Interview with Selim Ünlü

Featured Engineer

Selim Ünlü

Selim Ünlü - Professor with Multi-Disciplinary Research Program; Boston University

How did you originally get into electrical engineering and electronics?

I’ve always said that what you do in life is defined mostly by coincidence. I’ve always wanted to be an electrical engineer without even knowing what that was when I was a kid. Then again, I also wanted to be an aviation engineer when I was in elementary school. I grew up in Turkey, and went to a very special high school training people to become scientists. And when I graduated, the most popular professions in Turkey were either electrical/electronics engineers or medical doctors. So, about a third of my classmates became electrical engineers, more than a third become medical doctors, and about a fifth took other engineering jobs. I chose electrical engineering. I enjoyed the subject material. For two years, as a junior and senior, I worked at a factory building electronic circuits for handheld radios, transmitters, and receivers. It was very satisfying to build something that can actually be put to use. When I finished my undergraduate degree, I applied to graduate programs, and I got admitted to the Masters and PhD program at the University of Illinois at Urbana-Champaign. At the time, my background was in electronic circuits, electromagnetics, and microwaves. The devices we were working on were working at high frequencies. A professor – my MS and PhD advisor – recruited me to work on semiconductor devices based on having listed microwave devices on my application. This was a completely new field of study for me. I accepted the offer and worked in his lab for 6 years. I worked on a variety of different semiconductor devices from bipolar transistors to diodes. After my MS and a couple of years into my PhD, I started to work on optoelectronic devices—mostly photodetectors. While working on photodetectors, I got interested in enhancing photodetectors by using an interference filter. One of the most significant coincidences that shaped my career is having Katsumi Kishino – a professor from Tokyo – visiting our lab on his sabbatical. Working with him, we developed resonant cavity enhanced (RCE) photodetectors. This was highly beneficial to me, because I was finally able to use the experience from my undergraduate degree in applications on semiconductor devices. RCE concept was an interesting innovation and applied to both detectors and light emitting devices. Consequently, RCE attracted significant interest from researchers worldwide and became a very large field of research over the next decade. Today, there is even a Wikipedia listing.

What are your major research areas now?

I work in a variety of application areas. I feel very fortunate for being in Engineering at Boston University . We have a very collaborative, highly interdisciplinary and stimulating work environment. My laboratories are located in the Photonics Center that houses many research groups from engineering and sciences. We also have very strong ties to BU Medical School.

One of the focus areas is high resolution optical microscopy with an emphasis on semiconductor circuit imaging and fault isolation. Most recently, I branched into doing conformational measurements of biomolecules, biosensing, and neural stimulation. Our group’s name is Optical Characterization and Nanophotonics. Even such a general name falls short to describe the breadth of multi-disciplinary research we are involved in.

Can you give me a background on each one of those applications?

Let me start with the recent work on high resolution imaging of semiconductor circuits and devices using a solid immersion lens. The trend in the semiconductor industry is to reduce the dimensions and increase the number of transistors in a chip while increasing speed and reducing power. When you are making these finer structures, you have to use very short wavelength ultraviolet light in photolithography. Once the fabrication is completed, the top surface of the chip has many layers of metallic – opaque – structures and you no longer can look through the top surface to examine the device layer, which is the semiconductor layer, and see what devices are functioning as designed, or what is not working well. If there is a fault, you cannot identify it by looking from the top surface. Instead, most of the technologies for isolating faults in semiconductor chips are through backside imaging, that is, through the silicon substrate. For inspection, you can no longer use the very short wavelengths in the UV range that are utilized during the manufacturing process. Instead you have to use infrared wavelengths because silicon is not transparent in the visible. Basically, you have to use five times longer wavelength to measure the things you have written with short wavelengths. Thus, your imaging resolution is no longer compatible with with the nanoscale structures on the electronic circuit . We have developed a very simple technique using a high index silicon microlens that transforms the back side of the substrate, which is normally flat, into a curved surface, which allows tight focusing of light. That way we can achieve resolutions that are compatible with the feature sizes that engineers are trying to image.

We pioneered this technology for circuit and device design about ten years ago, and have been working on improving it ever since. Recently we received substantial funding the Intelligence Advanced Research Projects Activity (IARPA) totaling $5.3 million as an interdisciplinary research team from Boston University— along with an industrial partner, DCG Systems, Inc. of California. Our team will spend the next four years applying novel imaging approaches to pinpoint and resolve defects on next generation ICs. These tools are very important because the semiconductor industry is very large and extremely time sensitive. In the development cycle of new electronics chips, minutes are measured in many thousands of dollars. We have been working to develop the technology which enables rapid and accurate identification of faults. You can read more about the project here.

Are you building a prototype?

We will work with the state-of-the art chips from semiconductor manufacturers. We will work with them indirectly. Semiconductor manufacturing industry is very conservative and protective of know-how. Manufacturers will not reveal the specific structure of the metal layers, materials in the chips, etc. The layouts of the chips are also closely guarded secrets. Our development partner in semiconductor IC diagnostics has many customers among the chip manufacturers. We rely on their expertise and connections to ensure the relevance of our technology development.

