A blueprint for enterprise-wide deployment of advanced process control


James Moyne, Applied Materials, Santa Clara, CA USA

Advanced process control (APC), which includes both run-to-run (R2R) control and fault detection and classification (FDC), is now a required component in all fabs, providing process improvement and waste reduction. A major global semiconductor manufacturer is currently involved in an enterprise-level deployment of APC. Execution to specific technical and business models across the enterprise has resulted in a high ROI APC deployment with results including 50% improvement in process capability for key processes, and detection of key faults impacting productivity, such as etch arcing. The deployment process developed is proposed as a cost-effective approach for others to use in enterprise-wide APC deployment.

A major semiconductor manufacturer that has three fabs and four final assembly facilities distributed worldwide determined that APC technology is now a required component for all of their fabs. This manufacturer is currently deploying an enterprise-level APC solution that includes both R2R control and FDC [1]. Developing and deploying this APC system included meeting stringent requirements so that high and immediate ROI could be achieved through improved performance and reduced cost.

On the business side, the deployment had to be truly global which meant issues such as language and roll-out best practices had to be addressed. The personnel costs associated with APC roll-out, and especially maintenance, had to be small. Additionally, a phased and flexible approach to deployment was needed to support culture migration to APC and immediate ROI, while allowing migration to more complex integrated APC solutions as the technology and need evolve. On the technology side, integration with non-APC capabilities such as the existing MES was critical, a common data repository for R2R and FDC data was needed, and common and flexible interface for visualization of R2R and FDC data in a dashboard fashion was required that would allow for easy audit of runtime execution as well as enable worldwide remote access and support.

With these requirements in hand, the semiconductor manufacturer chose to develop both a technical and a business model for APC deployment to address these requirements. These two models can be used as a blueprint guiding other enterprise-wide industry APC deployment efforts.

Technical model deployment

Cost-effective enterprise-wide deployment of APC places very specific and in some cases unique requirements on the technical solution. First and foremost it has to be a complete APC solution, integrating both R2R control and FDC capabilities today, but also providing a mechanism to support future integration of other APC components, such as fault prediction, in a common manner. A consolidated solution of this kind is important to lowering costs as it supports a consolidated data storage, reduces integration (to both equipment and MES) costs, greatly reduces training costs, and improves ease of use. In providing this type of integrated multi-component APC solution, it is important that the SEMI Process Control System standard (E133) be leveraged for specifying integration between APC and non-APC components [2].

In addressing flexibility and reconfigurability requirements of the solution, it was determined that the solution has to utilize event-condition-action (ECA) technology to govern the interaction of APC capabilities with the outside world. ECA technology is a type of event-based configurable control technology whereby the control is defined as “on event, if conditions are met, perform actions” [1, 3]. As an example, Fig. 1b shows the graphical front-end of an ECA strategy depicting a control rule for the event “R2R update recipe settings.” Note that this rule defines the integration of the APC system with the outside world, but also defines the collaboration of APC components within the APC system. ECA technology is important to the implementation of a low-cost enterprise-wide APC system because it breaks down the complex logic of large implementations into small manageable control rule strategies, and is a user friendly logic management approach that is more aligned with the natural way of thinking about and documenting (as best practices) solutions.

Another important solution requirement is support for a consolidated dashboard-style user interface that could be customized to provide the appropriate information (from multiple components) to the appropriate level of detail to the user class. The use of consolidated common dashboards is important to ease-of-use in operation, but also reduces the training effort, and facilitates execution of the solution lifecycle feedback model.

Figure 1. Options that meet the requirements defined for this enterprise-wide deployment include common integration approach (a); ECA front-end drag-n-drop strategy editor (b); and dashboard-style configurable user interface (not shown).
Click here to enlarge image

An example of the solution deployed that meets these requirements is shown in Fig. 1. Figure 1a illustrates a common integration approach that leverages the PCS E133 standard and the support of both R2R control and FDC capabilities. Figure 1b provides an example of a graphical strategy builder, where icons are selected from a menu and a drag-n-drop approach is used to create strategies, and 1c shows the dashboard style user interface capability of the solution where both R2R control and FDC information can be displayed simultaneously to the logged in user.

One capability of this solution that was not specified during evaluation, but turned out to be of prime importance during deployment, is the strategy debugger. This capability tracks the execution of strategies step-by-step (e.g., icon-by-icon in the strategy of Fig. 1b). This helps users understand the strategies as deployed and investigate problems. It also serves as an excellent communication tool between the field deployment team and a centralized APC model and strategy development team.

Business model deployment

The APC deployment business model was focused on low upfront and continuing costs, flexibility, and support of an international deployment. The first key component of this business model is the concept of a small control team. Specifically, a small centralized team of APC experts is responsible for building and transferring globally APC solutions such as R2R and FDC models and strategies. This interaction is standardized through the implementation of well-defined life cycle models. In general, local site teams support minor customizing to address local control issues; these local teams also support testing and verification of models delivered by the small central control team and provide feedback that serves as in-bound marketing to support the development and refinement of models. Additionally, the central team provides standardized APC training, which is customized by the local teams to address language and culture differences.

