A supervised machine learning algorithm trained on a multi-petabyte dataset of inertial confinement fusion simulations has identified a class of implosions that robustly achieve high yield, even in the presence of drive variations and hydrodynamic perturbations. These implosions are purposefully driven with a time-varying asymmetry, such that coherent flow generation during hotspot stagnation forces the capsule to self-organize into an ovoid, a shape that appears to be more resilient to shell perturbations than spherical designs. This new class of implosions, whose configurations are reminiscent of zonal flows in magnetic fusion devices, may offer a path to robust inertial fusion.

Simulations using IPCC (Intergovernmental Panel on Climate Change)-class climate models are subject to fail or crash for a variety of reasons. Quantitative analysis of the failures can yield useful insights to better understand and improve the models. During the course of uncertainty quantification (UQ) ensemble simulations to assess the effects of ocean model parameter uncertainties on climate simulations, we experienced a series of simulation crashes within the Parallel Ocean Program (POP2) component of the Community Climate System Model (CCSM4). About 8.5% of our CCSM4 simulations failed for numerical reasons at combinations of POP2 parameter values. We applied support vector machine (SVM) classification from machine learning to quantify and predict the probability of failure as a function of the values of 18 POP2 parameters. A committee of SVM classifiers readily predicted model failures in an independent validation ensemble, as assessed by the area under the receiver operating characteristic (ROC) curve metric (AUC > 0.96). The causes of the simulation failures were determined through a global sensitivity analysis. Combinations of 8 parameters related to ocean mixing and viscosity from three different POP2 parameterizations were the major sources of the failures. This information can be used to improve POP2 and CCSM4 by incorporating correlations across the relevant parameters. Our method can also be used to quantify, predict, and understand simulation crashes in other complex geoscientific models.

Heavy-ion beams, each with current in the kiloampere range and particle energy in the giga-electronvolt range, must be focused onto a millimetre-size spot to provide the power required for ignition of high-gain targets for inertial confinement fusion. However, the focal spot size is always enlarged by chromatic aberration generated by the thermal spread of the beam ions in the direction of beam propagation. Enlarged focal spot degrades the target performance. For high-current beams, the conventional remedy for chromatic aberration using sextupole magnets has been shown to be ineffective. If novel correction schemes can be found, then the spot size can be reduced to below that previously believed possible. Smaller spots can mean lower energy targets so that the heavy-ion fusion (HIF) scenario can look more attractive. Success in laser cooling of ion beams in storage rings has inspired us to explore the feasibility of applying laser cooling for HIF, and the recirculator configuration proposed for HIF appears to be well suited for this purpose. However, using particle-in-cell simulations and theoretical arguments, we demonstrate in this paper that although laser cooling of heavy-ion beams is feasible in principle, the rapid velocity-space diffusion of ions in the bump-in-tail distribution, set up by the cooling lasers, limits the velocity-space compressibility of the thermal spread along the beam. Consequently, laser cooling is impractical for high-current, heavy-ion beams for the proposed recirculator configuration. Nevertheless, if the recirculator architecture or the target requirement can reduce the beam current, then the cooling scheme described here would be useful. This scheme may also be applicable to the RF linac and storage ring approach to HIF.

Several heavy ion beams, each with current in the kiloampere range and particle energy in the gigaelectronvolt range, must be focused onto a millimetre size spot to provide the power required for ignition of high gain, radiation driven targets for inertial confinement fusion. It is difficult to achieve this condition for three reasons: (a) The repulsive space charge forces within individual beams and the mutual space charge repulsion among the beams, which are strongest near the focal spot, prevent multiple beams from focusing onto a small spot. (b) Beam current increases from beam head to tail, so the beam-beam repulsion is time dependent. This prevents multiple beams from being focused onto a small spot throughout the entire pulse. (c) Radiation from the beam heated target photoionizes some beam ions into higher charge states, and these ions are expelled from the beam by the space charge force. To minimize these three effects, it is proposed to place a metallic cylinder in front of the focal spot and to cover the cylinder, at the beam entrance end, with a plastic film of submicron thickness. As the beams pass through the film, the beam space charge force draws electrons from the film to co-move with the beam, `autoneutralizing' the ion beams inside the cylinder. Although the film strips beam ions into higher charge states, particle-in-cell simulations show that autoneutralization is so effective that the residual radial electric field in the neutralized beam is considerably less than that in the unneutralized beam. The application of this scheme thereby results in a great improvement in focusing quality. Beams with higher charge state or lower mass ions, which have the advantages of reducing both the length of the high energy section of induction linacs and the strengths of the final-focusing magnets, can therefore be used

