Integrated Metrology Encounters Grudging Acceptance

Integrated metrology (IM) is the incorporation of a measurement system with a process tool to take measurements during a process (in situ), or at its conclusion without removing the wafer from the tool (in-line). This is unlike traditional stand-alone platforms (off-line), which require that the wafer be taken to them. IM systems may incorporate or require an array of systems and subsystems, including motion systems, various assemblies and robotics. IM's eventual goal is advanced process control (APC) in the fab.

"There's no cookie-cutter approach to IM," said Paul Blackborow, vice president of corporate marketing at MKS Instruments (Andover, Mass.). "There's an increasing level of complexity, so you must closely work with a sensor supplier who understands film chemistry or your deposition process's chemistry. The magic word is 'application-specific.'"

Dragan Podlesnik, general manager of the silicon etch division at Applied Materials (Santa Clara, Calif.), views IM as the next step in the evolution of process control. "Lot-to-lot control is already established within certain sectors of the industry with stand-alone tools. However, with further technology shrinks, the ability to reduce variations within lot, i.e. wafer-to-wafer control, is required. Realistically, only IM offers the flexibility and productivity to enable wafer-to-wafer control in volume production."

IM was first done successfully in chemical mechanical planarization (CMP), according to Gloria Johnson, product director of integrated products at Therma-Wave (Fremont, Calif.). "You had a process that was out of control. With CMP, each wafer is processed differently because the pad changes, the slurry gets a bit altered — it's very dynamic. Conversely, our experience over the past three years with IM for track and etch shows that these processes don't have a situation where you must inspect every wafer, because the process keeps changing."

The Applied Materials Transforma patterning system integrates Nanometrics’ Optical CD (OCD) metrology unit to reduce final transistor gate CD variation. (Source: Applied Materials/Nanometrics)

Device manufacturers are being buried by data. "Being a systems company, we have most of the critical data sources that surround the tool and influence the process environment — we maintain the process environment," Field said. Process windows for whatever manufacturers — either OEMs or users — are doing are shrinking, so applications that help manage the process window are necessary.

Monitoring the process window in real time is important. "We implement solutions through an overall system approach involving hardware and software," Field explained. "We're working on providing endpoint for the clean cycle required with CVD processes. Whenever you process a wafer, you also deposit dielectric on the chamber. Eventually, the chamber needs cleaning." Usually, this has been based on the number of wafers deposited. "This requires a considerable built-in safety factor," he added. "It consumes expensive NF₃ and can cause wear in the reactor's parts, requiring periodical replacement and chamber requalification affecting quality and time to manufacturing."

Optical solutions have been tried, such as a scanning monochrometer that monitors the plasma emission to detect effluent changes. However, it is not reliable because the clean etches the window that the camera looks through. Advanced Energy has an rf sensor that fits at the line and "looks" at things in real time. "The signals go to an application where we implement algorithms that work on a variety of process conditions, based upon the clean's specific nature," Field said. "Rf clean endpoint is reliable because nothing gets in the way of the rf sensor accurately 'seeing' the chamber."

"One of our etch products has OCD IM," said Applied's Podlesnik. "We've gone from optical to CD-SEM, back to optical, which is now far more sophisticated. IM's here. We already have it in a variety of tools — deposition, etch and CMP." For this, incoming material must be measured, and a chamber or system must respond by doing something (Fig. 1). "The idea is using on-board metrology to improve incoming material," he said. "This means we measure wafer parameters that are then used to adjust etch conditions. Already, the introduction of this technology into production has confirmed IM's capability to reduce incoming variation by more than 50%."

1. CD metrology has experienced a number of incarnations, dictated by the need to control processes — going from optical to CD-SEM and back to optical. (Source: Applied Materials)

Applied wants to offer a form of post-process metrology. "We're seeing significant benefits from this approach," Podlesnik explained. "Today, after etching or deposition, one removes the wafer, submits it to cleans and measures it. The data reveals whether you did well or if, in spite of in situ metrology, you didn't. Doing this in the same system is a benefit. Envision a metrology unit that measures every incoming wafer, and the same system also performs an outgoing measurement." Because the information would be shared with the system, the user would immediately know whether a wafer is processed to spec. In addition, the system could also do closed-loop control — that is, matching tool-to-tool, fab-to-fab, worldwide.

Another sought after improvement is fab cycle time. High-end chip runs are controlled by sending one wafer ahead through the entire process when the lot comes in. Everything is tested, it goes through metrology and, based on the data, the decision is made whether to send the whole lot. This can take four to 24 hours.

