The Measure of All Things

Without doubt, metrology is becoming increasingly more important as the industry works its way down to the smaller nodes. Some of the main challenges lie in the lithography area. Right now, we’re essentially facing two options: 193-mm immersion lithography with double patterning, or EUV litho.

At the recent SPIE Advanced Lithography Conference held in San Jose, I spent some time with Dr. Kurt Ronse, IMEC’s lithography department director. He is responsible for the Advanced Lithography Program, which covers 193-nm immersion, double patterning, and EUV lithography. During our conversation he mentioned that, whichever technology decision the industry makes, metrology will be an increasingly important tool to achieve it. Ronse pointed out that when approaching 22 nm and preparing to go beyond, it becomes more important to be able to measure with sufficient accuracy the dimensions produced to determine whether they meet CD uniformity requirements. In both IMEC and Albany Nanotech, an important part of the EUV and immersion lithography research taking place is the evaluation of different metrology alternatives, such as scatterometry, although old standbys like CD-SEM are far from being discarded.

The need for metrology of unprecedented accuracy is also making itself felt in the overlay field. Overlay must become staggeringly precise, which means that the metrology used to measure the overlay marks must evolve to match the requirements.

None of this should lead anyone to think that we are actually reaching fundamental limits in metrology; that will eventually happen, but not anytime soon. However, it does mean we are entering a new era during which engineers will have to take advantage of the capabilities offered by different techniques, to be able to measure overlay marks with sufficient accuracy. Traditionally, overlay has been measured by using imaging-base overlay metrology—essentially a camera. Increasingly, scatterometry is being seen as a choice, because it provides more precise data; there is less variation in the data obtained, reducing yet one more contribution to error.

                ASML's Alpha EUV Demo Lithography Platform. Source: ASML

IMEC has been working on process control for an immersion cluster, using a 193- nm ultra high NA immersion exposure scanner linked to their track. According to them, this combination of machines has been capable of controlling linewidths over an entire batch of wafers to the level required for <32-nm half-pitch printing. The problem is that to achieve that resolution in 193, it is necessary to print well below the wavelength, resulting in a considerably smaller process window.

Unprecedented process control will be required for production. Any small drift of any parameter in any of these tools results in CD uniformity or overlay performance going out of spec. Thus, in the wafer exposure and developing loop, going on down the line, there will have to be inline metrology techniques and tools to constantly monitor CD uniformity and overlay performance that, as soon as they detect anything deviating from spec they must be capable of identifying the deviation’s root cause—is it a focus or a dose drift, a hot-plate uniformity variation?—and automatically feed back the data to the scanner or the track, to immediately correct the problem and continue production, with everything keeping to the target linewidth and overlay performance. All this requires very precise metrology, because one is looking at very small effects—deviations of CDs that must be immediately be responded to, which, in turn, means that very fast metrology will be required, because if it takes too long, it is going to affect production.

IMEC has demonstrated that with some built-in dose correction schemes in the scanner and built-in hot-plate uniformity tuning, it is possible to obtain over a large number of wafers a CD uniformity >2 nm, 3σ, which is a factor improvement of 3 compared to what would be had if the tools were allowed to run without taking these fingerprints into account and doing the necessary feedback loops and corrections. Soon, all of this will be essential for volume manufacturing.

In the EUV lithography area, a problem of a magnitude only slightly smaller than that of the source power difficulties is reticle defects. We aren’t as yet aware of this problem’s full magnitude because there are no sufficiently good inspection tools that can measure and find the smaller defects and count them on an EUV reticle, in that way providing an idea of the problem’s true extent. Currently, we’re working blind in this area, with the only way to find these defects being to print them on wafers and then search for repeating defects. This is a current problem with the EUV Alpha tool, which is now working with larger feature sizes than the considerably smaller ones expected to be printed with high-volume manufacturing EUV scanners.

It is not too early to begin to develop the tools for EUV reticle inspection. However, because these will need to operate at EUV frequencies, they will be very expensive. This problem is compounded by the fact that the mask-making market is limited, and the developers and manufacturers of these tools could not expect a very high volume of customers; perhaps only five or so, each buying just one.

Although there is always the chance that the reticle problem will not be as large as currently believed, at present it is dangerous to assume that it is otherwise.

Faster and more precise technology, as well as superior inspection resources are starting to be needed now. And this metrology is not just needed for double patterning, EUV, and processing.

Apropos of Nothing…

As mentioned, I went to the SPIE Advanced Lithography Conference. The attendance was—to put it mildly—dismal. Several companies that I talked with at the event indicated that, prompted by the bad economy, they were either considering or already had decided to drop out of other shows, even SEMICON West. Although most technical papers featured in conferences could just as well be presented over the Internet as part of a webcast, including a Q&A session at the end of each paper, the networking that takes place at a conference or event could still be something best done at a personal level, not by telepresence. However, increasingly I hear people say that the time of the big conferences and trade shows is past. What say you?