skip to main content

Wafer Scrap Reduction in a 300mm Logic Foundry

An Applied Global Services Case Study

Dong Chae, Paul Turnbull and Marine Zhang

Device manufacturers face ongoing challenges to control variables, maintain productivity and improve their operations. Tool audits that benchmark a tool to best-in-class data, combined with sophisticated data analytics, can help manufacturers meet these challenges. Here is how Applied Materials used these methods to help a customer dramatically reduce wafer scrap from one toolset.


One of our foundry customers experienced a high scrap rate in gate stack fabrication, caused by a group of Applied Materials Centura tools equipped with ISSG, DPN Plus and PNA chambers. This tool group had one of the fab’s highest scrap rates, and excessive parts replacement costs were required as a result. The customer asked the Applied Materials FabVantage Consulting Group to identify solutions to cut scrap by half.


  • A FabVantage tool assessment identified root causes using data analytics to speed diagnostics. As a result, faulty parts were replaced, equipment settings were set to their BKM values, recipe adjustments were made, and correct PM procedures were implemented.


  • The customer exceeded its goal. Wafer scrap from the tool group was reduced to <0.2% and spending on parts was reduced by 4x. In fact, a tool in the group went from being one of the fab’s worst-performing tools to being the very best one.

A 300mm logic foundry was experiencing significant levels of wafer scrapping from a key gate stack fabrication tool set. The tools were Applied Materials Centura systems equipped with in situ steam-generated oxide (ISSG), decoupled plasma nitridation (DPNPlus) and postnitridation anneal (PNA) chambers.

The problem manifested itself as a dense streak of particles on the processed wafers (see figure 1). The customer was unable to identify the root causes, and hired Applied Materials’ FabVantage Consulting Group to perform an assessment.

Figure 1. Typical pattern of particle contamination seen on 0.6% of the processed wafers from one Centura tool at a 300mm logic foundry prior to a FabVantage tool audit.

Applied’s FabVantage consultants combine extensive fab and tool expertise with state-of-the art modeling and analysis tools. They engage in collaborative relationships with customers to help solve some of the industry’s most difficult manufacturing challenges.

Projects begin with a discovery session to identify key customer goals and challenges. Then a benchmarking process is performed to compare customer data with best-in-class data from Applied’s knowledge base. Subsequently, a formal assessment is conducted to define the problem and develop a strategy to achieve improvements in targeted areas.

FabVantage consultants conducted a comprehensive hardware and process audit at the foundry, which included:

  • Comparison of the product recipe to the initial Applied Materials recipe
  • An audit of the fab’s actual preventive maintenance (PM) procedure versus Applied’s recommended PM procedure
  • Checks of the settings of the equipment constants
  • Comparisons of the parts being used on the tool with those on Applied’s recommended parts lists

The FabVantage team used data analytics to quickly pinpoint the causes of tool misprocessing. Analytics tools included Applied’s E3 automation and equipment engineering system; a knowledge base of sensor and event collection plans for each chamber; and data visualization routines.

The team also performed sensor trace analysis to compare sensor values on good chambers versus the corresponding sensor values on bad chambers, to identify obvious abnormalities.

A total of 48 issues were identified. Several were related to damaged or non-best-known-method (BKM) parts, including a worn non-BKM throttling gate valve with a damaged coating; a scratched robot blade that was not levelled; a chipped quartz ring; a single-wafer loadlock (SWLL) with dirty O-rings and particles in the loadlock; dirty lift pin tubes; worn pads on the tips of the factory interface robot blades; and scratched reflector plates.

In addition, sensor data visualization revealed the following issues: excessive pressure oscillation; N2 overshoot in the DPN chamber; timing variations for pressure stabilization across chambers; and varying pump-down times across different chambers.

Finally, the FabVantage team’s audit showed that recommended PM procedures had not been followed.


Key among the findings was that the DPN chamber was unable to maintain a stable pressure. Sensor trace analysis identified pressure undershoot and spiking, as shown in Figures 2a and 2b.

Figure 2. Graphs showing DPN chamber pressure instability in the Applied Centura tool group. Figure 2a shows a 5x faster pressure drop in one chamber, which likely caused wafers to pop. Figure 2b shows pressure spiking due to faulty pressure control and throttling gate valves.

A faulty pressure control valve caused the DPN chamber pressure to overshoot. Additionally, when RF power was turned on, the chamber pressure and foreline pressure oscillated because the throttling gate valve was unable to maintain a stable pressure.

The particle streak shown in figure 1 typically might be assumed to be the result of faulty wafer placement. That normally would be a robot or local center finder (LCF) issue, but that was not the case here. Instead, the pressure oscillation caused the wafer to pop, which in turn led the LCF to react by recalculating as though there had been a wafer placement error.

This issue was discovered by overlaying the pressure oscillation data with the LCF correction distance, and observing that the LCF started to correct at the same time the pressure spiked or dropped.


Once the pressure control valve and throttling gate valve were replaced, the wafer popping stopped and the LCF stopped recalculating. Figure 3 shows the LCF distance before and after replacement of the valves. Figure 4 shows DPN chamber pressure stability before and after replacement of the valves.

Figure 3. Data plot shows the LCF correction distance before and after replacement of faulty pressure control and throttling gate valves. The data to the left of the arrow shows a large, varying LCF distance prior to replacement of the faulty valves. Data to the right of the arrow shows a short, uniform LCF distance after the valves were replaced.

Figure 4. DPN chamber pressure before and after replacement of faulty pressure control and throttling gate valves, showing a much more stable environment after they were replaced.

The combination of state-of-the-art FabVantage data analytics, supported by information from the Applied knowledge base, was instrumental in the rapid troubleshooting of the pressure control issue.

The knowledge base enabled the FabVantage consultants to quickly generate data collection plans and identify priority sensors. The LCF position does not have a sensor associated with it, so the team devised a data transformation to compute the LCF correction distance.

For easy visual identification of problems, the team wrote scripts to create charts that allow values from multiple sensors and chambers to be overlayed.


After implementing all FabVantage recommendations, defect counts fell on all tools in the gate stack module, as shown in figure 5.

Figure 5. The monthly defect count trended downward after implementation of all Applied FabVantage recommendations. Each color represents a different Applied Centura gate stack tool.

Wafer scrap fell to <0.2%, substantially exceeding the customer’s goal. As an additional benefit, customer spending on parts was reduced by 4x.

For additional information, contact