Hosing of a Long Relativistic Particle Bunch in Plasma.

Experimental results show that hosing of a long particle bunch in plasma can be induced by wakefields driven by a short, misaligned preceding bunch. Hosing develops in the plane of misalignment, self-modulation in the perpendicular plane, at frequencies close to the plasma electron frequency, and are reproducible. Development of hosing depends on misalignment direction, its growth on misalignment extent and on proton bunch charge. Results have the main characteristics of a theoretical model, are relevant to other plasma-based accelerators and represent the first characterization of hosing.

Vertical electron cyclotron emission diagnostic on the tokamak à configuration variable.

In this paper, a diagnostic for the measurement of the electron cyclotron emission (ECE) from non-thermal electrons in magnetically confined fusion plasmas is presented. The diagnostic employs a vertical viewing line of sight that allows us to directly infer the energy of the emitting electrons. Previous incarnations of this diagnostic on other machines have been limited by refraction, which can cause stray radiation to enter the line of sight, polluting the signal. By tuning the toroidal magnetic field on the Tokamak à Configuration Variable, TCV, and by carefully selecting the range of frequencies that are used to measure the ECE spectrum, refraction can be mitigated and background radiation power reduced to below the noise power of the instrumentation. A novel technique for calibrating the diagnostic based on plasma measurements and modeling has been developed. The paper will describe the Vertical ECE (V-ECE) diagnostic on TCV, the calibration method, and the first results from the measurements.

An analog-AI chip for energy-efficient speech recognition and transcription.

Models of artificial intelligence (AI) that have billions of parameters can achieve high accuracy across a range of tasks1,2, but they exacerbate the poor energy efficiency of conventional general-purpose processors, such as graphics processing units or central processing units. Analog in-memory computing (analog-AI)3-7 can provide better energy efficiency by performing matrix-vector multiplications in parallel on 'memory tiles'. However, analog-AI has yet to demonstrate software-equivalent (SWeq) accuracy on models that require many such tiles and efficient communication of neural-network activations between the tiles. Here we present an analog-AI chip that combines 35 million phase-change memory devices across 34 tiles, massively parallel inter-tile communication and analog, low-power peripheral circuitry that can achieve up to 12.4 tera-operations per second per watt (TOPS/W) chip-sustained performance. We demonstrate fully end-to-end SWeq accuracy for a small keyword-spotting network and near-SWeq accuracy on the much larger MLPerf8 recurrent neural-network transducer (RNNT), with more than 45 million weights mapped onto more than 140 million phase-change memory devices across five chips.

Controlled Growth of the Self-Modulation of a Relativistic Proton Bunch in Plasma.

A long, narrow, relativistic charged particle bunch propagating in plasma is subject to the self-modulation (SM) instability. We show that SM of a proton bunch can be seeded by the wakefields driven by a preceding electron bunch. SM timing reproducibility and control are at the level of a small fraction of the modulation period. With this seeding method, we independently control the amplitude of the seed wakefields with the charge of the electron bunch and the growth rate of SM with the charge of the proton bunch. Seeding leads to larger growth of the wakefields than in the instability case.

Determining 1D fast-ion velocity distribution functions from ion cyclotron emission data using deep neural networks.

The relationship between simulated ion cyclotron emission (ICE) signals s and the corresponding 1D velocity distribution function fv⊥ of the fast ions triggering the ICE is modeled using a two-layer deep neural network. The network architecture (number of layers and number of computational nodes in each layer) and hyperparameters (learning rate and number of learning iterations) are fine-tuned using a bottom-up approach based on cross-validation. Thus, the optimal mapping gs;θ of the neural network in terms of the number of nodes, the number of layers, and the values of the hyperparameters, where θ is the learned model parameters, is determined by comparing many different configurations of the network on the same training and test set and choosing the best one based on its average test error. The training and test sets are generated by computing random ICE velocity distribution functions f and their corresponding ICE signals s by modeling the relationship as the linear matrix equation Wf = s. The simulated ICE signals are modeled as edge ICE signals at LHD. The network predictions for f based on ICE signals s are on many simulated ICE signal examples closer to the true velocity distribution function than that obtained by 0th-order Tikhonov regularization, although there might be qualitative differences in which features one technique is better at predicting than the other. Additionally, the network computations are much faster. Adapted versions of the network can be applied to future experimental ICE data to infer fast-ion velocity distribution functions.

