In the single wafer cleaning process, liquid splash during the liquid dispense step can generate droplets that redeposit onto the wafer surface, leading to watermark defects. This study presents a three-dimensional computational fluid dynamics (CFD) model to investigate the mechanisms of liquid splash and its dependence on key process parameters. The model employs the Volume of Fluid (VoF) method and adaptive mesh refinement (AMR) to accurately resolve liquid splash and droplet formation. Simulations were validated against experimental data and successfully reproduced liquid splash and deposition regimes. Parametric studies show that a higher wafer spin speed and nozzle flow rate promotes splash by increasing the kinetic energy of the impinging liquid in the rotating wafer reference frame. A nonlinear dependence on liquid viscosity was observed, where intermediate viscosity promoted splash, while both low and high viscosity suppressed it. Hydrophobic wafer surface and low liquid surface tension enhanced splash. These findings provide mechanistic insights into splash formation and offer quantitative guidance for optimizing single wafer cleaning processes to minimize defect generation.
Molecular dynamics (MD) simulations were performed to investigate the effect of the nanoslit width on the evaporation behavior of isopropyl alcohol (IPA) molecules from nanoslits. The evaporation mass flux of IPA was calculated and found to decrease with decreasing nanoslit width. Since the evaporation rate is related to the saturated vapor pressure of the liquid, the evaporation mass flux of IPA depending on the nanoslit width was analyzed using the Kelvin equation, which describes the variation in saturated vapor pressure of the liquid confined within the slit. A comparison between the MD simulation results and the prediction of the Kelvin equation demonstrates that the change in the evaporation rate from nanoslit can be predicted by the Kelvin equation up to a lower limit of the slit width that can define the width of the liquid–vapor interface of IPA molecules within the nanoslit.
The objective of this study is to investigate the oxide removal and the reoxidation mechanisms occurring on indium phosphide (InP) surfaces in a cleanroom environment after HCl wet treatments. For this purpose, an experimental setup was mounted, tuned and validated consisting of a glove box connected to a X-ray photoelectron spectroscopy (XPS) chamber, enabling quasi in-situ sample transfer after HCl wet treatment without any air exposure. We showed that a dip of at least 3 min in HCl with a concentration of at least 0.5% (0.137 mol/L) effectively removed indium phosphates (InPO x ) from InP surfaces. Reoxidation of HCl-treated InP surfaces was observed after more than 1 min of air exposure.
Recent advancements in artificial intelligence (AI) have increased the need for high-performance, high-capacity semiconductors. Consequently, cutting-edge semiconductor structures have become increasingly complex, requiring highly precise optimization in manufacturing processes, and achieving high yield rates within short periods has become challenging. This study focuses on the use of various yield improvement and process development methods using AI in semiconductor manufacturing and development to address these challenges. Specifically, we report that Computational Fluid Dynamics (CFD) and Multi-Objective Bayesian Optimization (MOBO) can be used in the phosphoric acid etching during the word line formation of 3D NAND Flash, to determine optimal etching conditions and hard parameters for a batch-type wet etching bath without experimental verification using wafers.
Despite the silicon nitride (SiN) etching in both narrow critical dimension (CD) and high aspect ratio feature requires extreme precision for the structural integration of NAND Flash device, pattern dimension-dependent chemical kinetics and thermodynamics cause fluctuation in the SiN etch rate. Here we tackle both wet and gaseous process control methods to achieve the conformal SiN etching in superdeep and narrow pattern. Firstly, we studied the boundary layer and electric double layer in nanoconfined patterns to understand their significant impact on CD-dependent wet etching. Additionally, gaseous etching in Knudsen flow regime has been discussed to address the CD independent etching. Next, we investigated thermodynamic strategies to minimize the accumulation of byproducts during phosphoric acid-based SiN wet etching, which can reduce the etch rate lag at the bottom of superdeep holes. These strategies were experimentally validated, supported by computational calculations. Furthermore, we investigated the improvement in local etch rate lag within the top layer of superdeep holes, with a particular emphasis on the effusion of SiN gaseous etching products. This led to the demonstration of excellent top-bottom uniformity, characterized by a ratio of nearly ~1.
