Master of the flow

In his life as a researcher, Athanasios Mokos has already investigated a wide variety of phenomena that involve fluids: how the action of breaking waves shapes the coast, what happens in wet concrete, and – currently – how mineral deposits form on the fuel rods of a nuclear reactor. The 37-year-old Greek scientist has been able to do all this, however, without traveling to the sea, visiting a construction site, or setting foot in a nuclear power plant. All he needed was his sharp mathematical mind and a high-performance computer: Mokos is a modeler in the Laboratory for Waste Management, which is part of the Center for Nuclear Engineering and Sciences at PSI. 

His passion for simulating fluids and everything related to them began during his engineering studies in Athens. He gained access to the field through a difficult entrance exam he took after high school: “I’ve been interested in mathematics and physics since my school days,” Mokos recalls. “Luckily, I had great teachers who made everything very interesting.” During his studies, he was particularly fascinated by fluid dynamics – that is, the study of flowing liquids and gases – and the possibility of simulating very complex processes with the help of ingenious mathematical models.

In his latest project, Mokos wanted to find out how porous deposits on the fuel rods of a nuclear reactor can influence the heat exchange between the fuel rod and water. “It’s a bit like what happens in a teakettle – over time, calcium builds up, which can for example make the kettle louder when the water boils,” Mokos says. In a nuclear power plant, naturally, the effects are more far-reaching; deposits can have a significant impact on the plant’s efficiency.

Complex simulations with the supercomputer

For many years, researchers in the Laboratory for Waste Management and the Laboratory for Simulations and Modeling have been developing numerical methods for models of flows in nuclear reactors. This is where Mokos's simulations come into play. “To describe the complex interaction of the hot water in the reactor, the fuel rods, and the gas bubbles in the porous deposits mathematically, I use what’s called the lattice Boltzmann method,” Mokos explains. He was able to gain initial experience with this method through intensive collaboration with group head Nikolaos Prasianakis, who has been leading the development of lattice Boltzmann software for more than 15 years.

In this method, the liquid to be simulated is divided into small units and stochastically – that is, randomly –  modeled to determine whether neighbouring units collide and whether bubbles form within a unit due to boiling water. Because of the high spatial resolution required, just a few micrometres, this can’t be done on a normal PC: “The storage space required for the simulation alone is roughly equivalent to that of 40 graphics cards or GPUs,” Mokos says. Many years ago, for his dissertation in England, he was among the first to use the then-emerging high-performance GPUs (graphics processing units) to accelerate his simulations of multiphase fluid dynamics. For the reactor project, however, he now had 100,000 computing hours at his disposal on the Piz Daint supercomputer of the Swiss National Supercomputing Center CSCS. On this, his simulations ran roughly 20,000 times faster than on a normal PC.

Mokos looks back fondly on his time as a doctoral candidate in Manchester: “Thanks to the three universities there, everything is very international, there are lots of cultural activities, and you meet many people.” After four years in the doctoral programme and three years as a postdoc, however, he needed a change of scenery. In Paris, in 2018, the opportunity arose to collaborate with researchers from the Saint-Venant Hydraulics Laboratory, whose research he knew from conferences they had attended together, and which he greatly respected. There he developed simulations of flooding in large facilities – such as nuclear power plants, for example.

For the computer models, Mokos used a slightly different method called Smoothed Particle Hydrodynamics (SPH), which is better suited to the simulation of complex surfaces. After three years of research, however, the sponsor Électricité de France (EDF) abruptly stopped funding projects based on this method. This also ended Mokos's work in Paris. In search of a new job, he applied to PSI, among others, in 2021. “I had never been to Switzerland at the time,» he says, «but my doctoral supervisor in Manchester had spent a year there as a postdoc and recommended PSI to me as an excellent research centre.”

Prize-winning lecture

Mokos got the job and was surprised when he arrived in Villigen that no one spoke French there: “I knew that French was spoken on Lake Geneva and thought that was the case all over Switzerland.” In the lab, of course, he was able to communicate in English without any problems, and he says his German is now quite passable. He likes the good atmosphere at PSI, and the lively intellectual exchange with colleagues that often takes place during coffee breaks. He also enjoys life in Switzerland: “The cities here – even the big ones – are much smaller than Manchester, Paris, or even Athens, which means you are closer to nature.” Since he doesn't drive a car, he appreciates the good rail connections, which he often uses to go hiking. His most challenging excursion to date was a hike with friends last year in Monte Rosa. “We even had to cross a glacier,” Mokos says.

Mokos maintains his Greek roots even though he is far away. There is a fairly large Greek community in Zurich, for which he has been serving as treasurer for some time: «We recently organised a large event with 400 people for Tsiknopempti, the barbecue on the next-to-last Thursday before Lent.»

For his presentation on the work at PSI on heat transport in boiling water reactors, Mokos was recently awarded the gold medal at the Discrete Simulation of Fluid Dynamics (DSFD) conference in Zurich. The topic continues to occupy him: He is currently working on a large-scale model of the entire reactor. For this he is integrating neural networks into the simulations. He then wants to check the accuracy of the simulations against laboratory models that have been built by other research institutions.

Mokos can definitely picture himself continuing to pursue research in this field. It is therefore quite likely that he will continue to work on reactor simulations for some time. It is also possible he will turn his versatile computer models to completely different fluid phenomena.