What Do Scientists Do with Supercomputers?
When he was young, physicist Kirils Surovovs imagined the life of a scientist rather romantically – people writing formulas on a blackboard with chalk, thinking deeply and making discoveries. Today, the blackboard has largely been replaced by a computer – and not just an ordinary computer, but a supercomputer. As a researcher at the Institute of Numerical Modelling at the University of Latvia’s Faculty of Science and Technology, he uses it to virtually reproduce the process of growing silicon crystals, exploring ways to improve this complex and costly process.
As part of the European Union co-funded project “Development of Scientific Infrastructure at the University of Latvia in RIS3 Priority Areas”, the University of Latvia acquired five new high-performance servers this year. They have significantly increased the university’s available computing capacity, allowing researchers to tackle more complex and computationally demanding tasks. In Kirils’ case, a calculation that previously took an entire week can now be completed in less than a day.
The new servers have already changed the rhythm of his work. “Previously, while the computer was running calculations, I could process the previous results and write reports. Now I sometimes haven’t even finished processing one set of results before the next ones arrive. It’s a very pleasant problem to have,” he says with a smile.
Kirils specialises in the numerical modelling of silicon crystal growth. Put simply, he uses computers to study what happens inside a crystal-growing system and helps manufacturers understand how the process can be improved.
Silicon crystals are the raw material used to make the microchips on which modern electronics depend.
“Silicon is literally everywhere – in computers, tablets, mobile phones and other electronic devices. It remains at the heart of electronics. It is no coincidence that Silicon Valley is called Silicon Valley,” says Kirils.
Using specialised software, the researcher models the movement of molten silicon, temperature distribution and the spread of impurities. This makes it possible to predict how changes in the growth process will affect crystal quality before they are tested in a factory.
Such simulations help save both time and money. Silicon melts at temperatures above 1,400 degrees Celsius, making real-world experiments highly energy-intensive and expensive – a single experiment can cost as much as €30,000–€40,000.
“For that amount of money, you could employ a researcher for several months who could carry out hundreds of simulations. Our task is to make sure that these simulations are sufficiently accurate and useful to industry,” Kirils explains.
Virtual calculations do not completely replace experiments, however – their results need to be verified and compared with real-world data. For this reason, researchers at the institute work closely with scientific institutions and companies abroad, including the German company Siltronic, one of the world’s leading manufacturers of ultra-pure silicon wafers. University of Latvia researchers help address specific production challenges, improve models and test new approaches, while some of the results obtained through this cooperation are confidential.
Silicon crystal growth is just one application of numerical modelling. At the University of Latvia’s Institute of Numerical Modelling, it is also used to study building energy efficiency, metallurgical processes, weather, climate and groundwater. High-performance computing is also essential at the university for bioinformatics, pharmaceutical research, genome studies and the development of artificial intelligence solutions. For example, researchers at the institute combine numerical modelling and artificial intelligence to identify bubbles invisible to the naked eye in data from steel-casting experiments – defects that can reduce the strength of the material.
Asked what motivates him most in his work, Kirils highlights the opportunity to tackle increasingly new and challenging problems and to see how a clear, practical result gradually emerges from complex calculations. Of course, a researcher’s daily work consists not only of exciting simulations and impressive graphs, but also reports, communication with colleagues and partners, as well as administrative tasks.
“Sometimes you have to write not only poetry, but also prose,” he says with a smile, describing the two sides of a scientist’s work.
Kirils encourages young people not to be afraid of choosing a career in science – especially now, when artificial intelligence tools are rapidly changing the nature of many professions. More and more tasks can be automated, while science still requires the ability to navigate the unknown, formulate new questions and connect knowledge from different fields. For this reason, Kirils believes that a career in science could become one of the more resilient career paths of the future.
Moreover, studying science does not mean that one’s entire professional life must necessarily be spent in academia.
“Often, the best choice today is the one that expands your options for the future. The analytical and problem-solving skills gained through studying physics and other sciences can later be applied in research, information technology, artificial intelligence, finance and many other fields,” he says.
This story was produced with support from EU funds.
