Model-Based Optimization of CMP Process Parameters for Uniform Material Removal Selectivity in Cu/Barrier Planarization

Abstract

Chemical mechanical planarization is a process of achieving planar surfaces in the semiconductor manufacturing industry. The planarization of a surface is achieved by material removal from the wafer surface. The material removal depends on material properties and the process input parameters. Several studies have investigated the role of slurry chemistry to achieve a certain material removal selectivity for different materials. Here we propose a methodology of achieving a planar patterned surface of Cu/Mn/MnN using a model-based optimization for mechanical process parameters including applied force, slurry solid concentration, and abrasive particle size. The methodology has been developed via optimization using a genetic algorithm. The proposed methodology suggests that a lower downforce is the key parameter to achieve the desired material removal selectivity and planarity. The first part of the study suggests a low material removal rate (MRR) to achieve a lower standard deviation in MRR. The second part investigates the standard deviation in the thickness removed in the average time needed to remove a known thickness of the materials under consideration. It has been found that the application of lower downforce can also minimize the standard deviation in the thickness removed and a planar patterned surface can be achieved.

Surface Morphology Evolution during Chemical Mechanical Polishing Based on Microscale Material Removal Modeling for Monocrystalline Silicon

“…In the mechanic material removal model, we focused on the relationship between applied load and the morphology of the scratch. To simplify the problem, scratching was assumed to be the accumulation of Hertzian cracks along the movement direction of abrasive particles [14,15,19,20]. First, we considered single abrasive particles in contact with the pad and the workpiece.…”

“…With a molecular removal depth of 0.04 nm under chemical effects, the calculated MRR was close to the experimental value. Akbar et al [19,20] modeled the chemical effects during CMP as the diffusion of water into the workpiece. The moving reacted and unreacted sites have different hardness linked to plastic deformation and surface profile for scratching in the next polishing iteration.…”

Modelling and optimization of surface roughness in chemical mechanical polishing based on DNN-GA

“…6,7 During the CMP process, the procedure is so complicated that the empirical model and physical model cannot clarify the process thoroughly. [8][9][10] Due to the above reasons, how to build a model to study the relationship between parameters and surface quality quantitatively and how to optimize the parameters based on the model have always been worth-solving problems in CMP process. As a burgeoning artificial intelligence technology, machine learning methods, such as Support Vector Regression (SVR), Decision Tree (DT), Artificial Neural Network (ANN) can analyze the data quickly, by building a complex relationship between inputs and output(s) parameters.…”