He talks about neural networks with such enthusiasm that it is no wonder that his project also caught the attention of the evaluators of the prestigious domestic Junior Star grant. It is intended for excellent early-career scientists who already have experience abroad. Dr Filip Ligmajer, who works at the Institute of Physical Engineering (IPE) FME and CEITEC BUT, has received funding for five years for the new direction of research. However, the talented researcher was not in love with physics from his childhood, but somewhat unexpectedly, it was a book that led him to physics at the end of his high school studies grammar school.
What made you interested in physics?
I had good grades at grammar school but never studied much. (laughs) Then, before graduating from high school, I discovered the book Alice in the Realm of Quantas by Robert Gilmore. We were just taking quantum mechanics and this is an allegory for the fairy tale Alice in Wonderland. Excited by the book, I started reading other popular science books on physics, specifically by the famous American physicist Richard Feynman, and I completely fell in love with it. During my graduation year, I changed my ideas about university and considered whether to study engineering or try physics.
What finally decided?
It was one of the discussions at the Open Day at FSI. When I came to the IPE stand, Dr Jakub Zlámal was there, who explained to me what the study of Physical Engineering is about and that if I didn't succeed, I could transfer to the Basics of Mechanical Engineering. That calmed me down. Until then, I thought that only those who took part in the Physics Olympiads could go to study physics.
Currently, you work as a researcher and teacher at the IPE and also at CEITEC. With a project called Phase Transition Materials for Photonic Neural Networks and Neuromorphic Computing, you received the prestigious Junior Star grant. What is the project about?
The project concerns neural networks and neuromorphic computing, a computer technology that is not based on a classical computer (processor, memory, ones and zeros), but is similar to how neurons work in the brain. Neural networks today still work overwhelmingly digitally – that is, with the help of a classic computer, where we emulate these networks with ones and zeros, converting their behaviour into matrix multiplication... However, the trend for the future is that the hardware, i.e. the machines themselves, will be designed completely differently so that they are suitable for neural networks. Today, we graft technology onto something that was invented in the last century for classical counting. In this project, my team and I will deal with the hardware and optical implementation. That's a jump forward two levels.
How is the hardware supposed to work then?
There are two approaches – electrical and optical. In the case of electricity, it is, simply put, whether or not current flows in the function of neurons. It's very similar to what happens in the human brain. In the optical solution, we use photons instead of electrons, again for simplicity – it’s like the light is either on or off.
Hardware for electrical neural networks is at a fairly advanced level today. Some processors are tailor-made for certain tasks of neuromorphic computing or artificial intelligence. It's different with optics, we currently have optical cables designed for the Internet (communication in servers), but for example, computers in mobile phones use electrical processing units, and there are no photons. It's similar to hardware for artificial intelligence, where the electric approach is explored first and only then do we turn to photonics.
The project is planned to span over five years. What will you focus on during this period?
Our team will deal with how to create transistors or basic computing units for photonic neuromorphic networks. It is basic research, where we test in the laboratory whether and how the basic principles of physics work. With photonic systems, you have a choice of multiple approaches, and no one currently knows which one will work in the future.
Personally, I have been dealing with one particular material for the last decade, and that is vanadium dioxide, which shows phase transition. When you heat it to 68 °C, it will transform from an insulator to metal. The difference in electrical resistance is quite significant, like the difference between a steel wire and a wooden toothpick. Therefore, it is an extremely interesting material. I would like to implement it in photonic waveguides, i.e. in those places where photons "run", and I would appropriately modulate optical signals with the help of this oxide. In the project, we assume that we could produce one neuron or one more complex functional unit. Producing an entire processor with billions of neural connections is an investment that only the largest companies in the world can make today.
What attracted the expert evaluators who decided on the award of the Junior Star grant to your project?
I think the fact that I'm putting two things together. In the beginning, I only deal with basic functional principles, putting only a small strip of the material on the waveguides, and only gradually we involve nanotechnology approaches. We will move from the microworld to the nanoworld, where we will be nanostructuring down to the level of tens or hundreds of nanometers. We are looking at what changes it will bring, and whether the efficiency of the modulation will increase. It can be compared to burning a CD – information is stored in it by high-power light and read by low-power light. Vanadium dioxide can be handled similarly. If we exceed the temperature of 68 °C, we can write down the information and then read the information with a weaker pulse. The moment you suitably connect these waveguides, the network of these waveguides begins to function as a neural network.
How can a layman imagine working on this type of project?
In our field, it always has three sides like a triangle. The first part is a computer-aided design and simulation of physical effects. Then it is manufactured according to the design, which is extremely demanding work, for which we mainly use the CEITEC Nano shared laboratories. The last step is to analyze whether the manufactured model behaves according to our assumptions, or whether the design needs to be modified in some way. All our designs are more or less on the edge of technological possibilities and knowledge, so we will work in experimental mode.
How can you estimate where you will go in five years when working on an experimental project?
This type of project is about setting up a new laboratory with young researchers and starting to do research in a new area. The project is very well planned for the first two to three years, and we know exactly what we want to achieve. What's interesting about science is that you can produce an equally valuable result about something that is not working as I will rule out one of the ways. Of course, it's not as sexy as discovering what works, but that's the way it is, and you have to come to terms with that, not only in this field. I also count on the fact that at the end of the project, we will deal with something a little different.
When a researcher does basic research, he or she often experiences failure. How do you personally cope with it?
My mindset is that I don't do research for the result, but I'm interested in the journey. I focus on precise preparation. I want us to try to make it as well as possible, not to neglect anything, to be well prepared and to know the available literature and the work of our colleagues. Of course, it wouldn't be true if I said that I don't care about the result, but I don't derive my feeling of satisfaction from whether the experiment works out for me or not – because then I would be unhappy in 99 per cent of cases.
Students are just fully experiencing the exam period. What would you say to those for whom physics is a bit of a scarecrow?
If I am to be successful in life, to achieve something, to be satisfied with myself, I have to use my head for something useful. In physics, it's nice to train the ability to take on a task that I don't know how to solve, but I can make it using my head. The people I know in science – and they are not just physicists – are also skilled outside their fields. They don't care what problem they face, because they have it set up in such a way that they go and try to solve it. It's not about what rating I get, but about how I approached solving the problem.
How do you personally see the role of physics in the future?
The world is changing so rapidly today. Any specialisation we choose today may be replaced or changed in such a way that it will be useless in five years' time. But if I learn to think in a way that I understand mathematics, physics, materials, and design, that's something the world will always need. I will always be useful to companies and society as a whole.