One-on-one with the chairman of the group overseeing the shift to larger wafer sizes.
By Mark LaPedus
Semiconductor Manufacturing & Design sat down to discuss 450mm metrology challenges with Menachem Shoval, a former manufacturing executive at Intel and chairman of the Metro450 consortium. The Israeli-based consortium is developing metrology technology for the next-generation, 450mm wafer size. The group consists of Intel, Applied Materials, Jordan Valley, Nanomotion, Nova and universities in Israel.
SMD: In general, what are the challenges in metrology?
Shoval: Even without going to 450mm, there are huge challenges for metrology in terms of going down from 22nm to 14nm to 10nm to 7nm. The ability to detect defects that are 20nm in size is almost beyond the capability of light. There are high-resolution tools like e-beam, but e-beam is very slow. The challenge of 450mm is coming on top of this. In bigger wafers, the thickness of the wafer is not proportional to the size. These types of wafers will be kind of flexible. There is also a huge amount of data that needs to be calculated. The question is how do you handle the wafer? How many supporting points do you have? What about the contamination issues of these wafers?
SMD: What are the general requirements for a metrology tool at 450mm?
Shoval: The sizes of the defects are getting smaller and you have a larger number of them. A system must be smart enough to filter one defect from another. And you still must run at a given number of wafers per hour. This is a huge challenge.
SMD: Those are big challenges, but why start a consortium like the Metro450 in the first place?
Shoval: The semiconductor industry plans to migrate to 450mm wafers. But this could be a disaster for companies that are not so big. There are many companies that can’t invest six years in advance. So, several metrology tool companies requested support from the Israeli government for 450mm. And the government suggested that toolmakers create a consortium.
SMD: Israel is a hotbed of metrology activity, right?
Shoval: Together, Israel supplies over 30% of the metrology tools to the semiconductor world. It represents a $350 million industry, or roughly one-third of the worldwide metrology business. There are five big companies in wafer metrology located in Israel. I am referring to a division of Applied Materials, which makes SEM tools, defect-review tools, defect inspection and mask inspection. All of this is done in Israel. I am also referring to KLA-Tencor, which has a division for registration tools and optical CD. I am also referring to Nova, which is also doing optical CD. There is also Camtek and others dealing with inspection in the back-end. Jordan Valley is a unique company doing X-ray metrology.
SMD: What is in the Metro450 consortium?
Shoval: We formally started in July 2012. This consortium has five companies. There are four tool vendors—Applied Materials, Jordan Valley, Nanomotion and Nova. We also have Intel. We also have 14 researchers from various universities. About 60% of the money in the consortium is being invested by the Israeli government. The rest is coming from the companies.
SMD: What is Intel’s role?
Shoval: When joining the consortium, the companies expected to get money from the government. Intel did not request money. But Intel is invested in the technology. So for them, the ability to be here and to be familiar with the technology is important. I would even say they are able to lead the development in directions they think are relevant.
SMD: Is KLA-Tencor part of the group?
Shoval: Unfortunately, no. We want KLA as part of the group. We have a good connection with them. But for formal and legal aspects, they were not able to join.
SMD: How about GlobalFoundries, IBM, TSMC or Samsung?
Shoval: You have to have an entity in Israel. Intel has a fab in Israel and so they can join. Applied has an entity so they can join. If TSMC would build any development center here, they could join. We do collaborate with other consortiums. For example, we are talking to the G450C. We are also talking to other 450mm organizations in Europe as well.
SMD: What is the goal of this consortium?
Shoval: The goal is not to make a tool or a platform. Each company is developing its own technology. And they are competing with each other. But we can collaborate only on those parts that are common with all of them and are not the main IP of a company. In terms of the boundaries, we will work on platform technology, but not on detection. We also put in front of us a target, saying we will develop capabilities for high-volume manufacturing tools.
SMD: The consortium is looking at which node?
Shoval: We would like for companies to be ready for high-volume manufacturing. We are not saying that since 450mm will come at maybe 10nm, the tool needs to be ready at 10nm. This is the task for the company itself. We are dealing with the platform capabilities. We plan to complete our work in three years. So the companies will still have about two years to complete the development of HVM tools.
SMD: How will companies collaborate in the consortium?
Shoval: Each company will develop its own platform, but they are all sharing the ideas in the consortium. For example, if they come up with a good idea, it will be used in development. There will be many ideas. Each company will choose what’s best for them.
SMD: What has the consortium developed thus far?
Shoval: We have already done work on the ability to model the stages. The group has also shown the ability to develop non-linear stage capabilities.
SMD: You have five working groups, right?
Shoval: We have five working packages or groups. The first one deals with wafer handling, chucking and stepping. Larger and heavier wafers, faster stage movements and higher acceleration rates require new handling and chucking techniques. The question is how do you do that? The goal is to develop advanced, accurate servo algorithms based on nonlinear dynamic parameters.
SMD: Is that where Nanomotion fits in?
Shoval: Nanomotion has a way to drive stages. They have piezoelectric drive motors. Piezoelectric has a lot of advantages in terms of energy, contamination and reducing vibration. On the other hand, it isn’t fast enough. So in our consortium, they are working to develop faster stages, which may fit the need of 450mm wafer tools.
SMD: What about the second group?
Shoval: The second working group is developing sampling optimization. Some people call it adaptive sampling. Metrology is all about sampling. You don’t want to measure more than what you really need. In fact, metrology doesn’t add value to the wafer. If somebody could, they would eliminate metrology all together. But that is impossible. So, we would like to measure enough just to gain control. Today, the sampling is set by testing. At the beginning of the development, and later on, sampling can be reduced from time to time. If the tool is smart enough, it would be able to measure just enough. A 450mm tool would have kind of a smart engine in it to decide by itself how much to measure. This could be how many wafers in a lot or how many points on a wafer. Or, are we going to measure every lot or just every fourth lot.
SMD: What about the third group?
Shoval: The third one is dealing with wafer cleanliness, contamination and breakage. There are two aspects to this. Metrology can detect the damage to the wafer. We are also talking about airborne molecular contamination. The idea is to demonstrate a simple method for online airborne contamination detection. In another example, Jordan Valley, with their X-ray capabilities, might be able to detect micro cracks that can cause breakage on the wafer down the line.
SMD: And the fourth group?
Shoval: The fourth working package deals with standard calibration of the wafer. Each company is developing its own way in order to make sure the tool is calibrated and ready for production. But Intel, Samsung and others would like to test it with their own wafers. Usually, those are production wafers. The idea is to develop a standard wafer that will be used as a gauge for all metrology tools.
SMD: And finally, what about the fifth group?
Shoval: In the fifth working package, we are looking at fast data collection and processing. It’s known today that standard computers used for controlling the measurement tools are becoming as big as the tool itself. They demand a lot of energy, thereby exposing heat. And that’s because no one is designing a control system specifically for metrology. For example, if you inspect a wafer for defects, and you have thousands of defects on the wafer, you need to compare one die to another. This is a huge amount of data, which you cannot store. So what we are doing with the group is trying to develop a unique control system in which parallel capabilities will be taken into consideration.