Integrated Metrology Offers Great Promise

But Why Such a Slow Adoption?

Bijan Moslehi, PhD, is chief technology officer and senior vice president, semiconductor technology research, for The Noblemen Group (www.noblemengroup.com), a boutique investment banking, strategic advisory, and business development firm. Moslehi has some 20 years' experience working in the semiconductor and semiconductor equipment industries. He can be reached at bmoslehi@noblemengroup.com.

Since the mid-1990s, various integrated metrology (IM) strategies have been actively pursued, with early projections once heralding widespread use of the technology by the new millennium. However, with the exception of CMP and limited uses in postpatterning macrodefect inspection, the rate of adoption has been extremely slow. This has been quite disappointing to tool suppliers that have spent tens of millions of dollars to develop products for what many consider a very small niche market. CMP applications dominate what little market there is, with one supplier accounting for as much as three-quarters of the space. Even though integrated metrology is forecast to grow at a rapid clip over the next few years, mostly in etch and lithography, it is still projected to remain a small portion of the much larger process control market.

The primary IM drivers include early detection of process excursions, improved process control, and fast yield learning and ramp of new products and processes. Increased sampling with IM enables tighter process windows, lower "at-risk" materials, and wafer-to-wafer advanced process control (APC). From just a technical perspective, the need for the integrated approach can be determined from detailed analysis of traditional SPC charts, which remain the best indicators of tool and process stability and excursions, with excursion detection time, time-to-information, time-to-corrective action, and the dispositioning process among the most important factors to be considered. A clear case in point is the quick adoption of the integrated metrology approach for CMP, which has led to significant improvements in controlling a process known for its high degree of variability.

The 300-mm transition and increasing interest in wafer-to-wafer APC have fueled interest in IM, especially with regard to the need for increased sampling on the large, high-value wafers, and better control of the shrinking process windows in leading-edge process technologies. Integrated metrology's overall economic and operational viability for widespread implementation (in applications other than CMP) has come under increasingly close examination by many tool suppliers and chip manufacturers. Factors and criteria under scrutiny include process control requirements, capability and sensitivity of sensors, history of process excursions, process stability, operational and dispositioning issues, and overall economics for targeted applications.

The need for wafer-to-wafer APC capability versus conventional statistical sampling with run-to-run APC should be carefully examined and weighed against yield and cost factors. Integrated metrology promises the added benefits of zero (or a very small) footprint and a shorter "net cycle time," to be achieved through reduced delay times and faster time-to-decision—all envisioned with minimal impact on throughput and cycle time of the process tool. However, tool-to-tool material delivery capability in fully automated 300-mm fabs, when combined with an optimized fab layout and intelligent scheduling, effectively leads to virtual clustering of stand-alone metrology with process tools, which can significantly reduce and undermine the IM cycle-time advantage.

Integrated metrology could potentially improve overall process equipment productivity through increasing process tool availability by optimizing preventive maintenance schedules and reducing wait time for process qualification. However, there are major concerns about the reliability of these relatively low-cost sensors and their impact on the overall reliability of the host multimillion-dollar process tools. Ultimately, enhanced productivity must be achieved at a cost per functional die (or wafer) out that is comparable to or lower than that of existing practices.

The economics of integrated metrology has become a major focus area. The use of fabwide integrated sensors may not necessarily be the cheapest approach. With statistical sampling, each stand-alone metrology system typically handles the measurement requirements of several process tools. In the case of integrated metrology, every tool will need to be equipped with sensor modules, and some stand-alone tools will still be needed. Although low-cost integrated sensors may be suitable for excursion detection of tools and processes with a high degree of variability, they may not provide the sensitivity of stand-alone tools. The real challenge is to prove that for a given application, integrated sensors would meet all the measurement requirements cost-effectively and would also result in significant improvements in process control and yield over the conventional methods that rely on stand-alone metrology tools in fully automated fabs.

Suppliers claim that they have overcome the problems originally encountered while integrating metrology sensors into process tools. Despite this assertion, integrated metrology has essentially missed its window of opportunity for widespread, high-volume implementation in the early 300-mm fabs as well as in production lines running 130- and 90-nm processes. Most future market growth is expected in etch and lithography applications, where integrated sensors would be part of an APC system. For instance, macrodefect inspection sensors have been integrated into recently introduced next-generation lithography track tools. Tightening process windows have also led to increased interest in additional IM uses in other select critical etch and lithography applications, such as gate patterning and resist trimming, at the 65- and 45-nm nodes.

Despite strong interest in integrated critical dimension (CD) sensors, the market acceptance and adoption of this optical scatterometry-based technology have been very slow. This industry reluctance has been caused by several factors, some of which have been viewed as shortcomings of this impressive technology. Below the 130-nm node, with complex subwavelength lithography, there is even more interest to "see" the actual device features. Engineers want to fully measure three-dimensional CDs and understand information such as shapes and profiles for devices "within the die." Scatterometry CD metrology is based on measurements made on standard grating structures in the scribe lines, which must be correlated to the actual within-die CDs measured by a scanning electron microscope (SEM), making it an indirect measurement. Even if these correlations were fine, overcoming the negative perception of engineers has still proven quite difficult.

Furthermore, because of computational challenges, scatterometry has not been operationally suited for measuring contact and via structures. As more-powerful and faster computers become available, this limitation should be overcome. From the fab perspective, these issues make scatterometry a partial solution that needs to be combined with CD-SEMs. What further complicates the picture is that CD-SEMs themselves continue to be challenged by the future requirements of the industry. There are also concerns about the impact of process noise on scatterometry. The matching and calibration of integrated metrology sensors for a given application, as well as the cross-correlation of integrated sensors with the companion stand-alone metrology tools, are also essential requirements.

OEMs, IM suppliers, and end-users have certain competing, and often contradictory, objectives that have negatively affected the market. Most integrated-sensor suppliers would like to see an open plug-and-play integration standard and a more direct interaction with the end-user. Some process tool suppliers have adopted this approach, but others—including several market-dominating players—have a closed architecture, relegating the IM suppliers to one notch down the supply chain. Certain users would like to be able to install the integrated sensors of their choice on tools. However, many users would also like to see the process equipment manufacturer take full responsibility for the entire tool, including the integrated sensors. The historically slow pace of developing standards is yet another challenge. In the meantime, experience shows that customized, small-scale solutions will very likely be used.

Sensor suppliers have been working hard on offering many innovative solutions, such as the simultaneous measurement of multiple process parameters and high-throughput systems providing high sampling rates. One company's approach measures film thickness, CD, and overlay in a single system. Another vendor has come out with a high-throughput, stand-alone integrated metrology platform with multiple modules, offering film thickness, CD measurement, and defect inspection in the same tool.

Integrated sensors will continue to evolve over the next few years. However, users will defer broad implementation until, for each application, the benefits of IM are thoroughly and fully investigated, determined, understood, and proven. From this effort, enabling and cost-effective applications will emerge and be adopted. In any event, fabs will still use a mix of stand-alone tools and integrated sensors, with the ratio depending on the user application and market validation results.

One of the recurring themes I've noted in previous columns applies all too well to the integrated metrology space: the industry will avoid changes that it perceives as unnecessary, too costly, or too risky. In order to succeed, any new tool or process must provide enabling technology solutions that cost-effectively meet critical industry requirements that cannot otherwise be met with conventional technologies.