NIST’s Engineering Physics Division chief opens up on future directions for metrology.
David Seiler, chief of the Engineering Physics Division within the Physical Measurement Laboratory at the National Institute of Standards and Technology (NIST), sat down with Semiconductor Engineering to discuss the current and future directions of metrology. NIST, a physical science laboratory, is part of the U.S. Department of Commerce. What follows are excerpts of that conversation.
SE: What is metrology?
Seiler: Metrology is the science of measurements. So you are taking measurements–for example, what’s the thickness of a gate oxide or something like that. You want to be able to have reproducibility and precision in your measurements. That’s important because you don’t want to have to throw away wafers.
SE: NIST develops an assortment of current and future measurement technologies, right?
Seiler: NIST helps push the accuracy of the measurements, which would be needed for future devices. NIST has also produced and sold standard reference materials for years on various types of standards. If there is a piece of equipment, and you need to test it out, how do you know that you’re measuring things correctly? So a lot of times, you use standards. You can get them from NIST. And then the companies themselves would have their own way to check whether the equipment is operating with proper precision or accuracy.
SE: Metrology has been used in the IC industry for decades. Some say the demand for metrology began to increase at 45nm, when chipmakers introduced high-k/metal-gate structures in logic. Was that the case?
Seiler: This was about the right time. If you talk about high-k, you had the introduction of different and new materials. You wanted to know the composition.
SE: What else?
Seiler: In 2000, the cover page of our book for our metrology conference showed an atom probe. With that instrument, you can see the individual atoms in a certain device. So even 15 years ago or so, people wanted to know where individual atoms were at. Of course, they also wanted to know what kind of atoms they were. This was important for identifying defects and impurities.
SE: This is atom probe tomography, right? How does that work?
Seiler: With the atom probe, a high voltage is used to eject atoms from the surface. This is so images of the individual atoms on the surface can be reconstructed.
SE: Today, many of the current metrology tools can image structures in two dimensions. But the recent shift to new 3D-like devices, such as 3D NAND and finFETs, prompted the need for so-called 3D metrology. For this, the industry needs metrology tools that can measure in three dimensions, right?
Seiler: Now, you are talking about building in the third dimension. For 3D, you are talking about nano-metrology. You require more complicated imaging methods. Again, you want to know where the defects are. You also want to identify where the dopant atoms are.
SE: What are some of the challenges?
Seiler: You are seeing complex systems with new materials and very small dimensions. You have different kinds of structures. They need to have measurement advances to help enable the future of electronics. And, of course, if we talk about CMOS, there have been new materials and structures introduced. 10nm structures are being talked about and planned for production. And certainly, research and development are going on at 7nm and probably 5nm. People are even talking about 3nm. So we’re down to counting atoms here.
SE: What are some of the problems here?
Seiler: Certainly, the interconnects are going to be a challenge. If we talk about the transistor, we’re talking about so few atoms. One can talk about variability, yields or reliability. If you have just one or two defects, you are in hot water.
SE: What about the future?
Seiler: In the future, say 15 years from now, we are talking about totally different devices that will be based on different techniques. A whole new set of metrology and measurement techniques will need to be used.
SE: Today, we have a multitude of metrology tool technologies. How do you classify them?
Seiler: There are physical methods. There are optical methods. There are electrical methods. If you talk about optical methods like scatterometry or ellipsometry, these are contactless techniques. And that’s going to have some advantages. The problem with the optical techniques is that the resolution is going to be limited because of the wavelength of light. And if you have a very small transistor, the challenge is to characterize it optically.
SE: Will optical metrology run out of steam one day?
Seiler: You will see more advances in pushing the sensitivity and resolution of techniques like scatterometry. The state of the art today might be 25nm or maybe 30nm. But we need to go much lower. If we are talking about 10nm, and trying to get to 5nm, we need to make some breakthroughs in how we can push things like scatterometry at higher resolutions. This is so we can detect defects that are very small and are basically undetectable by any known techniques now.
SE: Any other thoughts?
Seiler: We are going to run into new gaps and problems. We do need new techniques because we are running up against complicated structures we’ve never measured before. They are going to be very small. But you will still need some standard workhorse tools. They are going to be needed and used. They can be refined.
SE: Is there a single metrology tool that can handle all needs for current and future devices?
Seiler: There are a lot of different ways to go. With the complex nature of these chips and transistors you need to have more than one technique. In other words, it’s a hybrid of techniques to produce the best accuracy, precision and understanding of what you are measuring.
