Brad Van Eck, International Sematech Manufacturing Initiative (ISMI)

All IC manufacturers must continuously control the cost of manufacturing. This activity must be done at a rate that offsets the rising cost of equipment, materials, chip and process complexity, and additional environmental, safety, and health (ESH) requirements if the industry is to maintain its historical ability to reduce costs per function. At the International Sematech Manufacturing Initiative (ISMI) Global Economic Symposium held in November 2004, several speakers noted that the semiconductor industry sells more than half of its products directly to the consumer market—a market even more volatile than the traditional markets to which chips were delivered in the past. This additional market volatility requires not only more-efficient manufacturing but also more-agile manufacturing with shorter cycle times.

Over the next decade, major technological changes are also anticipated as both traditional lithography and device structures reach their limits. Disruptive technologies such as novel materials and new, more-complex lithography solutions will increase R&D and equipment costs and complicate manufacturing. Process windows will inevitably narrow, tool complexity will escalate, and more-timely responses to rapidly changing markets will have to be incorporated into the way that factories are managed in order to meet customer demands.

These changes engender ever-more-challenging equipment and factory control issues in an environment where productivity has to be continuously improved to comply with the rule of Moore's Law. An acceleration of the technology nodes from three- to two-year cycles, something many leading-edge companies already claim to do, will stress these areas of concern even further. A transition in wafer size to 450 mm is still being hotly debated. This change, if required to meet productivity goals, will place greater demands on both chipmakers and equipment suppliers and stretch R&D dollars even further. The pressure to improve both equipment and factory productivity in the face of these new challenges is clearly anything but business as usual.

The ISMI Manufacturing Effectiveness series will address several of the key issues facing the industry in the coming decade. In subsequent articles over the course of the year, the authors will describe how advanced strategies and tools can be used to meet these challenges in the core areas of fab productivity, equipment productivity, and metrology. This issue's article presents an overview of those challenges.

Realizing the Roadmap to Productivity

The 2004 update of the International Technology Roadmap for Semiconductors (ITRS) calls for significant cost reductions in manufacturing and simultaneous improvements in productivity. Consequently, the "Realize the Roadmap" vision, championed by Sematech and others in the semiconductor industry, poses a variety of challenges. The recent demand for more data from all process and metrology tools in the factory indicates that a portion of the solution is a smarter factory, one that is continuously monitoring and optimizing itself. This optimization must occur in real time and must be largely automated if its full benefits are to be realized. The factory has to be truly data-driven and automated to improve productivity and reduce manufacturing costs to reach the goals required by the ITRS. Figure 1 shows some projected attributes of the roadmap's vision of the 300-mm e-fab for the 45-nm-and-below technology nodes.

Figure 1: Vision of a future factory, showing projected attributes of a 300-mm fab with 45-nm-and-below process capability. (Source: ITRS)

The 2004 edition of the ITRS calls for the introduction of larger wafers, most likely 450 mm, in 2012. If the market demand for semiconductor products continues to grow and the roadmap has accurately predicted technology-node transitions, larger wafers will be needed to meet the required productivity improvements. Based on the lessons learned from the transition to 300-mm wafers, the industry already lags behind in its work to prepare for this transition. In conjunction with other organizations and companies, ISMI has committed to provide leadership and investigate transition requirements as the industry considers this critical decision. A full treatment of issues pertaining to the 450-mm (or larger) wafer conversion will be one of the topics included in the article on fab productivity, scheduled for publication in the April MICRO.

Data-Driven, Automated Decision Making

Factories are preparing for an increased volume of data, but if factory information and control systems are not poised to convert this information into actionable decisions to redirect equipment and factory operations, the systems will produce no tangible benefits. Recent moves toward computer and Web-based standards are very encouraging. This trend must accelerate if the semiconductor industry is to take advantage of the equipment and factory control software already in use in other industries.

