Experts At The Table: MEMS Challenges

Semiconductor Engineering sat down to discuss the challenges of MEMS with Rakesh Kumar, senior director of the MEMS program at GlobalFoundries; Tak Tanaka, managing director for Applied Global Services at Applied Materials; Paul Lindner, executive technology director at EV Group; and Alissa M. Fitzgerald, founder and managing member at A.M. Fitzgerald & Associates. What follows are excerpts of that conversation.

SE: In our last conversation, we talked about the challenges in MEMS. Are there more challenges in the arena?

Kumar: If you look at the CMOS market, one fundamental thing behind it is standardization. You can have one process and you can serve multiple customers. In MEMS, CMOS foundries came into this market later than most. We started when MEMS customers were already in the market. They had proprietary processes. Then, they came to us and said: ‘Do this and we don’t want any change.’ So in some cases, people have been selfish when it comes to working with MEMS foundries. They look at their processes as a competitive advantage. They say: ‘We have the know-how and we don’t want to share it with the world. We want to maintain our competitiveness by keeping the process proprietary.’ And the traditional (IDM) foundries supported that. And so, these foundries did not try to establish common standards. They did not want to share that type of information.

SE: Then why not just develop standard processes to get around this problem?

Kumar: The development of standard processes is still in the early stage. But as a foundry, we have a big question. Let’s say we bring in or offer a standard process, but the question is will there be any takers? It is very easy for us to set up a line—if we get a large volume customer who has a proprietary process. We can do it. But if we want to have multiple customers with different technologies, then how do we make the best use of our equipment? But if you develop multiple processes, and you have only a limited number of customers, we can’t use the tooling to maximum the utilization. That will not help us in terms of reducing the cost. So, the development of standard processes will happen, but we need large volumes to succeed.

Tanaka: That is really an ROI challenge. Plus, everyone wants to differentiate their products at the same time. So that is also a challenge.

SE: Are they any solutions to the problems in MEMS manufacturing?

Lindner: We design our technology around a number of applications. And we try to come up with a universal process module, which feeds all of these needs. So we came up with a universal bonding chamber, which can do 50% higher temperature than 70% of our customer needs. The technology is not aligned to a standard or industry roadmap. It is aligned to our customers and what we see in the market. In addition, another way to obtain a cost-effective MEMS process after the design is to standardize. So there could be one packaging type, one hermetic seal, and one process module standard. But this has not happened in the last 20 years and I don’t see it happening for the next 10 years at least. Packaging technology, design and processes are proprietary and the competitive weapon for our customers. So we keep our equipment universal and have process development projects with individual customers to tune the processes and bring them into high-volume manufacturing.

Kumar: We think standardization is very much possible in MEMS. There are not many families of MEMS products that are in large volumes. Still, fabless companies have very innovative ideas, but can they afford the cost or the development times? We think we can develop application-specific standard process modules. This is our belief. But first, we are trying to serve customers who come in with their own, customized processes. Ultimately, as a foundry, our objective is to come in and standardize the processes. We want to provide design kits to our customers. This will take some time and it’s not easy to identify the common needs for various product segments.

SE: Any other ideas?

Kumar: We have to start with so-called basic building blocks, where we need at least some standardization in terms of some of the common processes, materials and tools. Even in the MEMS tools right now, there is a need for more development and improvement. Many of the tools are not fully automated and the throughputs are a big issue. They have a direct relationship to the cost of the product. If you have a bonder, which has a throughput of one wafer an hour, it’s not very cost-effective. MEMS must also be supported by design enablement. That means we need to come up with reference designs and design rules to bring up these products. Once these things are integrated, then the cost of development and the cycle times can be reduced significantly.

Lindner: Higher throughputs on specific tools obviously are a main driver on our roadmaps. Throughputs are, however, dependent on the processes that have been established. For example, every process that has been developed for bonding has its own roadmap. What can be done to accelerate it? There has been major progress over the years.

Tanaka: There is no magic in equipment manufacturing, especially for MEMS. We want to improve our customers’ device performance. We want to improve yield. Having said that, the MEMS challenge for us is two things. One is manufacturing. We don’t build like a thousand CVD tools for MEMS. Maybe a few here and few there. The second thing is that the standardization for MEMS is not there in equipment as well. We need to overcome the volume manufacturing and standardization issue. So, the question is how do we customize each process? Let’s take PVD aluminum, for example. PVD aluminum can be used for many applications. First, we try to see what the size of the market is in power semis, sensors, MEMS and other markets. Then, we try to combine these markets to determine our ROI. Then, we make a basic equipment technology. And then, we specialize it to each customer’s needs. That’s ideal. This is 200mm equipment. In CMOS, some big equipment vendors don’t care about 200mm. But a lot of smaller equipment vendors have 200mm equipment. So it’s very difficult for us to unite everybody and come up with standards. In order for us to differentiate, we really need to come up with a universal concept. That’s what we are trying to do.

Fitzgerald: There are needs on many fronts. Standards could be an enormous help and there are some decent efforts in standardization. NIST has made some efforts to standardize some material test approaches. Most recently, Intel and Qualcomm, in collaboration with MEMS industry groups, and a number of other people, came up with testing standards. These types of efforts will increase in intensity and focus, because people are beginning to understand that we have a time-to-market and cost problem in this industry. Certain market opportunities will be closed off to us—if we can’t get the products out in time and can’t get them there at the right price. I don’t think we can standardize every aspect of MEMS fabrication, but we can start with some of the easier places. For example, we could start with specifications with SOI wafers. There are no standards now. Every order is a custom order. They don’t create inventory. And this creates a major problem. If you place an order for an SOI wafer today, it’s a six-to-eight week lead time. That’s just one small corner of the universe. There are other areas, such as design rules, process capabilities, and best practices for certain types of devices.

SE: Any other thoughts on tool development?

Kumar: In wafer-level packaging, you had DRIE equipment and bonders. These tools were used for MEMS on a very small scale. There were only a few vendors, however. But once more advanced 3D wafer-level packaging technologies came into being, new DRIE vendors entered the market. Bonder vendors spent more R&D dollars to improve their throughput. With the growth of MEMS, and the entry of foundry companies like GlobalFoundries and TSMC, there will be more demand for tools and improvements of the tools.

SE: And what about standards?

Tanaka: Standardization takes time, but it will happen.

To view part one of this roundtable discussion, click here.