Project Introduction:
In the production of semiconductors, silicon wafers act as the foundation upon which integrated circuits are constructed. The wafer processing unit typically includes a variety of tools and procedures required for the creation of ICs as shown in the Figure
Wafers go through cleaning procedures to get rid of any debris or pollutants that could degrade the quality of the integrated circuits. Contamination of wafer is the biggest enemy for the chip manufacturers, this can happen by the dust particles & humidity present in the atmospheric air, which is the reason they maintain clean room environment.
Process Statement
# Loadlock system is a semiconductor wafer processing machine. Nitrogen flows into the loadlock system through a fan filter unit and sucked back through a return duct. To maintain a constant vacuum gauge pressure inside the loadlock system a vacuum pressure outlet is used. Inside the loadlock system the wafers are baked in a very high temperature. Wear occurs from contact areas in certain moving parts such as doors and seals inside the system. These comprise of breach, FIMS door, End effector. Small particles are released into the flow from different materials such as silicon, Alumina, Aluminum and stainless steel etc.
Problem Statement
# These particles contaminate the wafer by depositing on them. Our customer approached us In order to understand how the contamination occurs and to completely avoid particles depositing on the wafer.# With decades of experience in computational fluid dynamics industry and highly talented professionals, we were able to help our customer by creating a digital twin of the existing model and conducting a computational fluid dynamics study on the model.
Task Executed:
# Preparing the geometry and extracting the fluid volume from the geometry is the first step towards the computational fluid dynamics study. A fluid volume is nothing but the total volume inside the loadlock system, the nitrogen gas can enter and flow. By extracting the fluid volume, we can clearly understand and determine how its flow characteristics is going to be. This will be an effective way to validate the results. we considered three different cases for the simulation # Wafers being loaded in the system# During baking process# Wafer unloading # Fluid volumes were extracted for all three cases and verified with the customer before proceeding further. Once the fluid volume is extracted, the fluid volume is checked and resolved of any geometry issues that could cause convergence issue during simulation.# After the geometry extraction, the fluid volume was meshed using a fluid dynamic pre-processing software with high quality hexahedral mesh with not less than 0.1 orthogonal quality. Then the flow and temperature boundary conditions were applied on the meshed body.# Finally, the turbulence model and convergence criteria were updated, and the model was simulated.
Problems Encountered and Resolutions:
# During the Geometry preparation we encountered lot of problematic geometries such as surface bodies small faces and overlapping bodies. While creating fluid volumes we make use of Boolean geometry operators to isolate fluid regions from solids, but surface bodies makes it difficult to extract the fluid volume. To resolve this issue, we converted the surface bodies into solid bodies by patching and closing the surfaces. Once the solid bodies were created, we proceeded with geometry simplifications.# Once the solid bodies are created then comes the geometry simplification phase. Not all the curves of the geometry is important. Lot of curved regions and small faces, results in Increased cell count and thus increases the solution time. To over come this problem small fillets and faces under 5 mm were removed and replaced with corners. Complicated features and components that are not in the flow path and that does not affect the flow were either simplified or removed.<# Small holes that are below 5mm and has no effect on the flow were removed from the geometry.# Fluid volume extraction comes after the geometry preparation. The fluid volume is extracted from the model using Boolean operations and the fluid volume is again checked for faces that are below 5mm. Few small and bad faces along with the irregular corner were identified after fluid volume is extracted. These geometric features will cause element error and convergence error, so we took precautions and simplified all irregular geometry features without affecting the total flow volume.# The extracted fluid volume is then meshed with the help of powerful meshing tool. During the meshing process some lower quality elements were generated. Those elements were removed my optimizing and remeshing those regions
Results
The prepared model was solved in a CFD simulation software and results were extracted. By studying the results, the flow behavior and contamination particle’s flow paths were found, and the cause of particle deposition was identified.
Problems:
# Recirculation of particles increased particle resident time in the system.# Particles have tendency to settle on wafers in Foup where there is no flow.
Improvements were made to the load lock system to avoid stated problems.
Improvements:
# Flow speed optimized to avoid recirculation.# Diverters were provided in necessary zones to ensure no stagnation zones.# Ideal placement of Intake duct to ensure maximum coverage of flow volume.# Particle hitting the wafers were brought to Zero
Load Lock system flow Optimization
Initial Velocity Vectors
Particle Residence time
Benefits:
Need help with Particle Tracking and Inspection in Semiconductor Equipments? Reach out to us
We’re always here to help, Explore AES’s expertise in new product development! Step into the future with AES – Your End-to-End, Product Development Partner.