Figure 2: FAST II roadmap for equipment data acquisition for evaluation, showing the timing needed by chipmakers to evaluate software for the new data port. (Source: ISMI and SEMI)

There has been significant progress toward easy access to data. A second communications port dedicated to data acquisition, the equipment data acquisition (EDA) or Interface A port, has been standardized using the SEMI standards process. Figure 2 depicts the FAST II roadmap, developed jointly by ISMI and SEMI, for the timing needed by devicemakers to evaluate software for the new data port. Figure 3 shows the dates that devicemakers and suppliers must meet to produce robust software that supports the new port. As the figures reveal, most chipmakers are targeting 1Q 2005 for software evaluation and the latter half of 2005 for commercial availability of equipment using the standards that enable Interface A.

Figure 3: FAST II roadmap for equipment data acquisition for production, showing the timing needed by chipmakers and tool suppliers to produce software to support the new data port. (Source: ISMI and SEMI)

To guide this transition, both discrete factory simulation and economic modeling are needed. The industry must identify those areas where the most salient cost reduction and productivity gains will likely be found. Since resources are limited, the focus must be on those areas with the highest and fastest return on investment.

As processes become more complex and process windows narrow, lot-to-lot control must progress rapidly to wafer-to-wafer control and ultimately to real-time within-wafer control. Fault detection and classification (FDC) must be widely proliferated across the fab to rapidly reduce both scrap and rework. Since these systems are only as good as the data they depend on, proper access to information on the health of the process, the tool, and the wafer is required to reach the promise of this technology. More and more software suppliers have entered the market to fill this need. The integration of these new products into the fabs can be accelerated at a reduced cost if the data are of sufficient quality and can be delivered in a standardized format.

Although total equipment utilization rates of between 50 and 75% in the fab were once the norm, this area clearly represents a significant opportunity to improve productivity. There must be an evaluation of all manufacturing systems to identify and reduce or eliminate every root cause of nonproductive time. Factories that can detect drifting processes before they fall outside acceptable process windows will be able to take corrective action before the generation of scrap. The availability of these data will also enable more-sophisticated FDC systems that will anticipate process and equipment faults.

With this kind of early warning system, factories will be able to schedule the labor to make the repairs, ensure that the required parts are on hand, and provide the process and equipment engineers with the diagnosis of the problem before removing the equipment from production. These data will also enable faster requalification of equipment, speeding its eventual return to manufacturing. Early warning of process and equipment problems must also be employed by real-time factory scheduling software to minimize the impact on overall factory efficiency. When proliferated to the entire process and metrology tool suite, these techniques will reduce the total nonproductive time for the factory's most expensive assets.

For equipment with in situ metrology, this increased availability of data will allow for run-to-run (R2R) control. This level of control will enable tools with narrow process windows to continuously adjust the process to stay within those windows without the need for adjustment by equipment and process engineers. It will also reduce downtime and labor costs and reduce or eliminate the production of scrap. This is one of the challenges in advanced chip manufacturing that will be addressed in the installment dealing with metrology, slated for publication in the August/September isssue of MICRO.

For difficult or subtle process and equipment problems, proper access to these data likewise will enable easy, rapid access to experts both inside the chipmaking factory and at the equipment supplier's site for the diagnosis, repair, and requalification of equipment.

To optimize the entire factory, truly intelligent factory schedulers must have access to equipment and wafer status data as well as information on spare parts, labor, materials and chemicals, customer orders, finished inventories, in-line defects, and final test. To achieve this level of optimization, improved software is required. The increased availability of these data will enable predictive and preventive maintenance. A continuous real-time supply of information about the status of the factory, its equipment, wafer status, new orders, inventories of finished product, and spare parts must be used to continuously optimize factory operations in real time. Standardized interfaces for all systems and equipment that generate data are being developed to merge these diverse data sets.

Integration has been identified as the most significant problem facing the industry as it attempts to realize data-driven decision-making in high-volume manufacturing. While there has been progress in these areas, it is imperative that integration be completed quickly so that factories can focus on optimizing the FDC and R2R control algorithms, not data integration.

The role of software, already a critical component of every factory, is becoming even more crucial to achieving productivity improvements. Software reliability affects the time required to install and qualify new equipment. The additional software needed to acquire, analyze, and deliver actionable decisions must be robust and production-worthy. Unambiguous requirements, detailed usage scenarios and exception-handling guidance for high-volume manufacturing, and software testing strategies are all necessary to produce software of sufficient quality to meet these challenges. Both equipment software and third-party software must be tested before installation.