In terms of resolution, how small of structures are you able to resolve with the technology you’ve been developing?

Right now, there are very specific requirements for our current research program. The first milestone is resolving fault isolation at 150 nanometers. The semiconductor technology does not arbitrarily reduce the size of the devices continuously, but rather follows a roadmap with discrete improvements termed “nodes”. When we started to work on solid immersion lens technology, standard semiconductor IC technology node was transitioning from 0.13 microns to 90 nanometers. The following technology nodes were 65nm,45nm, and 32nm. The next generation technology that is going into production is 22 nanometers. Our research program is for four years, and the goals are for 22 nanometer as well as 11 nanometer nodes. The resolution requirements, for us, are 150nm for the first phase and 80nm nanometers for the next generation. It will be very challenging to isolate faults with that precision on the chip, but we are confident with a strong team and innovative technology.

What type of a sensor do you use as the image sensor?

There are different modalities of imaging semiconductor chips. Obviously, you want to monitor how well the electronic circuit is functioning. One of the specific analysis techniques our industry partner has developed is so-called laser voltage probing, or laser voltage imaging (LVP, LVI). In this modality, you have the chip working as it would normally operate with a clock driving the chip at a certain frequency and you have also data flowing through the chip at a different repetition rate. Thus, the transistors are switching at these frequencies. If they are functioning properly, you know that some of them have to be switching at the data frequency and some of them should be switching at the clock frequency. And then you shine a tiny spot of light on the circuit, and then you scan that light across the chip. So, at a given time, light is reflecting from a small area, and the resolution is determined by how tightly you can focus that light. When the transistors switch, there is a slight change in the optical properties due to variations in the carrier concentration. Observing the reflected light, you can look at the changes at those particular frequencies, either clock or data, and if you see a signal modulated at the clock frequency, or data frequency, for example, you can see the transistor switching at that speed. Most of the signal is not modulated at that frequency, you know it’s a very small change, and that is less than one part in one thousand. Extracting that information is a detection problem, but we do use somewhat ordinary photo detectors for this. The innovations are mostly in the optical systems. By scanning the focused light across the chip, we can form an image. Alternatively, by focusing on a single point and we can obtain the switching characteristics as a function of time. These are used to diagnose the circuits so you know that these transistors should be switching at this frequency, and if they aren’t switching as expected, you know that there is a fault.

Another method is called photo emission microscopy. When transistors switch, they emit a tiny bit of light. You can collect that light, and from determining the timing of the emission and its location, how much it is, and what the timing of that signal is, you can diagnose the circuit. Of course, that light is in very small amounts, so you need to built very sensitive detectors, very high collection efficiency optics, and high resolution optics to operate the transistor properly. You also have to work very closely with the manufacturers because, as the transistor parameters change, the amount of light that is emitted and the wavelength of light is also varying. It’s a moving target , thus a very challenging project.

When we started working on this project, we were the first people who actually applied solid immersion techniques to semiconductor circuit imaging. Now, all of the equipment manufacturers are using similar technology. Almost invariably, everybody has to use this solid immersion lens technology to image chips. It’s quite satisfying to see that our hard work is making an impact.

Can you tell us about your other project on Disease Diagnostics?

The current industry techniques for diagnostics are based on a sandwich assays. Most of the detection techniques rely on fluorescent labels for transduction. A common way of detecting the presence of a disease is to look for the protein biomarkers. Every pathogen has specific antigens and the immune system will develop antibodies against the infection. Typical assays use antibodies as capture agents to detect the presence of corresponding antigen proteins in a sample such as blood, urine or saliva. Capturing these proteins does not provide a sufficient signal. A secondary antibody that recognizes the same antigen used in a so-called sandwich assay, where the secondary antibody has either an enzyme or some fluorescent molecules linked to it. Conqequently, either the solution color changes, or it fluoresces when excited by another light source. That’s a transduction mechanism to make the presence of small amounts of target proteins captured visible. This kind of technology is typically called Enzyme-linked immunosorbent assay (ELISA). It’s a very common technology, but it does require a secondary antibody, and you have to refrigerate those, have a trained technician, etc. The problem is, you cannot have those types of tests in a doctor’s office. Today, you go to the doctor for an examination, the doctor takes your blood, and you have to come back three days later for the results. We want to build systems where people can, around the world, be tested right on the spot – at the Point of Care.

Are you at the point where you will start working on prototypes?

Label-free sensing is a recent trend in disease diagnostics. We all want to get rid of the secondary antibodies, the additional reagents and the labels. But it’s a challenging project, because you have to do the detection, not only sensitively, but also specifically. There are a lot of steps that need to be taken to prove these things. One of the technologies we have developed recently is, using our knowledge of optics and layered surfaces, based on our expertise in semiconductors, we were able to build the technology for a very inexpensive and compact system. We can detect individual bio-particles. This is something that people weren’t able to do prior to our work. Now, on a very simple surface, we can detect the presence of individual nanoparticles and viruses. We would like to apply this to diagnostics, which would allow multiple diagnoses on a single chip, with much more timely results. You can check out the project here.