"With IM, everything's done on the platform, without removing the wafer and thus reducing cycle times," Podlesnik said. "Every wafer acts as a send-ahead wafer. We can bring in a wafer, pre-measure it, process it and post-measure it. After any event — idle time, wet clean, whatever — every tool must be qualified. That cycle is also reduced because on the same platform we can measure pre- and post-etch parameters, and extract etch rate, uniformity and all other parameters." Podlesnik also noted that at least one stand-alone tool is necessary to ensure that all recipes work the same, and that everything should be networked. "You may can have fewer stand-alone instruments, but you still need some. Stand-alone will never completely go away."

There is a strong push for integrated or in situ inspection and metrology to increase throughput and reduce dwell time, especially in 300 mm facilities, where time is more expensive, said Anantha Sethuraman, vice president of the semiconductor business unit, microelectronics product group, at FEI (Sunnyvale, Calif.). "There have been some advances in integrated film measurement and integrated CD measurement. The cost of scrapping a 300 mm wafer is about two-and-a-half times or more than a 200 mm wafer, making a premium out of the capability to detect, monitor and control any excursion before it hits."

IM does not provide a stand-alone tool's level of accuracy, although precision and resolution advances have been made, Sethuraman said, citing in situ CD integrated into some of today's lithography tracks. "There isn't much of a solution available with other defect and metrology tools for integrated applications, partly because there is a resolution limitation in the configurations that you can use to perform the measurements. There's also the matter of space vs. resolution." Thus, the next best alternative to IM is in-line or in situ metrology.

Inspecting wafers with photoresist still within the lithography cell, which have not yet gone to etch, will become important, said Joe Danko, vice president of engineering at Inspex (Billerica, Mass.). "Users are concerned that IM on the system after the fact may require a new litho track, because the platform wasn't originally designed for on-line, on-track macro inspection." Another concern is that some systems run 120 to 140 wafers an hour but, on a litho track, they must run at the track's throughput, which may be half as fast. IM cannot just be put into a platform; both must be designed together.

Inspex is targeting the 90 nm node. "Some defects that we see on these 0.1 µm design rule wafers are so small that it would be difficult to detect them with in situ or some other integrated system, vs. a stand-alone where you get the full benefit of granite stages and other factors — they require positional accuracy to detect," Danko said. "For example, we do die-to-die comparison, time-delayed integration — both of which rely on X-Y accuracy and repeatability. This requires a stable, mechanically isolated environment. When it comes to looking at problems such as small defects, stand-alone is still king. Macro inspection (>1 µm) is easy to implement; micro (<1 µm) is not."

The integrated market is immature and the technology is still being qualified, according to Tom Long, senior vice president of business development for lithography and parametric solutions at KLA-Tencor (San Jose). "A major reason why IM hasn't been extensively adopted is that many sensors lack the performance capability of today's leading stand-alone tools. IM must offer the same or better process control capabilities than stand-alone tools at an acceptable CoO, and be extendible across two technology nodes for good ROI."

According to KLA, another issue with IM's minor penetration is that the ROI is unclear to the user. "We must question the value of looking at each wafer vs. doing intelligent lot sampling and making corrections on that basis," Long said. "Users also have concerns about the reliability impact of integrating these modules into process tools. When adding integrated sensors into a lithography cell for measuring macro-defects, CD overlay and film thickness, you're no longer just taking into account the track and stepper's reliability. You must look at the reliability of all those elements combined. If something goes wrong with one integrated sensor, the entire litho cell may no longer meet production process requirements."

KLA is developing IM sensor technology, starting with the premise that the measurement precision, accuracy and matching should be as good as stand-alone. "By using proven technology, such as scatterometry, we're making incremental investments to provide metrology sensors that can be integrated in a stand-alone cluster (stand-alone with two or more integrated sensors), or process tools depending on which implementation provides the best ROI," Long said. "The issue with IM is balancing reward, in terms of productivity and higher and faster yields, vs. the risk of missed excursions, lost productivity and lower-performing products."

Lam Research (Fremont, Calif.), however, designs its tools with IM in mind. "We cannot exactly predict what the metrology will be, but we can ensure that integrating different types in an almost plug-and-play manner will be as straightforward as possible," said Dave Hemker, vice president of new product development. "All of our products not only have the 'space' for IM, but the software and control system support as well."

Lam works not just with users, but also with metrology suppliers to target the most appropriate IM for various processes. Their main-line products — CMP and dielectric and conductor etch — have IM programs, with CMP focusing on copper measurement, thickness and residues; and dielectric and conductor etch looking at CD measurements for either feed-forward or feed-back control.

