Coronal mass ejections are solar eruptions driven by a sudden release of magnetic energy stored in the Sun’s corona. In many cases, this magnetic energy is stored in long-lived, arched structures called magnetic flux ropes. When a flux rope destabilizes, it can either erupt and produce a coronal mass ejection or fail and collapse back towards the Sun. The prevailing belief is that the outcome of a given event is determined by a magnetohydrodynamic force imbalance called the torus instability. This belief is challenged, however, by observations indicating that torus-unstable flux ropes sometimes fail to erupt. This contradiction has not yet been resolved because of a lack of coronal magnetic field measurements and the limitations of idealized numerical modelling. Here we report the results of a laboratory experiment that reveal a previously unknown eruption criterion below which torus-unstable flux ropes fail to erupt. We find that such ‘failed torus’ events occur when the guide magnetic field (that is, the ambient field that runs toroidally along the flux rope) is strong enough to prevent the flux rope from kinking. Under these conditions, the guide field interacts with electric currents in the flux rope to produce a dynamic toroidal field tension force that halts the eruption. This magnetic tension force is missing from existing eruption models, which is why such models cannot explain or predict failed torus events.
We implement unsupervised machine learning techniques to identify characteristic evolution patterns and associated parameter regimes in edge localized mode (ELM) events observed on the National Spherical Torus Experiment. Multi-channel, localized measurements spanning the pedestal region capture the complex evolution patterns of ELM events on Alfven timescales. Some ELM events are active for less than 100~microsec, but others persist for up to 1~ms. Also, some ELM events exhibit a single dominant perturbation, but others are oscillatory. Clustering calculations with time-series similarity metrics indicate the ELM database contains at least two and possibly three groups of ELMs with similar evolution patterns. The identified ELM groups trigger similar stored energy loss, but the groups occupy distinct parameter regimes for ELM-relevant quantities like plasma current, triangularity, and pedestal height. Notably, the pedestal electron pressure gradient is not an effective parameter for distinguishing the ELM groups, but the ELM groups segregate in terms of electron density gradient and electron temperature gradient. The ELM evolution patterns and corresponding parameter regimes can shape the formulation or validation of nonlinear ELM models. Finally, the techniques and results demonstrate an application of unsupervised machine learning at a data-rich fusion facility.
Stotler, D.; F. Scotti; R.E. Bell; A. Diallo; B.P. LeBlanc; M. Podesta; A.L. Roquemore; P.W. Ross
Atomic and molecular density data in the outer midplane of NSTX [Ono et al., Nucl. Fusion 40, 557 (2000)] are inferred from tangential camera data via a forward modeling procedure using the DEGAS 2 Monte Carlo neutral transport code. The observed Balmer-b light emission data from 17 shots during the 2010 NSTX campaign display no obvious trends with discharge parameters such as the divertor Balmer-a emission level or edge deuterium ion density. Simulations of 12 time slices in 7 of these discharges produce molecular densities near the vacuum vessel wall of 2–8 10^17 m3 and atomic densities ranging from 1 to 7 10^16 m3; neither has a clear correlation with other parameters. Validation of the technique, begun in an earlier publication, is continued with an assessment of the sensitivity of the simulated camera image and neutral densities to uncertainties in the data input to the model. The simulated camera image is sensitive to the plasma profiles and virtually nothing else. The neutral densities at the vessel wall depend most strongly on the spatial distribution of the source; simulations with a localized neutral source yield densities within a factor of two of the baseline, uniform source, case. The uncertainties in the neutral densities associated with other model inputs and assumptions are 50%.