Summary form only given, as follows. The discrete surface integral (DSI) algorithm for solving the Maxwell curl equations in the time domain provides the opportunity to accurately discretize extremely complex geometries. This algorithm is a direct generalization of the standard staggered-grid finite-difference approach using a 3d, non-orthogonal, unstructured grid composed of mixed-polyhedral elements. Little is known about the numerical properties of the DSI method when discretized on these more general grids. However, dispersion characteristics can be determined for idealized non-orthogonal grids which provide some insight on the behavior of the algorithm on more general grids. Results of the dispersion and stability analysis for the DSI algorithm when discretized on a 2d chevron grid are presented. This analysis shows that, for chevron grids, the DSI algorithm supports slowly growing electromagnetic oscillations. Unlike the usual Courant instability, these oscillations have complex frequencies with nonzero growth rates for a vanishing time step. Several numerical examples of the chevron instability, along with some comments about the possible impact of this instability on real problems are presented.

The development of time domain electromagnetic solvers for nonorthogonal grids is an area of current research interest, stemming from the need to simulate complex geometries in a wide variety of applications. A notable example is the discrete surface integral (DSI) algorithm which solves the Maxwell curl equations in the time domain using a 3d, unstructured, mixed-polyhedral grid. Although this method is an extension of the time proven Yee algorithm, little is known about the numerical properties of the method when discretized on these more general grids. Dispersion relations for the DSI algorithm can be derived using 2d idealized grids, such as the skewed mesh analysis done by Ray and Rambo for both triangles and quadrilaterals. The present work applies the same techniques used for the skewed mesh analysis to another idealized, but nonorthogonal, 2d grid.

The authors are continuing to develop electromagnetic plasma simulation techniques applicable to geometrically complex domains discretized by a non-orthogonal, unstructured, mixed-polyhedral mesh. Maxwell`s curl equations are advanced in time using the recently developed discrete surface integral (DSI) method. This algorithm preserves, in the absence of particles, the electric and magnetic field divergence constraints and is a direct generalization of the standard staggered-grid finite-difference method. The plasma is represented by particles which move through the unstructured grid obeying all of the normal particle-in-cell approximations. They show details of the DSI field algorithm and of the method for tracking and weighting particles on the distorted mesh elements. Several numerical examples will be presented which demonstrate the present capabilities to integrate the DSI and PIC algorithms. Preliminary calculations which include particle field emission and wave boundary conditions will be presented.

Summary form only given. There is much current interest in developing plasma simulation techniques able to accurately model the geometrical complexity of laboratory plasma devices. The discrete surface integral (DSI) method for solving Maxwell's equations in the time-domain presents an opportunity to achieve the desired geometrical realism. This method allows the use of general nonorthogonal mixed-polyhedral unstructured grids that are direct generalizations of the staggered-grid finite-difference method. The method is conservative in that it locally preserves divergence or charge. An effort has been made to include particles in the 3D electromagnetic field code DS13D.

Summary form only given. A version of the two-dimensional plasma simulation code Condor, has been retrofitted to use finite volumes as an option as well as standard finite differences to obtain the field solution. The finite volume solution takes place on a logically rectangular domain of quadrilateral cells. By allowing voids. the mesh can be distorted to fit a wide range of geometrically complex features