Léo RamosA Brazilian little known in his native country might be at the leading edge of a new kind of technology for computer memory and other electronic devices. Carlos Paz de Araújo was born in Natal, in the Brazilian state of Rio Grande do Norte, and has lived in the United States for 42 years. He is a professor at the University of Colorado in Colorado Springs, and a partner in the company Symetrix, which has gained worldwide importance in recent years for developing and licensing ferroelectric random access memory (FeRAM), used in more than 2 billion devices ranging from smartphones to electronic processors, from appliances to automobiles, DVD players and the newer generation of magnetic cards with chips. This is non-volatile memory, which does not lose information when the electrical current is turned off, and can be reused billions of times, contrary to the flash memory used in pen drives, which is slower and cannot be reused more than 100,000 times. Volatile memory is used in computers with hard drives and runs with the help of software when the equipment is connected.
The technology developed by Araújo has been licensed to companies such as Panasonic, Siemens, Delphi, Hughes, Sony, Sharp and many others. Although he has been abroad for many years, he still has the typical accent of his home state and does not seem like the type of executive who flies constantly between various Asian countries to exhibit his multi-million dollar technological innovations. He often visits Japan, where he is a visiting professor at Kochi University of Technology, and is also a consultant to Fudan University in Shanghai, China. Araújo is first and foremost a scientist. On his own, he spent 7 years studying resistive memory—which is expected to replace ferroelectric memory and could find universal use in pen drives, cameras, laptops, desktops, and low-cost mobile phones. This is different from what we see today, when there are several types of memory, one for each application.
The result of Araújo’s research is CeRAM (correlated electron RAM), which should eliminate computer hard drives and speed up processing. Developed at Symetrix, the new technology is already being tested by large companies in the industry. The company earns $4 million per year in royalties alone, and licenses no technology for less than $20 million. For all these contributions, de Araújo was the first Brazilian to win a prize for technological innovation from the Institute of Electrical and Electronics Engineers (IEEE), based in the United States, called the Daniel Noble prize and regarded as the Nobel prize in electronics engineering. This association has more over 300,000 members in 160 countries.
Araújo’s academic life began at the University of Notre Dame in South Bend, Indiana, where he studied electrical engineering, then pursued master’s and PhD degrees in the same area as well as obtaining degrees in philosophy and theology. At 60, married to Maureen Paz de Araujo, an American and also a professor at the University of Colorado, and with three sons, he seldom visits Brazil. During his latest trip, he consented to this interview with FAPESP during the 12th Meeting of the Brazilian Society for Materials Research, held in Campos do Jordão, São Paulo.
Was your journey from Natal, Brazil to the University of Notre Dame like?
I first went to Chicago as an exchange student at age 17. I was supposed to stay three months, but stayed six and then returned to Natal. The family that hosted me invited me to come back so I was an exchange student again at age 18. At the time, I was in the second year at a scientific high school. When I went back to Brazil, I began the third year and then they called me to come back, so I finished high school in the United States. The American family had seven children, but none with good enough grades to get into Notre Dame, known as the Catholic Harvard, and they wanted someone to go, representing the family. I got in and stayed there for 10 years. After my PhD I was invited to be a professor there, but I only stayed six months because the University of Colorado made me a good offer. I’ve been there since 1982.
| Age: 60 |
| Specialty: |
| Electronics engineering |
| Education: |
| Undergraduate degree in electrical engineering, master’s and PhD from Notre Dame University in the United States |
| Institutions: |
| University of Colorado, Colorado Springs and Symetrix Corporation |
| Scientific production: |
| 310 articles, 203 US patents and 321 patents in other countries |
When you moved to the University of Colorado, were you already working with microelectronics and semiconductors?
Yes, I worked with semiconductors, the physics of electronic devices, which is an area in electrical engineering, not physics, despite the name. A year later I began working on ferroelectricity.
Why? Incidentally, what is ferroelectricity?
