IRAJ EMAMI AND BHANWAR SINGH (AMD Fellows, Advanced Micro Devices): As feature sizes shrink below 90 nm, process tolerances have become significantly tighter. Many  tools have not been able to keep up with the process control requirements needed for 65-nm devices and beyond. The only way to tighten the tolerances of these tools and processes is by integrating real-time metrology, with the ultimate goal of achieving real-time process control—that is, using the feedback from metrology sensors to control the tool during processing. Run-to-run control measurement data, such as the fault-detection mode, can be used to prevent processing or shut the tool down. In addition, real-time measurement and feedback to the factory floor will help ensure that learning rates are improved and wafer-sort yield variability is reduced.

Integrated metrology (IM) provides low-cost, in-line measurement capabilities. It will play an important role for advanced processes, equipment control, and tool-to-tool matching. Scatterometry will also be a critical technique. Key IM applications in the 65-nm node and below will be for patterned layers (CD, overlay, layer thickness, macrodefects), gate oxides (thickness, uniformity, nitrogen content), metal layers (filling quality, grain size, boundaries), and low-k materials (thickness, refractive index, porosity, composition, uniformity).

Although the integration of metrology tools is progressing, standardized communication interfaces for data transfer and networking must also be put in place. Algorithms are needed for measurements, implementation, testing of integrated measurement systems, and data management for seamless factory automation.

While there have been major strides in accelerating this technology, it remains a key challenge. The integration of these tools and the ability to provide detection sensitivities between tool- and factory-level control systems require a great deal of coordination and upfront, thorough planning and investment.

MICHAEL PASSOW (senior engineering manager, technology development and deployment, IBM Systems and Technology Group): To continue with the current trend in device scaling, manufacturing processes require more-exacting controls. The proliferation  of such e-manufacturing control techniques relies on a critical concept– effective system integration. Whether used for real-time feedback control in the case of integrated metrology or remote access to troubleshoot and repair a bottleneck process tool with e-diagnostics, the integration challenge to successfully enable these and other mission-critical control techniques is the ability to convert accurate data into useful information in a timely manner.

The design of the modern 300-mm fab operating in a "touchless" (fully automated) manner is the most difficult of environments in which to establish the appropriate infrastructure for hosting these controls. This environment is designed to optimize development and manufacturing activities with precision and speed. A typical fab has more than 200,000 wafer step moves in a day, gigabytes of data generated and analyzed, and tens of thousands of statistical process control (SPC) charts updated, all automatically. Advanced process control (APC) run-to-run calculations are performed in a just-in-time manner with the most recently available metrology data. IM devices supplement stand-alone measurement data for real-time feedback control between wafers. To make this work without humans involved, data quality and timeliness become paramount.

To optimize a multitude of control schemes, the required data must be available in a timely manner for use at the optimum location. —Michael Passow

At the factory level, the manufacturing execution system (MES) must manage the execution of the thousands of operations in the factory simultaneously. The factory-level systems for APC, fault detection and classification (FDC), sensors, and SPC must be responsible for and direct the overall control plans. At the tool level, embedded metrology, APC, FDC, sensors, and SPC for local control are proliferating. Delegation of control activities from the MES level to the embedded level is required to maintain consistent performance of the factory in the automated, asynchronous environment. Thus, cascading of control is needed between factory-level systems and the localized control applications. To optimize a multitude of control schemes effectively, the required data must be available in a timely manner for use at the optimum location.

To support the suite of embedded control solutions, the communications infrastructure must be designed to support timely request-and-respond functions. XML communication of commands, logistics information, and data through a middleware layer is the most efficient and flexible implementation. As the capabilities of the embedded controls increase, so does the need for more efficient communication with the MES and other control applications. Pushing Web-based techniques to the embedded layer strengthens and streamlines the distributed nature of the open-architecture environment. Enhanced data flow and increasingly complex data analysis techniques at the embedded level drive the need for enhanced processing power at the embedded sensor, PLC, or application level.

To enable this functionality, which is desired in the advanced control environment of the fully automated fab, strict adherence to standards is required. Recipe management must address the need to integrate both IM and APC recipes. Data collection and high-speed sensor communication requirements must be addressed, along with the coordination of APC control plans and IM sampling plans. Data must be in a standard format and accurately time stamped. Data flow should be "fast enough" to meet application needs with minimal latency. Control parameters should be flexible and selectable, and process tools must universally accept real-time control parameter updates.

