Computational Process Control Solutions to Service Fabs in China
By David Lammers
China’s initiatives to build more sophisticated greenfield semiconductor fabs come at a time when equipment and services suppliers such as Applied Materials are developing new ways to gather and analyze data. New computational process control (CPC) solutions, which combine subject matter expertise (SME) with advanced analytics, may be exactly what is needed to support Chinese companies as they start operations in locations where the fab infrastructure is largely built from scratch.
Kirk Hasserjian, corporate vice president of service product development at Applied Global Services (AGS), said the current situation in China combines two factors: it is a large country with fabs being built or planned in many different cities; and it’s becoming increasingly difficult for Chinese manufacturers to find and recruit enough experienced engineers to fill their growing requirements.
“That creates a very good opportunity for these fabs to take advantage of our technology-enabled services and remote access capabilities. We can bring value-added service products and help companies accelerate their yield ramps,” said Hasserjian, who will deliver a keynote address at this month’s SEMICON China show1.
Figure 1 - An explosion of data captured in fabs creates opportunities to improve production yields. (Source: VLSI Research and Applied Materials)
There is an explosion of data due to multiple technology inflection points in semiconductor manufacturing (see figure 1). To fully leverage such data, Hasserjian said Chinese companies would benefit from CPC solutions using an Applied field service server (FSS), which is installed in the customer’s fab and connected to the fleet of Applied tools (see figure 2). These servers enable Applied experts to remotely access the FSS-connected tools over secure connections managed by a third-party vendor, substantially improving time to problem resolution and reducing the need for on-site troubleshooting resources. “From a service technology standpoint, if we have the customer tools hooked up to an FSS and have a remote connection, we can get experts from around the world involved at a moment’s notice,” he said.
Figure 2 - A Field Service Server (FSS), supported by Applied process and equipment expertise, is a critical part of Computational Process Control (CPC) implementation. (Source: Applied Materials)
“Customers do have concerns about the security of their data over remote connections and we are addressing it by using highly-respected third-party remote access vendors,” Hasserjian said. He added that using a third party adds a layer of assurance that the customer’s data, as well as Applied’s proprietary information, are independently being protected by a trusted go-between. “This remote access solution has also been adopted by other equipment suppliers, and we all have had success with this approach,” Hasserjian said.
Mike Armacost, managing director of advanced services engineering at AGS, said semiconductor fabs generate large data sets, including time series (trace) data. The challenges are somewhat different for databases stored at a fab site, or in an off-site data center that might have more computing horsepower.
AGS is working with several database companies to determine how to store and analyze equipment data, and automate the process of detecting any changes in equipment behavior. “This is an area of focus today for a variety of manufacturing industries. There are a number of innovative algorithms and data structures which enable this, when coupled with the appropriate SME. In our case, we are particularly concerned about analysis of time series data with the contextual equipment knowledge to better predict and control tool performance,” Armacost said.
One topic of intense research is deep learning, in which neural networks sift through labeled and unlabeled data to detect patterns. Hasserjian said machine learning techniques, including deep learning, are being developed within Applied Materials to analyze various data types, including time series or trace data.
James Moyne, a consultant to Applied Materials and an associate research scientist in the Mechanical Engineering Department at the University of Michigan, said deep learning is one of many techniques that Applied can use to detect issues and boost yields. “Deep learning is just another algorithm, one that is very tailored to big data so that it can analyze very large data sets. But there will always be a need for SME to develop any solutions. CPC allows us to combine the algorithm power with SME in a structured and reusable manner.”
Figure 3 - Computational Process Control (CPC) uses advanced data analytics techniques and large data sets from tools, processes, and fab infrastructure to accelerate yield improvements.(Source: Applied Materials)
“Applied has in-depth knowledge of our equipment design and sensors used in the equipment. Our tools are developed for specific processes and applications so we have a general understanding of the various process steps, even though we do not have exact knowledge of each customer’s process technology,” Hasserjian said. “Data analytics is not just algorithms. It is analysis that uses the expertise we have from a tool design and process understanding standpoint, and how the sensors react in a chamber environment. This is what we refer to as computational process control (CPC)” (see figure 3)
As companies move to leading-edge processes, “the technologies are becoming so complex, and the process margins are getting so narrow, that our customers are looking for help to control the tools and address some of these technical issues,” Hasserjian said. “That is driving technology-enabled services. We want to demonstrate that we are not just a tool maker, but a solutions provider. CPC is a big part of that.”
Mingwei Li, director of product marketing at AGS, said semiconductor companies are starting to realize that they must evaluate the tradeoffs: the value that advanced services and remote analytics bring in terms of keeping equipment and yields at optimum levels, balanced against data security concerns. Li said the answer is essentially the same in China as in other parts of the world. “We have to demonstrate that the benefits significantly outweigh the customer concerns. Over the last year and a half, as customers see what can be achieved with our services supported by CPC solutions, they are far more willing to work with us,” Li said.
Li emphasized that Applied’s intimate knowledge of its equipment is what differentiates its ability to analyze data and deliver services. “The CPC concept is that with Applied’s proprietary analytics software, and our process and equipment knowledge, we can help our customers leverage big data.”
For additional information, contact Michael_D_Armacost@amat.com.
1Hasserjian’s keynote, “Big Data Analytics for Smart Manufacturing in the Semiconductor Industry” is planned for Thursday afternoon, March 15, during the Smart Manufacturing Forum at SEMICON China in Shanghai.