Explore the vast range of titles available.

Nuclear fusion using magnetic confinement, in particular in the tokamak configuration, is a promising path towards sustainable energy. A core challenge is to shape and maintain a high-temperature plasma within the tokamak vessel. This requires high-dimensional, high-frequency, closed-loop control using magnetic actuator coils, further complicated by the diverse requirements across a wide range of plasma configurations. In this work, we introduce a previously undescribed architecture for tokamak magnetic controller design that autonomously learns to command the full set of control coils. This architecture meets control objectives specified at a high level, at the same time satisfying physical and operational constraints. This approach has unprecedented flexibility and generality in problem specification and yields a notable reduction in design effort to produce new plasma configurations. We successfully produce and control a diverse set of plasma configurations on the Tokamak à Configuration Variable 1 , 2 , including elongated, conventional shapes, as well as advanced configurations, such as negative triangularity and ‘snowflake’ configurations. Our approach achieves accurate tracking of the location, current and shape for these configurations. We also demonstrate sustained ‘droplets’ on TCV, in which two separate plasmas are maintained simultaneously within the vessel. This represents a notable advance for tokamak feedback control, showing the potential of reinforcement learning to accelerate research in the fusion domain, and is one of the most challenging real-world systems to which reinforcement learning has been applied. A newly designed control architecture uses deep reinforcement learning to learn to command the coils of a tokamak, and successfully stabilizes a wide variety of fusion plasma configurations.

Obtaining a burning plasma is a critical step towards self-sustaining fusion energy 1 . A burning plasma is one in which the fusion reactions themselves are the primary source of heating in the plasma, which is necessary to sustain and propagate the burn, enabling high energy gain. After decades of fusion research, here we achieve a burning-plasma state in the laboratory. These experiments were conducted at the US National Ignition Facility, a laser facility delivering up to 1.9 megajoules of energy in pulses with peak powers up to 500 terawatts. We use the lasers to generate X-rays in a radiation cavity to indirectly drive a fuel-containing capsule via the X-ray ablation pressure, which results in the implosion process compressing and heating the fuel via mechanical work. The burning-plasma state was created using a strategy to increase the spatial scale of the capsule 2 , 3 through two different implosion concepts 4 – 7 . These experiments show fusion self-heating in excess of the mechanical work injected into the implosions, satisfying several burning-plasma metrics 3 , 8 . Additionally, we describe a subset of experiments that appear to have crossed the static self-heating boundary, where fusion heating surpasses the energy losses from radiation and conduction. These results provide an opportunity to study α-particle-dominated plasmas and burning-plasma physics in the laboratory. A burning plasma, a critical step towards self-sustaining fusion, is achieved at the US National Ignition Facility, with a subset of experiments demonstrating fusion self-heating beyond radiation and conduction losses.

In a burning plasma state 1 – 7 , alpha particles from deuterium–tritium fusion reactions redeposit their energy and are the dominant source of heating. This state has recently been achieved at the US National Ignition Facility 8 using indirect-drive inertial-confinement fusion. Our experiments use a laser-generated radiation-filled cavity (a hohlraum) to spherically implode capsules containing deuterium and tritium fuel in a central hot spot where the fusion reactions occur. We have developed more efficient hohlraums to implode larger fusion targets compared with previous experiments 9 , 10 . This delivered more energy to the hot spot, whereas other parameters were optimized to maintain the high pressures required for inertial-confinement fusion. We also report improvements in implosion symmetry control by moving energy between the laser beams 11 – 16 and designing advanced hohlraum geometry 17 that allows for these larger implosions to be driven at the present laser energy and power capability of the National Ignition Facility. These design changes resulted in fusion powers of 1.5 petawatts, greater than the input power of the laser, and 170 kJ of fusion energy 18 , 19 . Radiation hydrodynamics simulations 20 , 21 show energy deposition by alpha particles as the dominant term in the hot-spot energy balance, indicative of a burning plasma state. In burning plasma, alpha particles from fusion reactions are the dominant source of heating. The design choices that resulted in reaching this state in experiments at the National Ignition Facility are reported.