I went to the university and saw that it was an option, because at the time very little was known about ferroelectricity and no one spoke of ferroelectric memory. The prefix “ferro” (iron) is misleading, it was used as an analogy with ferromagnetism. This happens when a magnet is placed near a material and the material becomes magnetic. This also happens in dielectrics, which are insulating materials that have the ability, with an electric charge, to change the position of an atom in a material. So, the ferroelectric memory changes the position of the atom and this leads to hysteresis [a condition in which a material stores information without the initial electrical stimulus that created it]. They used the name “ferroelectric” because of ferromagnetism, but ferroelectric memory has no iron.
Early on, in 1984, you established Ramtron with two other partners. What did it do?
Since the beginning of my career, my research has been with the purpose of creating new companies. Ramtron was a high-risk start-up company that carried out research to make ferroelectric memory possible. I headed up the research aspect and had staff to carry out design and financing. It was successful, became very famous, and was sold last year.
Was there already a positive environment at the University of Colorado for start-ups and innovation?
No. Actually, there was an environment for start-ups in my head, but not yet at the university. The University of Colorado was less sophisticated than Notre Dame, so I had already seen that obtaining funding from the National Science Foundation [NSF] would be very difficult. So when I saw the opportunity inherent in ferroelectricity, I started with a business plan instead of a research proposal. This created a way for me to work. All I want is to combine a deep level of research into a new phenomenon that few understand, with the possibility of immediate application.
And why did you establish Symetrix in 1986?
For the same reason. At Ramtron the founders, myself included, raised a lot of money and we had some operational problems, and then I got fed up and created my own company. There is a huge difference in technology, though. Later, Ramtron even bought a technology license from Symetrix, but isn’t using it. Symetrix ended up competing with Ramtron. Today there are two families of ferroelectric memory: theirs and ours. Theirs uses a lead alloy and our uses tantalum, bismuth and strontium. Our alloy is 10 times faster, uses less power and is cheaper to make. But, since ferroelectric materials are so desirable, some companies are following Ramtron.
What is the advantage of ferroelectric materials?
Ferroelectric memory stores information in nanoseconds. Flash memory, which is used in pen drives, uses a type of semiconductor memory that takes about 10 microseconds per bit to store information, so it is still very slow, and can’t be employed universally. Plus it self-destructs. The more you use it, the more it wears out and then it is no good.
In 1984, 1986, what kinds of electronic equipment used this type of memory?
Ramtron’s first major sale was worth $40 million, to a power company in Europe. They made meters that measure power consumption for a smart grid system. So when power fails, all the measurements are stored in memory in nanoseconds. The memory is fast, but does not require much power.
Going back in time, when you founded Ramtron and Symetrix, were there problems within the university?
Not with Ramtron, because it was separate, having been established with Australian venture capital. I served as a technology consultant. But when I established Symetrix, all of the initial capital stock of $4 million came from the U.S. government, specifically the National Science Foundation, through the Small Business Innovation Research [SBIR] Program.
The SBIR, incidentally, was the model for FAPESP’s Pipe (Innovative Research in Small Businesses Program). So did the SBIR have a very important role in the creation of Symetrix?
Yes. But when we reached revenue levels of $4 million per year, we no longer participated in the program. I had already obtained funding from Japanese companies. Later, I sold a license for military use to the American company Hughes Aircraft.
But you stayed on at the university, studying ferroelectric memory and doing research in your own company.
Actually, the company became a research center that funded the university. It was a partnership. We positioned ourselves as an independent entity that signed contracts with firms, sold licenses and paid the university as if it were a budget expense. At least $8 million was paid to the university, but the company’s revenues were much higher, from licensing and contract research alone.
What companies did Symetrix work with?
Companies that worked with me on ferroelectric memory were Panasonic, Sony, NEC, Siemens, Sharp, STMicro, and National Semiconductors. In all there were about 20 companies.
What devices use these advances?
