As the world’s energy demands increase, it is estimated that, over the next 100 years, the world will need a total of 0.2 yotta joules or about five times as much power as humans have consumed over the past several centuries. If the currently available energy deriving methods were to be continued, carbon emissions from fossil fuels would cause global temperatures to rise above tenable levels in just 30 years. Though, zero-carbon sources such as solar, wind, etc., might offer reasonable alternatives, these cleaner options are prohibitively expensive.1 Some energy experts believe that, power derived from nuclear fusion could become a promising option for replacing fossil fuels as the world's primary energy source and also, could play a pivotal role in addressing climate change issues.2
Fusion power generates electricity from the heat of nuclear fusion reactions in which two lighter atomic nuclei combine to form a heavier nucleus, while releasing energy, in the form of plasma. Fusion reactors are the devices designed to confine the plasma to harness energy. There are various methods for confining the plasma in nuclear reactors such as gravitational (e.g., alpha process, CNO cycle, etc.), magnetic (e.g., tokamak, stellarator, etc.), inertial (e.g., laser-driven, magnetized liner inertial fusion, etc.), electrostatic (e.g., fusor, polywell, etc.), and others (e.g., muon-catalyzed, magma, etc.).3
Even though the concept of nuclear fusion has been around for nearly a century, with the Russians making the first nuclear reactor in the 1960s to test various scientific theories,4 the major step toward reaching this goal is the initiation of ITER (International Thermonuclear Experimental Reactor) project, a 35-nation collaboration to design, build and operate an experimental reactor to achieve and sustain a fusion reaction for a very short period of time. ITER will be the world’s largest tokamak, a donut-shaped configuration for the containment of the plasma, where the reaction at temperatures hotter than the Sun will take place.2
Tokamak with doughnut-shaped vacuum chamber (Source)
Other tokomak devices that have been developed around the world include the Joint European Torus (JET), the world's largest and most powerful tokamak in operation today and the focal point of European fusion research is located at Culham Centre for Fusion Energy in the UK. H-mode type, ASDEX tokamak at the Max Planck Institute of Plasma Physics in Germany, was made in 1982. Alcator C-Mod, a compact tokamak capable of high magnetic field and dense, well-confined plasmas is located at The Plasma Science and Fusion Center at the Massachusetts Institute of Technology (MIT). The DIII-D advanced tokamak was developed by General Atomics in San Diego (US) as part of the international effort to achieve magnetically confined fusion. First plasma was achieved on the Experimental Advanced Superconducting Tokamak (EAST) in 2006, at the Institute of Plasma Physics in Hefei, China. HL-2M is a medium-sized copper-conductor tokamak under construction at the Southwestern Institute of Physics (SWIP) in China. ADITYA (synonym of Sun in Hindi) is the first indigenously designed and fabricated tokamak in India, located at the Institute for Plasma Research in Gujarat, under operation since 1989.5
Despite the potential benefits to society from nuclear fusion, its science remains one of the most pressing areas of experimental physics due to the complexity and challenges in controlling thermonuclear fusion for energy.6 The tricky part of nuclear fusion is generating more energy than is used in the process. Such reactors have to mimic conditions found only in deep space, a much more complex and costlier affair than fission. Heating plasma to temperatures higher than stars and then containing the ensuing reactions inside cryogenic cooling vessels can require machinery comprising of a million parts or more.6 Also, one of the most difficult challenges is the avoidance of large-scale plasma instabilities within these fusion reactors called disruptions, which can halt power production and damage key components.7 Thus, finding the right materials to construct the fusion reactor and developing the mechanism that will be used to extract the enormous energy/heat that is emitted are among the major tasks ahead.2 However, recent advances in areas such as development of exotic materials, 3D printing, machine learning and data processing etc., are changing the scenario of nuclear fusion energy production and paving the way for commercialization of fusion power.6
A tungsten alloy, developed by researchers at Los Alamos National Laboratory in New Mexico, can withstand unprecedented amounts of radiation without damage, making it suitable for the interior of fusion reactors. Moreover, mechanical properties of the alloy are retained after irradiation, while traditional counterparts degrade easily when irradiated.8
Researchers at the University of Rochester’s Laboratory for Laser Energetics (LLE) have created a dense plasma of deuterium by creating a high-density liquid deuterium by first lowering its temperature to 21 oK (-422 oF) and then rapidly increasing to almost 180,000 oF.9
General Fusion, in collaboration with GE, is using a three-dimensional printer to make mesh cages that will spin a combination of liquid lead and lithium to create a swirling sphere holed inside for injected plasma. This is a plastic prototype of the cage is shown below. Achieving and maintaining the right shape in the liquid metal was a problem during earlier attempts—the liquid metal had to move in such a way that it pressurized the plasma but never actually touched it.10
3D printed plastic prototype of the cage (Source)
The Nuclear Research and Consultancy Group (NRG) of The Netherlands has completed the irradiation testing in the High Flux Reactor of a new type of Eurofer 97-type alloy steel that will be used in a test programme of the ITER.11
Blocks of Eurofer97 alloy steel (Source)
Application of Deep Learning to nuclear fusion research is vital as it can provide valuable insights based on the tremendous data generated from the fusion reactors. For instance, JET (Joint European Torus) has over one hundred diagnostic systems to monitor the happenings inside the plasma, and each 30-second experiment (or pulse) generates about 50 GB of data. Reconstruction of the 2D plasma proﬁle inside the device can be carried out using Convolutional Neural Networks (CNNs) based on data coming from those diagnostics. Prediction of disruptions, which is a major problem affecting tokamaks today, can be done using Recurrent neural networks (RNNs). Training of such networks is performed on NVIDIA GPUs.12 With the aid of data science, Rochester’s laser lab moved closer to controlled nuclear fusion by bridging the gap between experiments and simulations. To create a predictive model, the team applied data science techniques to results from about 100 previous fusion experiments with OMEGA laser.13
