As semiconductor features grow smaller, inspection equipment developers race to find wafer errors before they hinder production.
Traditional lithography is based on a simple principle: Oil and water don’t mix. The method, first developed by an actor in Bavaria in 1796, used a smooth piece of limestone on which an oil-based image was drawn and overlayed with gum arabic in water. During printing, the ink was attracted to the oil, and was repelled by the gum.
Today, a new principle guides a very different lithography: Defects and wafers don’t mix.
Though lithography has advanced far beyond the simple technology used to create billboards, the same goal applies, except that the finished product must be a nearly perfect facsimile of the original. For semiconductor foundries and chipmakers, the goal is more elusive each year. As Moore’s Law states, the number of transistors in an integrated circuit doubles every two years. Even if this pace is slowing, as some experts suggest, the challenge is high to ensure that nanoscale features operate as designed.
Part of this effort is wafer inspection. Even though a typical 300-mm-dia wafer is made from an almost entirely pure, crystalline semiconductor, usually silicon, it will contain defects. These defects, as well as defects added in subsequent processing steps, can have a deleterious effect on chip performance, quality and consistency.
Wafer inspection is difficult for several reasons. First, feature size is extremely small. To ensure nanoscale devices work properly and reliably, top-down process control at less than 10 nm is almost mandatory. Second, chip development involves arbitrary patterning capabilities and fast prototyping. Inspection strategies must therefore be generalist in nature, able to cope with a steady succession of new designs. Finally, defects aren’t few and far between. Even in the exacting world of semiconductor fabrication, thousands of defects may exist on a single wafer. The goal for manufacturers is to minimize those defects. Or, at the very least, eliminate the defects that may have the worst effects on production.
KLA-Tencor Corp., Milpitas, Calif., has addressed this need for semiconductor and nanoelectronics industries with a variety of surface analysis tools that incorporate advanced imaging technologies. In July, at the SEMICON West conference in San Francisco, the company made a substantial revision to its core set of wafer inspection technologies. The new “Inspection and Review portfolio” includes four systems designed to provide advanced defect inspection and review capability for the development and production of integrated circuit devices at the 16-nm node and smaller.
Coping with size changes
The semiconductor industry is in the midst of a transition to 16-nm, 14-nm and even smaller design nodes, and chipmakers are experiencing increased process control challenges. KLA-Tencor’s latest innovations reveal the attention paid to this progression in the market.
As lithography technology continuously shrinks, the defect number becomes a significant issue in wafer manufacturing inspection. A certain number of defects results in what the industry refers to as “review and classification” problems. Defects are numerous enough (in the thousands) and occur with such variety that they have specific family names, including systematic, random and nuisance. Defects can result from any stage of the fabrication, from deposition to etching, and can include protrusions, polish marks, pits and scratches. The most effective way to reduce the defect numbers is to inspect the potential risk area instead of the entire chip. A highly capable optical inspection system carefully pre-selects this care area to get the most benefit from an inspection. This can dramatically reduce defects and enhance throughput.
KLA-Tencor’s 2920 Series defect inspection system encapsulates many of KLA-Tencor’s recent innovations in optics, built around a third-generation broadband plasma illumination source. This platform delivers twice the light of the previous iteration, the 2910, enabling the use of a new deep ultraviolet (DUV) wavelength band. The instrument is also equipped with the industry’s smallest optical inspection pixel, which provides increased resolution for detecting tiny defects like subtle protrusions.
The combination of high-resolution optics and a low-noise sensor allows the 2920 Series to deliver sub-0.5-µm defect coordinate accuracy, allowing it to accurately identify and target care areas. One standout innovation that allows the instrument to reach this level of accuracy while providing a 30% improvement in throughput over the 2910 is a technology called NanoPoint.
During chip development, this suite of detection technologies helps find the wafer defects that most immediately threaten yield, accelerating the identification of design issues that reveal the need for mask design. According to Keith Wells, VP and GM of the Wafer Inspection (WIN) Div. at KLA-Tencor, NanoPoint’s value has already been demonstrated on early metal layers, where line-edge roughness on dense patterns had previously limited the ability to detect “yield-killing defects” inline at advanced nodes.
“Our challenge is to design equipment that can discover defects whose size is further and further below the inspection wavelength,” says Wells. “In the past we have offered various improvements to the light source, optics and other subsystems, but NanoPoint addresses the issue from a new angle.”
According to Satya Kurada, product marketing manager at KLA-Tencor, the 2920 Series can generate care areas that are significantly smaller, and can choose them from smaller regions. These care areas are generated in the millions, and are based on user-defined patterns of interest. It can also remove noise from the pattern of interest, helping focus inspection resources on critical patterns.
NanoPoint’s greatest value is during chip development. The improved ability to find care areas can reveal a need for mask redesign in a matter of hours. This result can shrink redevelopment time from months to days. As volume fabrication is occurring, NanoPoint can also selectively track the presence or absence of defects based on patterns.
Supporting NanoPoint are Accu-ray and Flex Aperture technologies that quickly determine the best optical settings for capture of critical defect types. These measures have reduced the time to determine the best optical mode by a factor of four, while also doubling accuracy.
This improvement in capability is important, says Kurada, because of what’s happening with feature sizes.
Traditional optical methods are having trouble simply finding defects now. KLA-Tencor’s new approach, to use pattern-based inspection care areas, is a big step forward, he says.
Complementing the 2920 Series inspectors, KLA-Tencor’s Puma 9850 instrument captures defects across a wide range of wafer fabrication steps, such as photo-cell monitoring, chemical-mechanical planarization and process wafer qualification, at cost-effective throughput. It also handles the detection of etch defects. Improvements include a faster stage and better data rates for its sensors.
In addition to careful inspection of a patterned wafer, rough or smooth unpatterned films must also be inspected after deposition and polishing to ensure that process tools are not adding defects. Production-line examinations are also useful, particularly for sub-20-nm nodes. For this purpose, KLA-Tencor offers the Surfscan SP5. Like the 2920 Series, DUV technology produces high defect sensitivity at production throughputs. The Surfscan can find substrate or blanket film defects that can inhibit successful integration of multi-stack IC devices. In addition to defect detection, it offers the ability to create full-wafer “SURFimages,” or haze maps, which represent the variations of the wafer surface quality in response to changes in process chemistries or recipes. These maps help developers determine surface roughness and grain size.
Each of these inspection systems are designed to seamlessly connect with the fourth and final update—the eDR-7110 electron-beam review system. This system is an evolution of KLA-Tencor’s e-beam line and includes a scanning electron microscope (SEM)-enabled Automatic Defect Classification (S-ADC) engine that can produce an accurate representation of the defect population during production. S-ADC results can automatically trigger additional in-line tests, such as compositional analysis or imaging with alternative modes, while the wafer is still on the instrument.