Here in Brazil, this memory can be found in all printer cartridges, in the PlayStation and in almost all computer DVD players. Mobile phones had them in the beginning, and now they are coming back. It is starting to become something that every company wants. For example, the new smartphones and cameras come with NFC, or Near Field Communication, a system that facilitates communication and data exchange between two devices, with the ability to also turn on appliances inside a house. And the NFC systems used by Panasonic, Sony and Samsung use the memory that I developed.
So what Symetrix receives are license fees from patents and royalties. Everything was patented. How many patents do you have today?
More than 200 patents in the United States and 350, more or less, in other countries. And none is with the university. I paid for all of the patents. I do not even pay royalties to the university. I separated everything.
Were things together initially?
No. I paid the university as if I were the government paying myself. The university had no right to any of my patents.
But was it easy to do this, or was it a complicated deal?
At first it was easy because there was no money involved yet. When you start making money, everyone wants to undo something, but by then it was too late. The easy way was to use the theory of compensation. If the university were to pay for the patent, it would not be able to develop the technology because it would be very expensive. So the compensation took the form of my giving the university a 5% stake in the company, which is better than any royalty. And they signed this agreement without a problem. Then, later, the university had financial problems with its start-up incubation program and they needed to cover the deficit. To obtain the funds, they wanted to sell their share in the company and they sold for a very low price. But it was only part of the university that actually completed the sale; the other part didn’t know about it, and by the time they realized that they were no longer owners, it was too late. They were upset with me because Symetrix bought back the 5% that had belonged to the university.
The rule then is to register the patent immediately, and then publish the research?
Yes, I publish afterwards. Sometimes I publish a piece before the patent is approved. One has to make the decision carefully. For example, there is a new type of memory, resistive memory—known as ReRAM—which could replace ferroelectric memory. [He created correlated electron RAM, known as CeRAM] Much more care has been taken in patent publication. Everyone made this type of memory in a certain way, and I figured out what was not working. We deduced that the device’s paradigm was wrong. So, we did the science, patented it, and began to publish. Now it is new both in science and in technology.
Ferroelectric memory and CeRAM are non-volatile, which means that they do not erase information even without power. How do they work in an integrated circuit?
An integrated circuit operates using binary logic, with zeros and ones. All transistors and resistors are what we call the bottom. The memory is on top. As the emphasis is now on mobile devices, everything must have non-volatile memory.
How was CeRAM developed? Did you stop teaching?
I continued lecturing and didn’t bring up business matters because I knew that this memory would be different from the start. I sought out all the related literature to understand what was happening. I bought about 2,000 books and read around 7,500 scientific articles. A total of 72,000 pages. Then, after I understood everything, I began to publish. In 2011 we published an article on CeRAM memory that was the cover article of the Journal of Applied Physics.
Will CeRAM become universal memory?
It is a high-density, non-volatile memory. It also has a huge number of gigabytes of storage, great processing speed, low power, low cost and long life. It would eliminate the need for DRAM, SRAM [used in desktops and laptops], flash memory (pen drives), and computer hard disks. CeRAM would be used for all of these.
And how close to this universal memory are we? Is demand high?
It is a matter of investment. Everyone in the industry is looking for this next memory. And resistive memory is the first candidate. But there are two types of resistive memory: one that has been around for 12 years and can’t be improved, and CeRAM. That’s why we changed the name, instead of being called ReRAM, we called it CeRAM because the effect that makes it memory is the effect that in physics is called strongly correlated electrons. We treat the electron as a wave and not as a particle because we are in the quantum state. We made the material with nickel, a transition metal with a property—understood since 1937—that no one had figured out how to use yet. Nickel oxide is able to manipulate the quantity of electrons passing through the material. In fact, it is a new transistor that can withstand up to 400°C without loss of memory.
How was this discovered?
I did not discover this phenomenon, what I did was control it, first on the chemical level. When the material was stable throughout, I tried to understand the phenomenon in order to write the equations and explain the memory states and conduction, through two fields, one called mesoscopic physics [the miniaturization of macroscopic phenomena] and the other quantum physics.
