeling of Plasma Material Processing

In this thesis, a nano-scale model of feature evolution is developed and used to illustrate several of the physical mechanisms at work during state-of-the-art plasma etching of semiconductor materials. A new model for plasma etching of silicon, silicon dioxide, and silicon nitride in fluorocarbon containing plasmas was developed. This new technique uses physically based models to take into account the transport of reactive species and ion energy through the finite thickness fluoropolymer overlayer which develops during etching to the etch interface beneath. This etch model was used to explore the underlying physical mechanisms behind the aspect ratio dependence of etch rate in high aspect ratio etching, as well as the wafer scale uniformity of plasma etching. The results presented here indicate that, for a wide range of applications, diffusive neutral transport of reactive species is responsible for the aspect ratio dependence of etch rate. Etching silicon in a chlorine containing plasma was shown to have an etch rate which depends linearly on ion flux for a wide range of conditions. Because of this dependence, any non-uniformity of ion flux over the target wafer results in non-uniform etching. Both of the issues described above can be addressed by introducing a technique referred to as atomic layer etching (ALE). This technique separates the reactive radical and ion fluxes in time, by dividing the etching reaction into two self-limited half step reactions. Only by cycling between the radical passivation phase and the ion etching phase is etching observed. The conditions under which this system is effective was explored using the ALE of silicon as a prototypical example. It was found that ideal ALE can provide total independence from aspect ratio etch rates, and perfect wafer scale uniformity. By introducing non-ideal fluxes that are more representative of conditions found in a typical plasma etching reactor, both of these benefits were reduced, but not eliminated. Applying the ALE technique to SiO2 is more difficult than bare silicon, as chlorine does not effectively passivate the surface. Instead a fluorocarbon containing plasma is employed. Unlike using a pure halogen gas, the inclusion of fluorocarbon species was found to result in non-self-limited passivation of the surface, with a fluoropolymer layer forming having a thickness that depends on the passivation time. Results indicate that this fluoropolymer layer functions as a fuel for continued etching during the ion bombardment phase, linking the etch per cycle to the passivation time. The selectivity of the ALE of SiO2 over Si3N4 was also studied. It was found that very high selectivity can be obtained in the steady state, but a transient period exists at the beginning of etching during which the selectivity is low. The implications of this transient etching were explored using a self-aligned contact etch as an example application. It was found that thicker passivation in each cycle improved selectivity, but at the trade-off of causing more tapered features, referred to as critical dimension loss.PhDElectrical EngineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttps://deepblue.lib.umich.edu/bitstream/2027.42/146025/1/chuard_1.pd