SE: There is a lot of talk about hybrid metrology. Can you explain that?
Seiler: For example, you use small-angle X-ray scattering. Then you compare it with some advanced scanning electron microscope data, and then you combine the two measurements. With this hybrid metrology technique, you can come to a better picture of the actual structure you are trying to measure. Hybrid metrology is the wave of the future.
SE: For years, NIST and others have been working on small-angle X-ray scattering, or CD-SAXS. It’s still in R&D. What is CD-SAXS?
Seiler: The structures are so small now that you need something like CD-SAXS to be able to give you an image. If you want to measure the pitch properly, it’s a way to get a three-dimensional image. From there, you can extract the dimensions so that you are able to reproduce the structures. CD-SAXs has demonstrated a tenth of a nanometer accuracy in measuring industrial relevant 32nm pitch structures. That’s why people are excited.
SE: CD-SAX is limited due to the power source. NIST has installed a new and more powerful source based on a liquid-metal jet source technology. What’s the status?
Seiler: It’s very promising.
SE: Will CD-SAXS replace optical metrology?
Seiler: I don’t think it can do everything. You will still need things like ellipsometry. You will still need optical techniques. Again, it depends on what you are measuring. Each of the techniques has its own advantages. But the structures and compositions are complex. So you need a variety of tools.
SE: NIST is also developing something called electron spin resonance (ESR), right?
Seiler: Electron spin resonance is a technique that has been around for a long time. It’s very useful for identifying the defects and atoms in liquids and solids. It is also useful for the symmetry of where these defects are. The difficulty is that the normal technique requires a big electromagnet. It’s bigger than we are. And it’s heavy.
SE: Is NIST developing a new version of ESR?
Seiler: This is a technique that is revolutionary. You use only a small magnet. And you use a very small tip. In fact, we have a new tool coming in at NIST. You can use an AFM tip. The goal is to get down to a single defect. This is not only to detect it, but to identify it. We’ve advanced the technique and sensitivity by a factor of 20,000. The goal is to go beyond that by another factor of 20,000 or maybe 50,000.
SE: The industry is also working on another advanced technology called neutron scattering, right?
Seiler: You can use cold neutrons to do depth profiling or diffraction. Neutron scattering would be useful to help unravel complex structures. NIST has a cold neutron source. Of course, with neutrons, you must have a source to produce the neutrons. Here at NIST, that means we have a reactor—a nuclear reactor. With this, you can slow the neutrons down. And then, because they act like waves, you can do neutron diffraction, scattering, depth profiling and things like that. But the disadvantage is that you need a neutron source. If you want enough neutrons for measurement sensitivity, you need to have that at a reactor site. You, of course, need protection. So neutron scattering is going to be rather limited in applications. But it could be used in R&D.
SE: Several R&D organizations have synchrotron or particle accelerator facilities. They have giant beamlines, which can be used for X-ray metrology and other apps. Any thoughts?
Seiler: Neutron scattering and synchrotrons all have their usefulness. They are qualified to measure aspects of a material structure, property or composition. They all have their places.
SE: Earlier, we talked about atom probe tomography. This is generating a buzz, right?
Seiler: Atom probe tomography is an important advancement. You can use it to extract a structure atom by atom. And you are able to model the data. So you get an accurate picture of what you are doing, where the atoms are, and how to identify them. That’s critical for atomic and nanoscale chemical and structural analysis.
SE: What else?
Seiler: NIST has two of them. Imec is working on it. Every major institute may have one. In one of our last conferences, we brought all of the atom probe people together. They were asking themselves questions. For example, wouldn’t standards be great to have? So then, you can test one’s measurements, understand how they are extracting the data, and model the physical structure. You can compare that across the world. That’s the frontier and the gap with atom probe. How realistic and accurate is the picture of the atom probe? People are working on that and there are going be improvements.
SE: Is this the ultimate tool that can measure atoms?
Seiler: This is just one. There is another project at NIST. We can turn every TEM into a high-speed camera for atoms. That means you couple the spatial atomic resolution with high-temporal, sub-picosecond resolution. It can be incorporated into a traditional TEM microscope. It’s like an advanced stroboscope or advanced camera. So, you can record magnetic spin waves, for example, with time resolution. The stroboscopic transmission electron microscope is a good one to keep your eye on.