What kind of investment is needed for a production line?
There are several types of factories with different levels of integrated circuit printing technology. For example, today, a company that makes microcontrollers and wants to add memory needs 130 nanometer printing technology to lay down the lines of an integrated circuit. By comparison, our technology is at the 10 nanometer level, 20 times finer than the wavelength of light. And the machines that print on the chip, to mark the location of the transistor, are large and sophisticated. Just one printing machine costs at least $40 to $50 million. And you can’t have just one. At this level, a factory costs at least $4.5 billion. The state of New York, for example, announced in September that it plans to build three semiconductor plants and is investing $45 billion.
And why do they want to do this there??
It has become a center for high technology, for nanotechnology. There is no federal government influence, just the state government and several companies that are financing it. The state was very clever because it took money from the lottery and established a high-level nanotechnology center with about $500 million and this attracted companies.
But is your CeRAM already at the point of being transferred to the research and development (R&D) department of a company?
It is quite an interesting process, because it’s like a political campaign and the candidate is the memory. Generally, the best thing in the world is to have a bad example for comparison, because anything better looks good. So the bad example is that everyone did research on resistive memory using the same material, the same thing. Everyone did it, especially Samsung, which is the world’s leader in memory. Then you appear in this campaign [via press releases, interviews and conferences] saying “That one is worthless, ours is what works and here it is.” Then the phone calls and emails flood in. We already have two large companies working with us, but I can’t tell you the names.
What is this process like?
They have R&D. But a company like Samsung spends $4 billion per year on R&D. They go crazy when someone like me appears to solve the problem. They need to buy a license, but first they want to make sure we are telling the truth. They agree and we start a period in which they can ask anything and we respond and even prepare a chip, so that they can manufacture with our technology. This also depends on the company, some buy the license right away, but others want to test the technology, get a feel for it.
How much does the license for this kind of material cost?
It costs at least $20 million plus 3% of sales. So, if they sell $10 billion, 3% goes to Symetrix.
What does Symetrix take in and how many researchers does it employ?
Today it earns $4 million just in royalties, not counting licenses and projects. There are currently 10 employees. We earn a lot of money transferring technology. When we run out of business, we create. When it is ready, everyone wants to know about it and learn how it works. Generally, the scientific community does not even know what we’re talking about. For example, when we introduced ferroelectric memory, hardly anyone knew what it was and even today few people understand it.
Is there is no research on this at the universities, even in the US?
We were the first, then a lot of people copied us. But university laboratories do not have the discipline to make a memory that can become a product. And they often waste a lot of time publishing, in a slow process that academia itself promotes. This is seen most clearly in the memory business, because if the research does not turn into a product, it has no value.
But this is not the purpose of universities.
Well, it should be in engineering. But they don’t do this because professors are always looking for funding. Our company partners with a large company and we invite their engineers to live in Colorado and work with us. They then spend three or four years at the company, learning. There are no development costs, only research. When it passes into the hands of the company that is licensing it, they are already developing a test chip. Then, when all the parameters are understood, we can say: “now it can go to the assembly line.” The engineers stay at Symetrix to transfer the technology to their companies.
You tried to set up a ferroelectric memory factory in Brazil in 2008. Why didn’t it work out?