JON MADSEN (director of engineering, Nanometrics): As APC gains widespread adoption, semiconductor manufacturers and process toolmakers are coming under increasing pressure to provide fully integrated process control solutions. The push for very tight process tolerances and productivity improvements from the billion-plus-dollar factories drives the need for APC. The ultimate goal of providing metrology directly on the process tool—i.e., integrated metrology—is to enable closed-loop control (CLC) of the process tool without operator intervention.

Reliability requirements for an IM module are considerably higher than for a stand-alone metrology tool. —John Madsen

Several hot button issues are inherent with the introduction of IM as a solution for tighter process tool control. Two important considerations are system reliability and rapid/automatic tool recovery. Since metrology is an integral part of the CLC system, the reliability requirements for the IM module are considerably higher than for a stand-alone metrology tool. Depending on how critical the CLC is to the process, an IM module failure can result in the production line halting for several hours until service is performed to correct the problem. A second issue is that, without an operator monitoring the measurements in real time, if metrology system performance degrades and is not detected early in the cycle, erroneous information may be sent to the process tool. This can cause incorrect changes in the process parameters, which in turn can result in catastrophic misprocessing of product wafers.

One solution to this problem is the routine performance of such system checks as precision, stability, and accuracy on a periodic basis, which can range from hours to days. In the case of thin-film and optical critical dimension (OCD) metrology tools based on optical spectroscopy (reflectometry or ellipsometry), wafers with standardized films or dimensional features made of silicon dioxide and other stable materials are commonly used. These precharacterized standards use highly accurate measurement systems and are compared to standard reference materials.

The performance of in-depth reference checks and calibration procedures on the process tool is particularly disadvantageous in the case of IM, since it requires the introduction of a standard or monitor wafer into the tool. Obtaining monitor wafers from the fab stocking system and loading them onto the IM module wastes both time and process tool resources.

A better solution would be to implement a methodology that optimizes the IM strategy for closed-loop control. There needs to be an approach that minimizes interference with the process cycle, yet enables automatic qualification of metrology system performance without the introduction of a standard or monitor wafer into the process flow. This can be accomplished using a reference standard located within the IM tool that can be easily and automatically accessed by the tool without operator intervention.

Once this approach is implemented on each process tool, it opens a new way of viewing the entire integrated metrology process. By using networking software, correlations with other process tool calibration and reference data can be performed. Questions such as why some process tools perform better than others based on IM measurements (i.e., CD and film thickness) can be easily answered. Autodiagnostics and autocalibration will enable semiconductor and process tool manufacturers to manage their fabrication processes without the need for operator intervention.

TIM ASH (vice president of engineering, core products group, Advanced Energy): Significant progress has been made in the industry toward improving overall equipment effectiveness (OEE) through International Sematech Manufacturing Initiative (ISMI) projects such as e-manufacturing and standards programs. A significant gap, however, exists in the OEE strategy—namely, the connectivity of critical subsystems such as RF generators, mass-flow controllers, and other sensor-actuator devices into the overall fab information architecture.

ISMI is calling for standard interfaces that allow seamless information flow, resulting in higher OEE and reduced manufacturing costs. The focus has been placed on SECS/GEM; interfaces A, B, and C; and infrastructure standards such as the common equipment model (CEM). Unfortunately, efforts have stopped just short of where they need to go. One look underneath the hood of state-of-the-art 300-mm etch, CVD, and PVD plasma processing tools demonstrates the lack of a standard information and control interface at the critical subsystem layer. The hardware interfaces in use all have proprietary software protocols.

SEMI has adopted a family of standards for this class of equipment—the E54 Sensor/Actuator Network Standards. E54, however, is not driven as a business requirement, and as a result, it is not followed. This standard represents many years of work and industry experience, yet little if any of this work is under consideration in the OEE strategy. The E54 subcommittee issued a consensus statement in the February 2005 North American Standards meeting that documents this concern: "Tools today are collecting and manipulating data internally utilizing E54 SEMI standard sensor bus technology. The equipment data acquisition (EDA) standards within the SEMI CEM standard define mechanisms for extracting data from equipment at high speeds for applications such as diagnostics and fault detection. It is important that these two standards are aligned so that sensor bus data collected internally could be easily exported via the EDA interface. The best place to start in this alignment effort might be extensions or changes to the CEM object model definition."

E54 is not perfect, but it is a good starting point. In order to achieve the productivity improvement called for in the OEE strategy, the industry must first regard E54 and the information from the critical subsystems as essential to e-manufacturing standards. Second, the industry must drive to establish Ethernet as the single standard for network communications. No industry can meet overall cost and quality objectives if the supply chain is required to develop and support more than eight different complicated network technologies with proprietary protocols.