More data improves throughput, helps extend the life of equipment.
Third-party sensors are being added into fab equipment to help boost yield and to extend the life of expensive tools, supplementing the sensors that come with equipment used in fabs.
The data gleaned from those sensors has broad uses within the fab. It can measure process module performance, identify defect sources, and alert fabs of impending equipment failure. And when coupled with machine learning, the data from all sensors can improve process control and improve uptime.
While equipment makers build numerous sensors into their products, they cannot identify all of the data that might be needed for yield improvement and equipment maintenance. This is especially true for each increase in wafer size or improvement in a manufacturing process. It’s why aftermarket sensors are so important. These include sensors that measure incoming and outgoing fluids and gases, which can provide insightful data about subtle equipment failures that need attention. Instrumented wafers also speed up equipment set-up and provide additional information that can be used for maintenance and process improvement.
Vidya Vijay, senior product line manager for WaferSense & MRS at Nordson Test & Inspection, pointed to several reasons for adding third-party sensors, including:
New sensors also generate data that needs to be integrated into data management systems in the fab. Fortunately, this rarely presents a barrier. Providers of the enterprise systems note that after-market sensors often uses existing SEMI standards, e.g. SECS/GEM, though at times a custom transformation is needed to move data from a new sensor to a data platform.
“We are quite agnostic to how the sensors are integrated,” said Joe Fillion, director of product management at Onto Innovation. “Whether they’re integrated into the machine, or alongside the machine, our product has the ability to understand the values for the sensor. And we can trace and track that to either put it into a production hold or stop. These always are based upon set analysis units, which are configured for that particular type of sensor.”
Value in more sensor data
Sensors for temperature, pressure, humidity, gas concentrations, vibration and current can be found throughout the process tools and associated sub-systems. Third-party sensors are often added after installation and initial process development. Those sensors also can be applied to sub-fab intake and outtake pipes to monitor parameters that illuminate issues affecting process variability and defectivity. In addition, equipment teams can use instrumented wafers as a source for equipment set-up, post maintenance qualification and for predictive maintenance algorithms.
“Aftermarket sensors are used both within equipment and in auxiliary systems such as pipes and manifolds feeding critical materials. In addition to monitoring internal conditions like temperature, pressure, particles or vibration within equipment, sensors in pipes are essential to ensure the purity, pressure, and flow of gases or chemicals,” said Nordson’s Vijay. “Monitoring these feeds helps prevent contamination and ensures process stability, which is critical for yield optimization in semiconductor manufacturing.”
For example, a residual gas analyzer (RGA) can measure and fingerprint elements found in a space, generating specific and actionable data that equipment or process engineers can utilize in multiple places.
“They’re used for physical vapor deposition (PVD) tools and get installed primarily on de-gas chambers to detect undesired photoresist remaining on wafers,” said Boyd Finlay, director of solutions at Tignis. “They also are used on process chambers, buffer stations, and in exhaust lines for auto-leak detection (oxygen, moisture), contamination monitoring (wrong gas), or APC process control applications by measuring target gas presence, and therefore good/bad gas ratios in time series processing. RGAs are used mostly for vacuum leak detection. While an ion gauge shows gross leaks, most leaks are very small (e.g. a bellows leak), and those are the ones that are adding killer defects before you see the gross leaks. RGAs are also used for ligand ratio monitoring on atomic layer deposition (ALD) exhaust lines. During ALD the wafer goes into the chamber, and then the recipe calls for pulse/purging of the precursors and carrier gases to deposit the monolayer down on the substrate. The RGA is right there at the exhaust port, and it’s monitoring the byproducts coming out of the chamber. If the signal stays the same, then the process is consistent. But if it changes, the process has changed and there’s probably something wrong with an exhaust valve, precursor valves, or some other high-speed valve that’s happening within the ALD tool. Having that one RGA sensor on your exhaust, you see all the other problems coming in, including gas concentration issues.”
Instrumented wafers are used for temperature profiling of chambers, as well as a number of other parameters. Ease of use comes from their ability to access a tool by going through the AMHS (automated material handling system). Data collected from the sensors over the lifetime of a tool, and within a fleet of tools, goes well beyond just post-maintenance qualification. That data can be combined with manufacturing data analytics for predictive maintenance. In addition, third-party sensors can be used to diagnose defectivity events associated with wafer handling.
“Today the reason instrumented wafers became necessary on 200mm and 300mm tools is because the chambers are difficult to reach and measure under vacuum,” said Tignis’ Finlay. “They are required to run remotely through the tools to assess temperature profiles and wafer handler vibration. On 200mm and 300mm tools, you need a wafer device that could go through the tool with onboard power to take the measurement.”
Fig. 1: A 300mm wafer sensor with cameras for inspection and measurement of components above a sensor shower head and photoresist dispensing nozzles. Source: Nordson Test & Inspection
“As legacy semiconductor companies embrace smart manufacturing, we are seeing fabs adding 3rd party sensors to their manufacturing environments to gain additional visibility and control of their process steps,” said David Park, vice president of marketing at Tignis. “Whether the end goal is to address process variability concerns or to predict the mechatronic work rate degradation of equipment, raw sensor traces and their contexts are essential elements for these fabs to automate signal discovery using AI/ML.”