This issue is very complicated. Symetrix’s strong point is inventing. It’s not every day that someone stands up and says: “I’m going to invent the next type of memory.” The semiconductor industry is worth around $300 billion per year and a large part of it is memory. Twenty years ago, in the United States, companies without factories that design chips (fabless) were the target of investments. That was a success. Ninety-nine percent of chips are designed in the U.S., but made in China. This started in Taiwan with a firm called TSMC, which has $77 billion invested in factories. The governments of these places invested to provide employment, then the price of cheap labor was a partnership between the United States and Taiwan. Then China began to copy this and now has three huge factories, but its investment is nowhere near that of Taiwan. The idea comes from the United States and the factory is in China. Then the Americans sell to the rest of the world—this is what we call a butterfly firm. One wing is manufacturing, the middle is control over the design, and the other side is marketing. So, about eight or nine years ago they did a study in Brazil, at the beginning of Lula’s term as President, on how the country could manufacture semiconductors, and they came to the conclusion that the government should invest $5 million in the design sector and establish some design companies here. The business model would be the same as in the United States, to design here and manufacture in China. However, China has become increasingly complicated because they have also learned to design over time. Their idea is to dominate, but they haven’t been able to dominate the industry that much. Innovation is creating the design, or the type of chip, which goes into a new iPhone, and that is still only done in the United States. In Brazil, the federal government also decided to invest in a semiconductor factory in Rio Grande do Sul, Ceitec, but they used equipment donated by Motorola that was very out-of-date. They never completed the factory. It is an example of both academic and government inefficiency in Brazil in this area. When I came here with ferroelectric memory, I went to the Brazilian Development Bank (BNDES). I was interested in establishing a branch of Symetrix in Brazil, but I didn’t want to invest my money because it was too great a risk. I would transfer the technology to someone else. We would start with a wafer [a thin, circular semiconducting piece of material] purchased for $500 in Taiwan and put the ferroelectric memory on it, in a Brazilian plant that would cost no more than $50 million, not $4 billion. The idea was to export, for the most part. If we had done that, today Brazil would be exporting microcontrollers with ferroelectric memory. The microcontroller is cheap and can be used in various types of integrated circuits, in cars, stoves and refrigerators, for example. We would be supplying the market and have a company with initial revenues of $150 million per year, with an investment of $50 million.
And the ferroelectric memory cards?
Another idea was to use ferroelectric memory here to produce a smart card, already used in Japan to pay subway and bus fares, which could also serve as an identity card and credit card at the same time. This advanced chip is a microcontroller with a radio station, with radio frequency identification (RFID) memory. If we introduced that card in Brazil, the factory would be up to full production within a year. This is already so evolved that you can’t buy new computers [desktops, laptops] in Japan that don’t have a card reader. You just put the card on top, get on the Internet and see everything you bought, what you did, or recharge it with more credits. Thus, a Brazilian company would have something more advanced and have an advantage over other businesses.
Returning to the Brazil question, you presented the technology here and talked to the Multidisciplinary Center for Development of Ceramic Materials, one of the FAPESP Research, Innovation and Dissemination Centers. Why?
Professors Elson Longo and José Arana Varela had already studied the science of ferroelectrics so they knew the value of what I was presenting and wanted to collaborate in personnel training. But the project did not go forward because there was no investment. There were promises from businessmen from São Carlos, from governments, the BNDES and ministries, but nothing came of it. It was very frustrating and I would rather not even talk about it. I kept waiting for someone to say “let’s do it because it’s good for Brazil,” but this doesn’t happen, I never had that conversation. But I’m not angry.
How much would the investment have been at that time?
Look, it wasn’t an investment for me. It was for Brazil. Someone had to say “I want to set up a semiconductor factory,” but Ceitec was already operating in Rio Grande do Sul, funded by the federal government, there was no room for us. The investment in São Carlos would have been a maximum of $70 million.
Where was the factory built?
In Japan. Panasonic built the factory within an existing one, with $20 million, and today, it’s the company’s most profitable integrated circuit plant.
How do you see the field of semiconductors in Brazil today?
Today there are huge complications. Brazil needs a semiconductor industry, primarily due to the size of the economy. Chips are the oil of electronics. The world is divided into countries that have chip independence and those that don’t. Some companies in Brazil are designing chips, but it is still just beginning. The fundamental problem, in terms of technology, is that we mix science with technology. Brazil needs a ministry of technology. We need to separate high tech from basic and academic science.
Not a ministry of science and technology?