"We consider IM because people want process control," said MKS' Blackborow. "IM's purpose is to quickly show whether you got what you wanted. There's no silver bullet, particularly when you look where processing is headed with all the new films. Until recently, we dealt with only four or five chemical elements to make a chip. Now we're dealing with many more in the chemical mix — to make low-k and high-k dielectrics, new gate metals, and copper and barriers. No one technique can manage all films and processes."

MKS is focused on process as well as wafer metrology. "Process complexity is driving towards tool instrumentation," Blackborow said. "Most fab processes are gas-based processes, whether ALD, CVD, PVD or etch. We provide much of the instrumentation that measures the flow rate and controls it, measures the pressure and controls it, measures the rf power, etc. However, considerably more metrology is being done of gas composition, and that's where add-on gas monitoring comes in. We use FTIR-based spectroscopy and mass spectrometry to determine exact gas mix in the chamber. On the rf side we don't just measure power, but also impedance, harmonics, etc., because these show what's really happening in the chamber. We're working with customers to combine the data they get from the traditional process control sensors with new process state sensors like the rf probe and FTIR. In combination, they can better hit their process targets and know when the process is drifting out of spec." (Fig. 2)

2. Key process parameters are requiring a level of control unprecedented in the industry’s history. IM modules are becoming a crucial part of the new processing environment. (Source: MKS Instruments)

However, processes have many variables, and platforms must be instrumented to cope. "You could instrument every parameter, but for CVD there are hundreds of potential parameters — flow gas rates, pressure signals, rf signals and harmonics, gas composition spectra, etc.," Blackborow added. "A data flow of this magnitude could kill the fab's IT system. The concentration will be on reducing that data to manageable fault detection information. We're using our expertise in data management and connectivity to ensure that we enable a flow of information, rather than data to the fab or OEM."

Jason Rollo, director of business development for integrated metrology at Nanometrics (Milpitas, Calif.), believes current manufacturing process and fab economic developments favor IM. "A chief IM driver is process efficiency. Unfortunately, in today's economic conditions this emphasis has tapered. However, higher wafer sampling is essential for wafer-to-wafer control, and the only viable solution to achieve this efficiently and cost-effectively is through IM."

There are two schools of thought on IM. One is where the equipment manufacturer wants to deeply embed it in system, making it transparent to the marketplace. The difficulty with burying the IM module in the tool is that most OEMs have a sketchy knowledge of metrology, which complicates IM support. The solution is to have it as an integral part of the tool, but available for the metrology company to support it. "There's a serious trend for companies adopting IM, where the OEM owns the process and the IM supplier owns the metrology, with both going to the user's premises," Rollo said.

Robert Loiterman, senior vice president of technology and general manager of integrated metrology at Rudolph Technologies (Flanders, N.J.), agreed with this view, and said he sees three parties involved with IM. "There's the IM supplier, the process tool supplier and the device manufacturer. For IM to succeed, these must partner and work together to identify yield improvement applications and areas where IM provides cost-saving solutions. As leading-edge fabs implement APC, IM that links the process tool to the APC system will become an essential component of yield enhancement."

Therma-Wave's Johnson views different requirements for IM. "Lithography's needs are different than CMP's. Things change at 300 mm — you have automated wafer handling and the logistics to go from track, stepper and metrology station, and then on to the next level. Wafer handling costs almost surpass metrology's.

"Still, cost-wise, metrology is a safe bet. At 300 mm, you really want to catch a problem sooner. At 300 mm and 120 wafers per hour, IM can pay for itself very quickly by rapid fault detection. Even if it's a reworkable catastrophe, such as no resist on the wafer or the wrong reticle, you still want to catch it earlier." With shrinking geometries and more chips on a larger wafer, metrology is entering the APC sector, refining the process to get the best possible yield.

Weng Yang, director of marketing at Timbre Technologies (Santa Clara, Calif.), believes users are testing the waters to identify applications that they cannot do without. "Most have identified scatterometry's future in an IM setting; integrated into track and etch. So even when the evaluation process is taking place, mostly on stand-alone systems, the focus is on IM. For 300 mm they want more metrology data from the wafer for control."

For the foreseeable future, the purpose of IM is not to replace the stand-alone system. No conscientious OEM will tell his customer, "I'll sell you my system with IM — cancel the stand-alone tool." IM must add value to the device manufacturer's process and produce a direct, positive result. Then, and only then, should it be designed into systems.