New Apps For 3D Chips

Thick interposers and proven reliability will go a long way toward making stacked die a reality.

popularity

By Mark LaPedus
Semiconductor Manufacturing and; Design sat down to discuss the 3D device challenges and applications with Peter Ramm, head of the department for device and 3D integration at Fraunhofer EMFT Munich, one of Europe’s largest research organizations.

SMD: Fraunhofer was a pioneer in 3D chip R&D, right?
Ramm: We are the oldest microelectronics institute in Germany. We started in the mid-1980s with a big project in 3D. We were one of the first to work on this after NEC, IBM and others. We started with a 3D integrated circuit. We made an NMOS and PMOS layer. We made a second layer. We made vias through the layers at 2 microns. At the end of the 1980s, we realized a CMOS ring oscillator that was 3D integrated. But we were much too early with 3D IC integration. There was no real application division of a company that said: ‘Yes, this is exactly what we need for a product.’ So, we started looking at lower integration densities. The next project for us was in the direction of less vertical integration density using more smart technology.

SMD: What are some of the projects you are working on?
Ramm: We are working on interposers. We don’t like the term 2.5D. In fact, what they today call an interposer was called an MCM several years ago. MCMs existed some 15 years ago, but it was not a success story. So, no one is talking about MCMs. Now, the industry is talking about interposers. The disadvantage of the interposer is that it’s expensive. It’s not a cheap kind of 3D integration. You have to deal with thin silicon. You have to do a redistribution layer.

SMD: So what’s the solution?
Ramm: We are working on a thick interposer technology with Siemens. This is a stable substrate. The only problem is that the pitch is normally large. We are also working on a technology based on photo-assisted etching. With that, you can have diameters of the TSVs in the range of down to 2 or even 1 micron. So, you can have a thick wafer or substrate with a low pitch. You can metalize it with alloys.

SMD: What are the applications for this?
Ramm: This would be an alternative for a thin interposer. You can use it for anything. It could be a memory stack on a processor and so on. In this project with Siemens, the application is for MEMS-ASIC integration. A typical example is that you can have two ASICs. You have an RF receiver/transmitter and you have a controller. You can integrate them with an interposer and add a MEMS device. For example, you could end up with a pressure sensor.

SMD: What’s the status of this device?
Ramm: This is still in research and development.

SMD: What are the advantages of a thick interposer?
Ramm: The handling is easier, so it’s more cost-effective in fabrication. It’s like a standard wafer. For example, the thickness is 700 microns. You can, of course, make it thinner, as long as it stable.

SMD: Fraunhofer is part of the Best-Reliable Intelligent Ambient Nanosensor Systems, or e-BRAINS, project. What is that project?
Ramm: The e-BRAINS project is involved in the research and development for the integration of heterogeneous systems and ambient assisted living. So in fact, it’s about the combination of sensors and MEMS with ICs for the application of ambient assisted living. A part of this is 3D integration. There are not so many vertical interconnects in these types of systems. In this case, it can be hundreds or thousands of interconnects. The direction for us is in miniaturization. The task is to use 3D when it is needed. Typically, it’s not performance driven. It’s form- factor driven.

SMD: When did the e-BRAINS project start? Who are the partners in the group?
Ramm: It started in September 2010. It will end in September 2013. We have around 20 partners. The partners are Siemens, Infineon, SensoNor, Sorin, IQE, 3D Plus, DMCE, Vermon, Magna Diagnostics, EESY-ID, Fraunhofer, SINTEF, IMEC, CEA, EPFL, Tyndall, Technische Universität Chemnitz, ITE and TU Graz.

SMD: What is the goal?
Ramm: The demonstration vehicles will be in the areas of medical applications, such as a smart biosensor grain. In effect, it’s a DNA sensor. We are also looking at infrared imagers and active medical implants like pacemakers. For Siemens, we are looking at air quality systems for the applications in the home. It’s also interesting for applications in cars. The fifth application is smart ultrasound imaging for medical applications.

SMD: The e-BRAINS project is just a part of your 3D R&D activities, right?
Ramm: This is a part of our 3D activities. But looking into the future, I see it as a more and more important part.

SMD: What are some of the challenges?
Ramm: In the case of a pacemaker, it’s really miniaturization. The 3D in this case is without TSVs. It’s not needed. We just use a cost-effective method. It’s 3D, but it’s done it in a way where are you stacking chips and doing a sidewall or maybe a through-polymer vertical metallization. So you don’t need TSVs because you don’t have the integration density. The driving factor is the form factor.

SMD: How do you make a DNA sensor?
Ramm: A DNA sensor is based on a TMR principal. The integration with a corresponding ASIC is done via 3D integration. In this case, it’s done with a couple of through-silicon vias. If it is cost-effective, it could be a success.

SMD: Where is the industry at with 3D?
Ramm: Concerning the variety of concepts and bonding techniques, we are in very good shape. We have a variety of choices. But you have to make the right choice. You have to choose a reliable 3D integration technology to fulfill the performance and pitch requirements. You need robust processes working in fabrication, not on a power point presentation. There are reliability requirements. They are completely different from product to product. For automotive, you have very high requirements on reliability.

SMD: Where is 3D heading?
Ramm: The answer is that it’s not heading in one direction. You will have a variety of technologies. You will also have a variety of applications. The thing that has to happen soon is that 3D devices must go into fabrication and not face too many problems. This has to be demonstrated. It must go into a product.

SMD: There are still so many challenges, right?
Ramm: The challenge is handling the thin silicon and producing TSVs without any cracks. In the supply chain, not everything can be done by one manufacturer anymore. The reliability must be demonstrated. From my point of view, this is the key problem and challenge—3D must demonstrate the ability to be reliable. It’s one thing to show nice pictures of TSVs, but we have to show it works.



Leave a Reply