Time-depending yield degradation at microchip manufacturing
Silicon wafers are processed to create hundreds of dies. During manufacturing, transistors and interconnections are made. At the end of the processing, the upper layers and bumps are fabricated outer connections of a die.
A typical processed wafer is shown below:
Bumps made of the Cooper are located on the very top of each die. Typical bumps are shown below:
The bumps are used for the final test of all dies. The final test is an electrical test that is performed using prob-cards equipped with micro pins. Micro pins touch the bumps that allow them to perform electrical tests: send signals and receive results. Only good dies are taken for the packaging. All dies that fail the test will be scrapped. The process yield depends on the amount of "good" and "bad" dies.
It was revealed that the time between the end of the process and the final test impacts the yield. The longer the dwelling the more dies fail the final test. If the dwelling exceeds hundreds of hours the amount of failed dies becomes dramatically high, which results in the scrapping of the whole wafer.
Time-dependence yield degradation is illustrated in the picture below. Failed dies are marked by blue an red colors.
The end of the process is shown in the picture below:
The bumps are built from Cu on the top of the dielectric layer. Ti thin layer is used to improve the adhesion of the Cu to the dielectric material. The purpose of the operations shown above is to disconnect the bumps.
The process integrates both wet etch and plasma etch operations.
Chem 1 - is the mixture of H2SO4 + H2O2 - classical solution to remove Cu. Chem 1 does not interact with Ti.
Dry etch of Ti is performed with fluorine-containing gas that removes Ti and does not remove Cu.
Chem 2 - is the HF-diluted solution that dissolves Ti and does not interact with Cu and the dielectric material.
The time depending dies degradation is observed.
TRADITIONAL APPROACH
The main reason is an oxidation of the Cu - Proposed solution - to keep the wafers in an inert atmosphere.
This is the wrong concept: even oxidized Cu remains conductive, the experiments with an inner atmosphere did not show any improvements.
The managerial problem solution is to avoid dwelling between the end of the process and the final test: plan processing and perform the final test just after processing.
Applying such a solution increases production costs and is not always viable.
The project aims to analyze the phenomenon and find a creative solution.
Similar projects
The process is related to microelectronics - microchip manufacturing. The purpose of the process is to create a SiO2 layer on the surface of a Si wafer. Equipment: Vertical furnace to heat the wafers in the Q2 atmosphere and perform oxidation on the wafer surface. Process: The oxidation occurs on the front side and on the back side of the wafer Requirements: Create a SiO2 thin layer with a certain thickness and low sigma - low standard deviation of the thickness between the wafers and within the wafer Failure: Wafers from the lower zone have higher thickness and significantly higher within wafer sigma (standard deviation of the thickness within the wafer)
Wet cleaning is widely used in microchip manufacturing. Single wafer equipment is working as follows. A wafer rotates, and chemistry is poured from a movable nozzle. Water rinsing is performed at the end of the process. Loading of a new batch of the chemistry resulted in excursion - a strongly increased amount of defects was observed on the wafer after the processing. The project is dedicated to the failure analysis and creation of innovative solutions.
This project investigates radial non-uniformity of CDO low-k dielectric deposited by PECVD and its possible impact on wafer performance. Final test data indicate that faster dies are mainly located closer to the wafer center, while slower dies are more often found near the wafer periphery. The corresponding RC-delay trend also increases toward the wafer edge. The focus of the project is the capacitance component of RC-delay. CDO is expected to provide a low dielectric constant between BEOL metal lines, but the deposited film may not have uniform properties across the wafer. Near the wafer edge, the process can be influenced by different gas flow, plasma behavior, residence time, temperature, and pumping conditions. This may lead to stronger oxidation and deeper precursor decomposition near the periphery, forming CDO that is more oxygen-rich, less carbon-rich, denser, and closer in behavior to SiO₂. As a result, the dielectric constant k may increase toward the wafer edge, causing higher interline capacitance, higher RC-delay, and lower die speed. The project uses Functional Modeling to understand how PECVD chamber components and process conditions affect CDO film properties. The goal is to identify the main functional and problematic interactions and propose directions for compensation, such as radial gas-flow tuning, multi-zone showerhead design, edge compensation flow, and local plasma or temperature control.