For stable and efficient fusion energy production using a tokamak reactor, it is essential to maintain a high-pressure hydrogenic plasma without plasma disruption. Therefore, it is necessary to actively control the tokamak based on the observed plasma state, to manoeuvre high-pressure plasma while avoiding tearing instability, the leading cause of disruptions. This presents an obstacle-avoidance problem for which artificial intelligence based on reinforcement learning has recently shown remarkable performance 1 – 4 . However, the obstacle here, the tearing instability, is difficult to forecast and is highly prone to terminating plasma operations, especially in the ITER baseline scenario. Previously, we developed a multimodal dynamic model that estimates the likelihood of future tearing instability based on signals from multiple diagnostics and actuators 5 . Here we harness this dynamic model as a training environment for reinforcement-learning artificial intelligence, facilitating automated instability prevention. We demonstrate artificial intelligence control to lower the possibility of disruptive tearing instabilities in DIII-D 6 , the largest magnetic fusion facility in the United States. The controller maintained the tearing likelihood under a given threshold, even under relatively unfavourable conditions of low safety factor and low torque. In particular, it allowed the plasma to actively track the stable path within the time-varying operational space while maintaining H-mode performance, which was challenging with traditional preprogrammed control. This controller paves the path to developing stable high-performance operational scenarios for future use in ITER. Artificial intelligence control is used to avoid the emergence of disruptive tearing instabilities in the magnetically confined fusion plasma in the DIII-D tokamak reactor.

Nuclear fusion is one of the most attractive alternatives to carbon-dependent energy sources 1 . Harnessing energy from nuclear fusion in a large reactor scale, however, still presents many scientific challenges despite the many years of research and steady advances in magnetic confinement approaches. State-of-the-art magnetic fusion devices cannot yet achieve a sustainable fusion performance, which requires a high temperature above 100 million kelvin and sufficient control of instabilities to ensure steady-state operation on the order of tens of seconds 2 , 3 . Here we report experiments at the Korea Superconducting Tokamak Advanced Research 4 device producing a plasma fusion regime that satisfies most of the above requirements: thanks to abundant fast ions stabilizing the core plasma turbulence, we generate plasmas at a temperature of 100 million kelvin lasting up to 20 seconds without plasma edge instabilities or impurity accumulation. A low plasma density combined with a moderate input power for operation is key to establishing this regime by preserving a high fraction of fast ions. This regime is rarely subject to disruption and can be sustained reliably even without a sophisticated control, and thus represents a promising path towards commercial fusion reactors. A magnetic confinement regime established at the Korea Superconducting Tokamak Advanced Research device enables the generation of plasmas over 10 8  kelvin for 20 seconds with the aid of fast ions without plasma edge instabilities or impurity accumulation.

The tokamak approach, utilizing a toroidal magnetic field configuration to confine a hot plasma, is one of the most promising designs for developing reactors that can exploit nuclear fusion to generate electrical energy 1 , 2 . To reach the goal of an economical reactor, most tokamak reactor designs 3 – 10 simultaneously require reaching a plasma line-averaged density above an empirical limit—the so-called Greenwald density 11 —and attaining an energy confinement quality better than the standard high-confinement mode 12 , 13 . However, such an operating regime has never been verified in experiments. In addition, a long-standing challenge in the high-confinement mode has been the compatibility between a high-performance core and avoiding large, transient edge perturbations that can cause very high heat loads on the plasma-facing-components in tokamaks. Here we report the demonstration of stable tokamak plasmas with a line-averaged density approximately 20% above the Greenwald density and an energy confinement quality of approximately 50% better than the standard high-confinement mode, which was realized by taking advantage of the enhanced suppression of turbulent transport granted by high density-gradients in the high-poloidal-beta scenario 14 , 15 . Furthermore, our experimental results show an integration of very low edge transient perturbations with the high normalized density and confinement core. The operating regime we report supports some critical requirements in many fusion reactor designs all over the world and opens a potential avenue to an operating point for producing economically attractive fusion energy. A stable tokamak plasma has been demonstrated with a high plasma density and a high energy confinement quality, both of which are simultaneously important for fusion reactors.

Divertor detachment offers a promising solution to the challenge of plasma-wall interactions for steady-state operation of fusion reactors. Here, we demonstrate the excellent compatibility of actively controlled full divertor detachment with a high-performance ( β N ~ 3, H 98 ~ 1.5) core plasma, using high-β p (poloidal beta, β p  > 2) scenario characterized by a sustained core internal transport barrier (ITB) and a modest edge transport barrier (ETB) in DIII-D tokamak. The high- β p high-confinement scenario facilitates divertor detachment which, in turn, promotes the development of an even stronger ITB at large radius with a weaker ETB. This self-organized synergy between ITB and ETB, leads to a net gain in energy confinement, in contrast to the net confinement loss caused by divertor detachment in standard H-modes. These results show the potential of integrating excellent core plasma performance with an efficient divertor solution, an essential step towards steady-state operation of reactor-grade plasmas. Plasma fusion devices like tokamaks are important for energy generation but there are many challenges for their steady state operation. Here, the authors show that full divertor detachment is compatible with high-confinement high-poloidal-beta core plasmas and this prevents the damage to the divertor target plates and the first wall.