Nordson’s Vijay noted that instrumented wafers are battery operated and the data they collect is wirelessly transmitted. “These sensors enable real-time monitoring and provide data related to leveling, vibration, particles, humidity, and gapping,” she said. “All of these are critical factors for process equipment optimization.”
For instance, leveling is crucial for uniform deposition in CVD and PVD chambers. Existing wafer sensors offer horizontal accuracy of ±0.03 degrees.
The advent of IoT devices facilitates easy integration of sensors on a factory floor, or internal to a machine.
“I can give you three sensor examples that are outside the equipment and of value,” said Ranjan Chatterjee, vice president of smart factory solutions at PDF Solutions. “First, there’s ambient temperature of the facility or inside the particular machine. On older equipment, maybe those sensors didn’t exist. Now you can add them. The second is vibration sensors for the motors. Once motors are close to failing, they vibrate more. The third one is humidity, both in the facility and within a chamber. Humidity affects a number of processes.”
Third-party sensor applications
Third-party sensors are used for data collection for a myriad of reasons, including extending the lifetime of older equipment, controlling process variability, improving maintenance, and reducing defectivity. In all cases, the sensors need to be connected to data analytics products.
The good news is that integrating 3rd-party sensor data into factory data management systems is straightforward, especially when sensors use established SEMI standards like SECS/GEM. Even if they don’t adhere to those standards, the barrier to connect them to a factory’s data infrastructure is low. But to maximize their usefulness, several industry experts emphasized the importance of traceability and the ability to correlate data from different sources in context.
Fig. 2: Factory data infrastructure for connecting data from equipment, sub-fab and 3rd-party sensors to the various data systems, e.g. MES, APC, FDC. Source: PEER Group
Usage models for third-party sensors vary, but there’s always a well-defined need and rationale.
“All tiers will use third-party sensors, because as each organization learns, they say, ‘I really need to know more about X,'” said Onto’s Fillion. “That might not be something that comes with the tool. So whether it’s a tier-one or tier-three customer, there’s always going to be a need to do some sort of analysis after integration. The sensor itself would be transcendent upon any time, any tier customer. How the data is used and managed is going to be dependent upon the customer and their needs.”
Sometimes the equipment manufacturer is not involved at all. “They could use their standard data collection system to collect data from an additional five sensors that they just wired into the panel,” said Alan Weber, vice president of new product innovations at Cimetrix Connectivity Group at PDF Solutions. “Fabs typically want to do something the tool supplier is either not aware of or wouldn’t support. That’s why they’re putting the sensor on there. They’re trying to figure out what’s going on in the process so they can push it beyond its normal envelope.”
In other cases, equipment suppliers may identify a need for more sensors, often in conjunction with their fab customers. “Equipment suppliers often identify new monitoring requirements after the equipment is deployed and operational,” said Nordson’s Vijay. “They may add sensors to address issues such as unexpected drift, missing measurements, or to collect additional data that aids in optimizing performance or troubleshooting. Collaboration between equipment suppliers and fabs also drives these upgrades.”
Others concur that collaboration between OEMs and fabs can motivate additional sensing capability. “They usually do that in the context of a joint development program,” Weber said. “Specifically, they’ve agreed to do process development and take some ideas that either the tool supplier or their customer has experimented with, in order to see if the extra sensing and different control or different operating envelopes make sense in a given context. If they do, then they go back and probably add those sensors to their production tool.”
Another common application is adding capability to tools in a legacy fab. Even for equipment that is more than 20 years old, fab engineers systematically apply improvements throughout their useful life.
“The first problem is that these tools are either old, i.e., end of life, and the OEMs are out of business or they got acquired by somebody bigger that doesn’t want to add sensors,” said Bob Kane, director of North American sales at PEER Group. “Second, the software is locked down. And the last one, which is probably more prevalent, relates to new purchases. If I start putting sensors on this old tool, you’re not going to buy my new tool. There exist many reasons for adding sensors on these legacy tools. Ultimately, they need the data.”
Others agree that continuous improvement drives new sensors in old equipment. “Identification of new sensor needs could be process-driven,” said Nordson’s Vijay. “Consider all the capital tools of older generations that are already in the field. How will they go back and add a sensor to an old tool that is installed and out of warranty? So that is when aftermarket sensors are needed, especially those made to be wafer-like. These travel wherever a wafer can travel inside the tool. Sensors have to be completely viable and designed in a way it is really easy to be used in the tool.”
Conclusion
Adding third-party sensors to existing sensors is an established practice across wafer fabs. The data obtained from those sensors helps to improve yield, identify excursions, and maintain expensive equipment with reduced downtime.
“Our customers know that every 1% yield gain equals X dollars,” said Onto’s Fillion. “So they’re hyper focused on scalability, efficiency and yield. All of this sensor data is critical to the process, and thereby understanding it. One percent could really make a big difference to that overall bottom line for a company.”
Related Reading
Using Predictive Maintenance To Boost IC Manufacturing Efficiency
Smarter tools enable smarter fabs but the logistics of predictive maintenance challenge implementation.
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
Leave a Reply