Brazil needs a ministry of science, to provide scholarships to students, and to fund research, but technology at the level that can transform the country, which determines how we compete with other countries, has to be seen as a way to protect Brazilian technological development. Remember the time of the Falklands War? No one was selling anything to Argentina. They had a chip shortage. Since everything will be increasingly dependent on chips, more and more information, a country the size of Brazil, with the seventh largest population in the world, absolutely cannot be a chip technology colony, because right now all of the chips used here have to be imported. Thus, we are in the information age without the basic raw material, the chip.
You have given some lectures in which you speak of the future of the future. What are they about?
I gave a talk a long time ago, in Japan, called The Future of the Future. It was a talk to several Japanese executives and became somewhat famous. Last year I did a series of lectures funded by the American consulate in Japan, along with other inventors, the one who invented the plasma screen, the Blu-ray laser, etc. The lectures were later turned into a book that is now required reading in MBA courses in Japan. But The Future of the Future was quite different. I noted that the thing that ages fastest is our vision of the future. When I was young, everyone knew there was going to be a future, that man would go to the moon, that there would be microwaves and telephone watches. But I came to the conclusion that we no longer have a vision of the future. Everything happens quickly and those who have the power to capitalize on ideas create a future that the rest have to follow. Before, you could buy one device for music and another for video, but now they are integrated into a single device and the level of integration will continue to increase. Next you’ll have iPhones that conduct medical examinations. So, if you are the owner of Sony, you will have a perpetual headache, because everything that is done is quickly destroyed. So, how do you create value? And this goes along with the Brazilian need to build a factory, have manpower, assemble things, but it’s more than that. The important thing is creating these things. Today, a 3D printer does almost everything. With this high level of integration, the value of information transfer is now greater than the value of the object that you use for information or even to manufacture parts. The Internet, cloud computing—in the United States these things are now leading to an area called big data, which is like doing a search and finding data connections, looking for trends, defining products and manufacturing them with robots. Labor needs to be very advanced and educational institutions have to anticipate that.
And where does Brazil fit into this?
How will you create jobs for all those engineering school graduates if you have only one Apple? How can you generate quality employment? Is no longer the size of the factory, but rather the size of the idea. If you have no ideas, and there is no direct financing channel to create these genuinely innovative, Brazilian ideas so Brazil can export to the world, Brazil will remain perpetually behind. And the politicians are unable to understand this. Not even the ministries are prepared for it. In the United States, Japan, anywhere, the fundamental question is: where are the white-collar jobs for scientists and engineers? But wealth creation is the reason for new technologies. And as technology is always making itself obsolete, it creates wealth and destroys wealth.
Is obsolescence a problem?
Yes, all over the world.
Now, changing the subject a little, you have also worked on artificial neural networks and have already published in biology journals.
I have a background in electronics engineering divided into systems and components. From time to time I do something different in systems, such as artificial intelligence with artificial neurons. The latest is with a student from Spain. We’re using an inexpensive brainwave reader we bought through Amazon that reads brainwaves through a crown placed on the head via Bluetooth, using a smartphone, for example. The waves are separated by frequency, for example, a 20 Hz beta wave, a 4 Hz theta wave. Thus, we can use a binary frequency identifier for these waves, so if you relax there is a theta wave, whereas if you focus there is a beta wave. In two months, he was able to write software to create a text processor for paraplegics. The person focuses on the letter A in a word processor. If the person blinks or relaxes, the letter changes, for example. It’s very easy to do this type of project with a $100 sensor and a mobile phone or tablet.
This work is only at the university?
Yes, these more esoteric things I do through the university. Now I’m connecting the word processor for paraplegics to a digital 3D printer, because then the paraplegic can look at a computer screen and look for an object that he wants. Just by thinking. He chooses the menu and selects something. Thus, a person’s thoughts can be transformed into something in 3D on a printer, but sometimes it goes wrong. This still requires a lot of work. This kind of work gives me a great deal of pleasure and students love these projects that can be carried out much faster than spending years inventing non-volatile memory.