When a heavy atomic nucleus splits (fission), the resulting fragments are observed to emerge spinning 1 ; this phenomenon has been a mystery in nuclear physics for over 40 years 2 , 3 . The internal generation of typically six or seven units of angular momentum in each fragment is particularly puzzling for systems that start with zero, or almost zero, spin. There are currently no experimental observations that enable decisive discrimination between the many competing theories for the mechanism that generates the angular momentum 4 – 12 . Nevertheless, the consensus is that excitation of collective vibrational modes generates the intrinsic spin before the nucleus splits (pre-scission). Here we show that there is no significant correlation between the spins of the fragment partners, which leads us to conclude that angular momentum in fission is actually generated after the nucleus splits (post-scission). We present comprehensive data showing that the average spin is strongly mass-dependent, varying in saw-tooth distributions. We observe no notable dependence of fragment spin on the mass or charge of the partner nucleus, confirming the uncorrelated post-scission nature of the spin mechanism. To explain these observations, we propose that the collective motion of nucleons in the ruptured neck of the fissioning system generates two independent torques, analogous to the snapping of an elastic band. A parameterization based on occupation of angular momentum states according to statistical theory describes the full range of experimental data well. This insight into the role of spin in nuclear fission is not only important for the fundamental understanding and theoretical description of fission, but also has consequences for the γ-ray heating problem in nuclear reactors 13 , 14 , for the study of the structure of neutron-rich isotopes 15 , 16 , and for the synthesis and stability of super-heavy elements 17 , 18 . γ-ray spectroscopy experiments on the origin of spin in the products of nuclear fission of spin-zero nuclei suggest that the fission fragments acquire their spin after scission, rather than before.

In nuclear fusion reactors, plasmas are heated to very high temperatures of more than 100 million kelvin and, in so-called tokamaks, they are confined by magnetic fields in the shape of a torus. Light nuclei, such as deuterium and tritium, undergo a fusion reaction that releases energy, making fusion a promising option for a sustainable and clean energy source. Tokamak plasmas, however, are prone to disruptions as a result of a sudden collapse of the system terminating the fusion reactions. As disruptions lead to an abrupt loss of confinement, they can cause irreversible damage to present-day fusion devices and are expected to have a more devastating effect in future devices. Disruptions expected in the next-generation tokamak, ITER, for example, could cause electromagnetic forces larger than the weight of an Airbus A380. Furthermore, the thermal loads in such an event could exceed the melting threshold of the most resistant state-of-the-art materials by more than an order of magnitude. To prevent disruptions or at least mitigate their detrimental effects, empirical models obtained with artificial intelligence methods, of which an overview is given here, are commonly employed to predict their occurrence—and ideally give enough time to introduce counteracting measures. Tokamak plasmas are prone to sudden collapses that terminate the nuclear fusion reactions. This perspective discusses the prediction of these so-called disruptions with artificial intelligence techniques.

Nuclear reactor safety and efficiency can be enhanced through the development of accurate and fast methods for prediction of reactor transient (RT) states. Physics informed neural networks (PINNs) leverage deep learning methods to provide an alternative approach to RT modeling. Applications of PINNs in monitoring of RTs for operator support requires near real-time model performance. However, as with all machine learning models, development of a PINN involves time-consuming model training. Here, we show that a transfer learning (TL-PINN) approach achieves significant performance gain, as measured by reduction of the number of iterations for model training. Using point kinetic equations (PKEs) model with six neutron precursor groups, constructed with experimental parameters of the Purdue University Reactor One (PUR-1) research reactor, we generated different RTs with experimentally relevant range of variables. The RTs were characterized using Hausdorff and Fréchet distance. We have demonstrated that pre-training TL-PINN on one RT results in up to two orders of magnitude acceleration in prediction of a different RT. The mean error for conventional PINN and TL-PINN models prediction of neutron densities is smaller than 1%. We have developed a correlation between TL-PINN performance acceleration and similarity measure of RTs, which can be used as a guide for application of TL-PINNs.