Research On Multi-objective Optimization Design Of Plasma Etching

In the process of chip preparation technology, the application of plasma etching technology to generate demand pattern is one of the most important process. Using experimental means to gain experience is time-consuming and costly, and it strongly rely on the experience of the experimenters.In order to overcome these shortcomings, it is very necessary to apply numerical method to simulate the plasma etching process. Besides, in order to reduce the blindness brought by simple measuring method, based on the numerical simulation, to optimize the evaluation index of the plasma etching morphology has a certain practical significance. In this paper, the process of mixed gas C4F8/AR etching SiO2 is numerically simulated by using CFD software. Process parameters in the macro chamber include RF power, chamber pressure, gas ratio and plate voltage. And structural parameters of micro devices includes the thickness of photoresist, photoresist angle and slot width. The sensitivity of process parameters in the macro chamber and structural parameters of micro devices to evaluation index of etching results, including etch rate and transverse etching width, is studied. The study find that the effects of various parameters on the etching rate is in conformity with the physical and chemical mechanism of the plasma etching technology, and the effects of each parameter on the etching rate and the lateral etching width is different, even opposite. The macroscopic and microscopic parameters, including plate voltage, chamber pressure, photoresist angle and notch width, sensitive to etching rate and lateral etching width are chosen as design variables in this paper. And multi-objective optimization is conducted on etching rate and lateral etching width on micro scales and macro-micro scales. By applying full factorial experiment design method and the optimal Latin hypercube test method, the approximate model is established, and optimization design is carried out.First of all, the single objective optimization is designed on the etching rate and the lateral etching width respectively. Based on response surface method, radial basis function neural network model and Kriging model, different surrogate models are constructed and the genetic algorithm and particle swarm optimization are used to construct the optimization design process to optimize the solution. In terms of the optimization of microstructure parameters of etching rate, the genetic algorithm is applied to the model based on response surface methodology, which is proved to be efficient. In terms of macro and micro scale optimization of transverse etching width, the particle swarm optimization algorithm is applied to the Kriging model, which is proved to be reliable.Secondly, considering that the etching process is not only concerned with a single goal, etching rate and etching profile quality are very important. Therefore, the etching rate and the lateral etching width are chosen as the objective functions, and the microscopic structure parameters are chosen as design variables. Multi objective optimization design for the above two objective functions is developed. It is found that archival micro genetic algorithm (AMGA) constructed by radial basis function neural network model and second generation non dominated sorting genetic algorithm (NSGA-?) are better. The parameters of etching process and microstructure are improved by this studied.Finally, the cellular algorithm method is applied, which preliminarily develops the three-dimensional numerical simulation of the profile evolution of plasma etching. Slot electric field, ion trajectory, the interaction between ions and etched surfaces and the other items are simulated, and the simulation results can be used to evaluate the device surface topography and roughness.