An artificial intelligence project to help bring the power of the sun to earth is picked for the first U.S. exascale system. Princeton Plasma Physics Laboratory’s deep-learning software called the "Fusion Recurrent Neural Network (FRNN)", is composed of convolutional and recurrent neural nets to quickly predict when disruptions will break out in large-scale tokamak plasmas14. A method for predicting and preventing disruptive instabilities in controlled fusion plasmas through deep learning, has been developed by researchers at Harvard University and MIT.7,15 The team at Lawrence Livermore National Laboratory (LLNL), used validated models of the interactions of neutrons and protons and a powerful ab initio reaction method to accurately predict the properties of the polarized deuterium-tritium (DT) thermonuclear fusion. The research establishes a better understanding of the rate of DT fusion in a polarized plasma.16
General Fusion is hopeful about succeeding in the development of fusion reactor because of the data analytics techniques it has implemented with help from Microsoft. These techniques may help during its research phase, enabling engineers to optimize the reaction conditions quickly. Data analytics methods could also assist during reactor operation, helping adjust pressure and other conditions and to keep the liquid metal and plasma in proper state. By making these small corrections, one could avoid losing control of the plasma, which otherwise would forcibly turn off the machine.10 Google has teamed up with nuclear fusion company TAE Technologies (Formerly Tri Alpha Energy) to bring artificial intelligence into the world of plasma research. Researchers from both companies have come together and created a machine-learning tool based on what’s called as the Optometrist Algorithm.17
The central confinement chamber (Photo courtesy of Tri Alpha Energy Inc.) (Source)
Andrew Maris et al. from Lawrence Livermore National Laboratory have also applied machine learning to analyse the experimental data derived from inertial confinement fusion experiments performed at National Ignition Facility.18 Machine learning has been applied by researchers from Los Alamos National Laboratory for small X-ray data sets collected from the National Ignition Facility (NIF) images, in order to study the compression and asymmetry propagation with single- and double-shell inertial confinement fusion targets.19
A few interesting patents have been presented below:
A plasma confinement system disclosed by TAE Technologies (EP3452842A1), comprises a vessel, a magnetic coil positioned about the vessel, a combination probe comprising a flux loop and a B-dot probe positioned about the interior wall of the vessel, wherein the combination probe is configured to prevent twisting by fashioning it into a curved shape.
WO19008349A1 from Tokamak Energy Ltd. deals with a method of controlling a plasma in a nuclear fusion reactor. Using a machine learning approach, the control system is trained during the initial period of reactor operation, such that when primary sensors inevitably fail or degrade, the control system can determine the state of the plasma from the secondary sensors and the primary sensors, which are still active. Machine learning approaches are particularly relevant, as the actual sensors used are likely to vary between reactors - both due to primary sensors failing at different rates in different machines, and due to small deviations in manufacturing resulting in differences in the relationship between the plasma state and secondary sensors.
A tabletop reactor for controlling fusion activities covering a spectrum of reactions including aneutronic reactions such as proton-boron-11 fusion reactions has been developed by Alpha Ring International (WO2018208858).
Boeing has patented a unique method for electromagnetic control of plasmas in fusion power reactor environments using liquid lithium walls (US20170221589).
UJK Management GmbH has developed a method for generating electrical energy based on the fusion of protons with boron isotope 11, using laser radiation and magnetic fields (US10410752B2).
With billions of dollars having been invested in government initiatives, company projects, and dozens of nuclear fusion start-ups., Lockheed Martin's compact fusion reactor is an example of a company initiative in nuclear fusion technologies.20 The company expects to build a prototype fusion reactor in five years and deploy ones that are small enough to fit on the back of a truck in the next 10 years. The size of its device will enable its usage in spacecraft, ships, and city power stations.21 Amazon’s Jeff Bezos and companies such as Cenovus Energy have sunk more than $127 million into General Fusion, a start-up trying to commercialize fusion energy.22 Commonwealth Fusion Systems (CFS), a Boston-based start-up run by MIT researchers and funded by Breakthrough Energy Ventures led by a group of billionaires- including Bill Gates, Jeff Bezos, Jack Ma, Mukesh Ambani, and Richard Branson- believes it can bring fusion power to market in 15 years. Entrepreneur Vinod Khosla has also invested in CFS.23 MIT fusion collaboration receives renewed funding from Department of Energy for Plasma Science and Fusion Center to advance fusion studies on the world’s largest stellarator.24 Constructions industrielles de la Méditerranée (CNIM), in collaboration with ITER Organization, is designing and manufacturing high-precision handling equipment to assemble components of the future fusion reactor.25
Nuclear fusion power generation is still in the developmental stage and has great potential in the near future. Development of novel materials with outstanding properties as well as advances in technologies such as 3D printing are a boon to the fabrication of sophisticated nuclear fusion reactors. Furthermore, advancements in the field of Artificial Intelligence and Machine Learning will provide an impetus to the establishment of nuclear fusion reactors enabling production of clean, affordable and sustainable energy.