Computer simulations show Swedish fusion initiative could have global impact

Magnetic mirror machines 

Novatron is modern variation of a magnetic mirror machine, an architecture that was first demonstrated in 1955 at what is now the Lawrence Livermore National Laboratory. The basic concept is to arrange two large magnets in such a way as to reflect charged particles back and forth between regions of the strong magnetic fields. 

The magnetic mirror is a simple approach to fusion, with some clear advantages, according to Oden. A low-cost solution with easy fueling and no need for special equipment to get rid of excess heat caused by neutron flux, it offers continuous steady state operation and a very high beta, which is the ratio of plasma pressure to magnetic pressure. High beta machines achieve higher output with less magnetic forces, making them more economically viable than those with low beta approaches – all other things being the same. 

The biggest challenges with traditional mirror machines are the instability and poor confinement time. Confinement time – how long a machine can keep plasma fuel in place – is one of the three conditions necessary to start and sustain a fusion reaction, with the other two being density and heat. According to Oden, the scientific community has come up with solutions over the years to improve stability and confinement time in mirror machines. 

One solution is based on ideas from Jan Jäderberg, co-founder and CTO of Novatron. Jäderberg found ways of minimising problems with plasma instabilities that have stood in the way of making fusion energy a reality.  

The solution Novatron aims to build is an axisymmetric tandem mirror (ATM) that combines two basic fusion concepts – magnetic mirrors and biconic cusps. The classical magnetic mirror has a magnetic field pointing in the normal direction of the symmetry plane and offers good confinement but is unstable. Biconic cusps, on the other hand, produce a magnetic field that is tangential to the symmetry plane and is stable but with poor confinement.

Novatron’s unique concept is a new category of magnetic confinement, with a normal magnetic field at the symmetry plane such as the classic mirror, but with an added pair of biconic cusps.

“We create a magnetic field that is convex, looking from the inside, and concave, looking from the outside,” said Oden. “This results in very good confinement – and it’s inherently stable.”

According to Oden, all the magnets are circular, so it’s easy to produce and it’s easy to stabilise mechanically. Novatron also achieves a very high beta and has many different ways of increasing performance, making the confinement longer and increasing the plasma volume. Moreover, the ATM design makes it relatively easy to keep the plasma burning without having to apply the complicated heating devices required by other approaches. 

The power of computer simulations 

Like all fusion projects, Novatron has a roadmap that consists of several phases, starting with simulations that allow researchers to validate and fine tune the underlying physics and engineering of their approach and ending with a commercial fusion reactor a decade or so away.   

To simulate their architecture, Novatron relies on specialists in computer simulation and in physical modelling – including Rickard Holmberg, computational physicist and software engineer at Novatron. The simulation platform the company uses most of the time is WarpX, which was written and is maintained by a team of researchers, mainly centered at Lawrence Berkeley National Laboratory, SLAC National Accelerator Laboratory, and Lawrence Livermore National Laboratory.

“We made some improvements and adaptations to the base WarpX to make it more suited to our needs,” Rickard Holmberg told Computer Weekly. 

The simulation software runs on a very wide range of hardware – from a single GPU on a workstation to large clusters. It runs on both AMD and NVIDIA GPUs, and it can use OpenMP and MPI for CPU parallelisation. Novatron runs more simulations on their own hardware, but they sometimes use small cloud resources. “We are looking to expand to large clusters,” said Holmberg. 

The company has performed extensive computer verification and stress-test simulations that show Novatron is stable – in stark contrast to the notoriously unstable classical mirror approach.

“Our calculations have also indicated that we will have an energy confinement time improvement of a factor of 100 over traditional magnetic mirror machines,” said Oden. 

A series of experimental machines 

“We’re now commissioning our first experimental reactor, called Novatron 1, in Stockholm at the Royal Institute of Technology at the LVM Laboratory,” said Oden. “This will be a core cell where we show stable plasma. We plan a second experimental reactor, Novatron 2, where we will create our first axisymmetric tandem mirror and add performance-boosting functionality. We expect to achieve fusion conditions by 2027.”

Like all other fusion companies, before putting power on a grid, a prototypical machine will be produced to show the commercial feasibility. This will be Novatron 3, a pilot reactor that will be completed in the 2030s, assuming the company sticks to its current roadmap. 

Finally, Novatron 4 will be a full-blown fusion reactor for a commercial power plant. If all goes well, that will be accomplished in the 2030s. The company hopes its unique architecture will become the preferred approach to commercially viable source of clean, safe and virtually limitless energy on a global scale.