skip to main content

The 2013 International Technology Roadmap for Semiconductors (ITRS): Major additions in prediction and big data technology areas

James Moyne, Ph.D.

The International Technology Roadmap for Semiconductors (ITRS) is probably the single most important document governing the direction of the semiconductor manufacturing industry. In addition to setting targets for future manufacturing capabilities, it identifies areas where technologies do not yet exist to meet a capability, and also highlights emerging technologies that could be harnessed to address a capability requirement. The ITRS is divided into chapters to represent the various aspects of semiconductor manufacturing. The Factory Integration (FI) chapter seeks to define a roadmap for factories and enterprise systems that are designed and integrated for efficient and effective development and manufacturing [1].

During the past five years the technology space that encompasses “Factory Integration” has seen a rapid evolution. Concepts such as system-wide (web) integration, the cloud, mobile devices and apps, predictive analytics and big data should now be an integral part of any industry technology roadmap. The FI technology working group (TWG) decided that a new roadmap was needed to capture galvanizing concepts such as commonality across all areas, fab-wide integration and data-driven systems. Consequently the entire FI chapter was restructured as shown in Figure 1. This new vision keeps the existing functional areas (now called “thrust areas”) of Factory Operations, Production Equipment, Automated Material Handling Systems (AMHS), Factory Information and Control Systems, and Facilities. It then extends the roadmap to incorporate new thrusts impacting all FI areas. The new overarching thrusts defined for 2013 are “Augmenting Reactive with Predictive,” “Big Data” and “Control Systems Architectures,” and new sub-chapters were created for each of these.

Figure 1. New Factory Integration Vision

Augmenting Reactive with Predictive (ARP)

In justifying ARP as one of the overarching FI thrusts the ITRS states that “industry needs to augment the existing reactive mode of operation, changing reactive operations to predictive operations wherever possible, but continuing to be able to support reactive operation. This will provide significant opportunities for cost reduction and quality and capacity improvement.” The ITRS prediction vision is a state of fab operations where:

    1. yield and throughput prediction is an integral part of factory operation optimization, and
    2. real-time simulation of all fab operations occurs as an extension of the existing system with dynamic updating of simulation models.

The scope of ARP is the set of technologies that collectively and collaboratively are required to achieve this vision. They include Predictive Maintenance (PdM), Fault Prediction (FP) Virtual Metrology (VM), predictive scheduling, yield prediction and augmenting predictive capabilities of the factory through simulation and emulation. Figure 2 illustrates current thinking regarding the roadmap of these technologies as fab-wide solutions. The prediction vision generally is the same for 300mm and 450mm facilities and full implementation of the vision is expected to become a requirement for remaining cost competitive in both facility types.

Figure 2. Augmenting Reactive with Predictive (ARP) Roadmap for Technology Requirements [1]

Big Data (BD)

The fab is continually becoming more data driven and requirements for data volumes, communication speeds, quality, merging, and usability need to be understood and quantified. Challenges and solutions associated with these issues are provided in the BD sub-chapter, categorized according to the “five ‘v’s”: volume, velocity, variety, veracity and value.

Volume—With the increase of data collected per tool and per wafer, storage of large amounts of data (petabytes) places considerable load and cost on existing infrastructure, such as analyzing, storing, processing and cleansing data. Algorithms to optimize the storage of data are needed. Data models that enable access of the data in an optimal and reliable way must be developed and standardized for applications to plug and play.

Velocity—Issues include data generation speed, speed of compression as needed for transmission, speed of transferring, speed for pre-processing for storage, speed of storage and speed of analyzing. The rate of data generation is exceeding the ability to store it in the underlying systems.

Variety—Merging different data sources and data types is often difficult, time-consuming and results in data quality degradation (Veracity). A factory must make huge volumes of data meaningful to the product flow and process steps such that multiple applications can take advantage of the data to create meaningful and actionable information.

Veracity—Veracity refers to the accuracy or truthfulness of the data. For example, data store reduction can be accomplished by new and emerging techniques used to compress data without impacting the quality of the data and ensuring no loss of information. Data quality of all data stores will be of increasing concern, with issues such as quality of data in maintenance databases and quality of manually entered data being roadblocks to cost-effective implementation of new technologies such as predictive maintenance.

Value—The cost of big data needs to be balanced with its potential value. Costs include collection, storage, and processing of the data. Benefits are identified in other roadmap areas such as prediction technologies. Items that are in the scope of BD value include applications that efficiently mine data, and specific data analysis systems such as expert systems, yield management, and maintenance management.

Control Systems Architectures (CSA)

Control Systems Architectures (CSA) encompasses control aspects that are common across FI technologies. It covers control at the higher levels such as process run-to-run control and manufacturing execution systems. It addresses the challenges associated with both the evolution and potential revolution of these control systems. Evolutionary items include more granular control, higher speed control, higher control quality, and higher levels of control capability. Potentially revolutionary items include the possibility of new control paradigms and new control platforms, such as cloud-based systems, distributed and autonomous control, machine learning, and artificial intelligence.

Looking Ahead

As we move through 2014, the focus will be on conducting the appropriate research to complete the challenges and potential solutions tables in the Augmenting Reactive with Predictive and Big Data sections (see, for example, Figure 2). Working with organizations such as SEMATECH, surveys of the industry are planned that will help define the starting point and near-term roadmap for these technologies.

[1] For more information on the ITRS FI Chapter, access the electronic chapter links for Factory Integration highlighted as links throughout this chapter and online at The 2013 materials will be available at this site in late March 2014.

James Moyne is Chairman of the ITRS Factory Integration International Technology Working Group and serves Applied Materials as a Factory Systems Consultant for the Applied Global Services business.