Figure 4 shows the average aggregate percentage improvement (issues corrected) from more than 20 process and metrology equipment suppliers' software after a series of test, improvement, and retest cycles. The data clearly show that software quality can be dramatically improved using the test/improvement cycle/retest methodology. This type of approach must be used for all new equipment software as well as for revisions to existing software before being installed on the factory floor.

Figure 4: Average aggregate percentage improvement by equipment software from more than 20 process and metrology tool suppliers after test, improvement, and retest cycles. (Source: ISMI data derived from testing.)

The ability to share data among the various segregated databases for processing, equipment status, spare parts inventories, in-line defects, final test (both parametric and yield), inventories, order status, hot-lot status, customer returns, failure analysis, and design is essential to make the best decisions for factory optimization. There need to be systems that simultaneously collect, analyze, and automate simple decisions as well as conveniently present data to process engineers, equipment engineers, and middle and upper management. While some factories already have systems deployed that can perform a portion of these actions, the cost of integration, as well as the time and manpower to operate the systems required, has become too high. The pivotal role of software in transforming the capabilities of factory tools will be one of the topics explored thoroughly in the article on equipment productivity, set for publication in the June MICRO.

The ITRS calls for a host of new capabilities to support e-manufacturing. Figure 5, taken from the roadmap document, shows the status of process and equipment control, automation, and other e-manufacturing capabilities in a current leading-edge 300-mm fab. Figure 6, also excerpted from the technology roadmap, lists the automation and equipment capabilities that must be included. This graphic also reveals that the data volume rate with the new EDA port will increase exponentially: the new port's data volume will rise to approximately 10,000 data points per second compared with the current equipment rate of 300 data points per second.

Figure 5: APC, AEC, automation, and other e-manufacturing attributes of current leading-edge 300-mm fab running high product mix. (Source: ITRS)

Rising to Meet the Challenges

The semiconductor industry faces the largest assortment of technical and productivity challenges in its history. This daunting array features a large number of disruptive technologies, including the introduction of new high- and low-k dielectric materials and strained silicon, as well as new, more-complex lithography solutions such as 193-nm immersion, extreme ultraviolet, maskless, and nanoimprint. With the industry selling more than half of its products directly to consumers, markets have become more complex and volatile, requiring shortened cycle times and more-agile manufacturing.

Figure 6: Automation, equipment, and related e-manufacturing capabilities needed for future fabs. (Source: ITRS)

The possibility of accelerated technology node introductions, which will be required should the industry return to a two-year cycle, would further test IC manufacturers. The ITRS-forecasted upgrade to 450-mm wafers by 2012 will drive the industry to make a belated push toward this difficult productivity goal. Software, a key segment in every factory, must become more robust and production-worthy in order to acquire, analyze, and deliver actionable decisions; improved testing is especially critical.

These and other manufacturing effectiveness challenges, which will add to both R&D costs as well as to equipment and operational budgets, must be addressed by the semiconductor industry in the coming decade. This article series will detail some of those challenges and provide possible solutions that will allow the chipmaking community to stay on the roadmap over the next decade.

---

Brad Van Eck, PhD, manages ISMI's factory productivity program. Over the course of his career, he has worked in various R&D positions at RCA, GE, and Harris Semiconductor, where he managed the unit step process development group, focusing on advanced interconnect process development. Harris assigned him to Sematech in 1990. As an assignee, Van Eck developed vapor-phase process tools and cleaning processes. He moved to the consortium as a direct hire in 1993. During his tenure at Sematech, he has managed both the sensor development and integration project and the RTP development project, focusing on temperature measurement. He chairs the annual AEC/APC Symposium, contributes to the ITRS in the areas of factory integration and metrology, is a founding member and current vice president of the Integrated Measurement Association, a member of AVS, and is cochair of the sensor bus subcommittee of SEMI standards. He has a BS in chemistry from Calvin College (Grand Rapids, MI) and a PhD in inorganic chemistry from Michigan State University in East Lansing. (Van Eck can be reached at 512/356-3981 or brad.van.eck@sematech.org.)