Click here to enlarge image

The second key business model component is utilization of a phased approach to deployment (Table 1) that generates immediate and continuous ROI. An important realization in the APC deployment was that significant ROI could be achieved just by replacing existing highly manual control solutions and business rules with automated counterparts. Thus, this first phase of deployment reduced the culture shock of automated APC by delivering an automated solution that was familiar to the fab. Migration to improved APC solutions could then proceed with minimal resistance. The process can be summarized as: start by automating existing practices, keep it simple, provide a path for evolution that allows for flexibility, and support a collaborative environment (especially between central control team and individual fabs) for continuous improvement.

A third key component of the business model is the development of and adherence to best practices to reduce both upfront and recurring costs. In addition to the APC life cycle models and phased deployment approaches described above, specifications are pre-defined for standardized (E133) APC integration, training is standardized, and common ECA strategies are pre-defined for R2R control and FDC deployment. As an example of the latter, common strategies utilized in R2R control solutions are defined for events of pre-metrology data received, used settings data received, recipe request, and post-metrology data received. These strategies are of the form illustrated in Fig. 1b (for example the particular strategy in Fig. 1b could be part of the strategy associated with the event post-metrology data received).

A sample of results

To date, FDC has been deployed in over 200 processes in areas at this semiconductor manufacturer including CVD/PVD, etch, implant and lithography, while R2R control solutions have been developed and deployed for CMP, etch, diffusion, lithograph overlay, and lithography CD. In the remainder of this section, a few key results are highlighted in the areas of R2R control and FDC that have been adopted as a best practice and rolled out to the enterprise utilizing the technical and business models identified above.

R2R result. Improving process Cpk is oftentimes the primary motivation for implementing R2R control. This improvement is usually attributed to the use of closed-loop model-based control. However, improvement gains due simply to the automation of the process are often overlooked.

In this example, a phase 1 controller was written for the ILDO process, the first inter-layer dielectric, which simply emulated and automated the current manual control process for calculating polish head removal rates. An increase in Cpk from 1.12 to 1.72 was achieved corresponding to a process improvement of 50%. Note that this example also illustrates that the removal of the human element that comes with automation of a process can be a major factor in process improvement. Even well documented processes run by highly qualified operators, which is the case in this example, can benefit from automating the process, as an automated process enforces exact rules and logic.

FD results. Etch arcing can of course be very destructive to product, and thus detection of the etch arcing problem is a source of high ROI in an FDC system through scrap avoidance. Direct detection however is very difficult because extremely high sampling rates are needed to catch the DC bias arcing spikes. An advantage of a comprehensive FDC system, such as the one utilized here, is that it allows for exploration of more cost-effective alternatives.

Figure 2. Fault detection solution for etch arcing. (Arcing creates a photoresist byproduct that hinders main-etching of next wafers; this byproduct can be monitored by optical emission spectroscopy (OES); a windowing feature of the FDC system allows for zeroing in on the plasma etching step to detect anomalies indicating definitively that arcing has occurred.)
Click here to enlarge image

In this case, investigations revealed that the arcing phenomena could be captured indirectly by monitoring the chamber plasma (Fig. 2). Specifically, arcing creates a photoresist byproduct that can be monitored by optical emissions spectroscopy (OES). The OES intensity profile at a particular wavelength is different if arcing has occurred indicating that the photo-resist byproduct has been created. A windowing tool of the FDC system allows for high resolution “zeroing in” on this feature so that an arcing fault can be clearly determined.


The process of clearly defining the technical and business models presented herein for enterprise-wide APC deployment ahead of actual implementation proved valuable and resulted in improved ROI, as well as a better understanding of how ROI would be achieved. Looking ahead, it is clear that APC deployment is an ongoing effort and these models need to take this into consideration. In the deployment presented in this paper, we have achieved ROI in both R2R control and FDC related to the early deployment phases. As we move forward, the APC technical and business models will allow us to complete the remaining phases outlined in Table 1, and migrate into new areas such as fault prediction and APC incorporation with yield management.


The author wishes to acknowledge Mark Yelverton and Kevin Chamness for their contributions to this article.


  1. J. Moyne, “Making the Move to Fab-wide APC,” Solid State Technology, Vol. 47, No. 9, (September 2004); pp. 47-52.
  2. J. Moyne, H. Hajj, K. Beatty, R. Lewandowski, “SEMI E133: The Process Control System Standard: Deriving a Software Interoperability Standard for Advanced Process Control in Semiconductor Manufacturing,” IEEE Trans. on Semiconductor Manufacturing-Special Issue on Advanced Process Control, November 2007, pp.408-420.
  3. J. Bae, H. Bae, S.-H. Kang, Y. Kim. “Automatic Control of Workflow Processes Using ECA Rules,” IEEE Transactions on Knowledge and Data Engineering, 16(6), 2004.

James Moyne received his PhD degree from the U. of Michigan (where he is currently an associate research scientist), and is a standards and technology specialist for the Applied Global Services Group at Applied Materials, 3050 Bowers Ave., Santa Clara, CA 95054 USA; ph: 734-516-5572;