Transition between Instability and Seeded Self-Modulation of a Relativistic Particle Bunch in Plasma.

We use a relativistic ionization front to provide various initial transverse wakefield amplitudes for the self-modulation of a long proton bunch in plasma. We show experimentally that, with sufficient initial amplitude [≥(4.1±0.4) MV/m], the phase of the modulation along the bunch is reproducible from event to event, with 3%-7% (of 2π) rms variations all along the bunch. The phase is not reproducible for lower initial amplitudes. We observe the transition between these two regimes. Phase reproducibility is essential for deterministic external injection of particles to be accelerated.

Persistent random walks of charged particles across magnetic field lines.

We investigate the time evolution of the mean location and variance of a charged particle subject to random collisions that are Poisson distributed. The particle moves on a plane and is subject to a magnetic field applied perpendicular to the plane, so it is constrained to move in circles in the absence of collisions. We develop a procedure that yields analytic expressions of the mean and variance. These results are valid for arbitrary times after the start of the walk, including early on when, on average, less than one collision is expected. As an example of their applicability, we use these expressions to model experimental results and simulations of suprathermal ions propagating in a turbulent plasma in TORPEX (the TORoidal Plasma EXperiment).

Cathodoluminescent screen imaging system for seeded blob detection in toroidal plasma experiment.

We designed and built a diagnostic based on a cathodoluminescent screen for the detection of turbulent plasma structures with high spatial resolution. The screen is coated with a low threshold energy cathodoluminescent powder that emits light when exposed to a plasma. The emitted light is imaged with a fast frame camera combined with an image intensifier and an optical bandpass filter. The diagnostic is used to study turbulent structures and seeded blobs. The results are analyzed with pattern recognition algorithms to track the turbulent structures and study their evolution in time.

Proton Bunch Self-Modulation in Plasma with Density Gradient.

We study experimentally the effect of linear plasma density gradients on the self-modulation of a 400 GeV proton bunch. Results show that a positive or negative gradient increases or decreases the number of microbunches and the relative charge per microbunch observed after 10 m of plasma. The measured modulation frequency also increases or decreases. With the largest positive gradient we observe two frequencies in the modulation power spectrum. Results are consistent with changes in wakefields' phase velocity due to plasma density gradients adding to the slow wakefields' phase velocity during self-modulation growth predicted by linear theory.

Truncated Lévy motion through path integrals and applications to nondiffusive suprathermal ion transport.

Fractional Levy motion has been derived from its generalized Langevin equation via path integrals in earlier works and has since proven to be a useful model for nonlocal and non-Markovian processes, especially in the context of nondiffusive transport. Here, we generalize the approach to treat tempered Lévy distributions and derive the propagator and diffusion equation of truncated asymmetrical fractional Levy motion via path integrals. The model now recovers exponentially tempered tails above a chosen scale in the propagator, and therefore finite moments at all orders. Concise analytical expressions for its variance, skewness, and kurtosis are derived as a function of time. We then illustrate the versatility of this model by applying it to simulations of the turbulent transport of fast ions in the TORPEX basic plasma device.

Particle transport at arbitrary timescales with Poisson-distributed collisions.

We develop a model to investigate the time evolution of the mean location and variance of a random walker subject to Poisson-distributed collisions at constant rate. The collisions are instantaneous velocity changes where a new value of velocity is generated from a model probability function. The walker is persistent, which means that it moves at constant velocity between collisions. We study three different cases of velocity transition functions and compute the transport properties from the evolution of the variance. We observe that transport can change character over time and that early times show features that, in general, depend on the initial conditions of the walker.

Time intermittency in nondiffusive transport regimes of suprathermal ions in turbulent plasmas.

Intermittent phenomena have long been studied in the context of nondiffusive transport across a variety of fields. In the TORPEX device, the cross-field spreading of an injected suprathermal ion beam by electrostatic plasma turbulence can access different nondiffusive transport regimes. A comprehensive set of suprathermal ion time series has been acquired, and time intermittency quantified by their skewness. Values distinctly above background level are found across all observed transport regimes. Intermittency tends to increase toward quasi- and superdiffusion and for longer propagation times of the suprathermal ions. The specific prevalence of intermittency is determined by the meandering motion of the instantaneous ion beam. We demonstrate the effectiveness of an analytical model developed to predict local intermittency from the time-average beam. This model might thus be of direct interest for similar systems, e.g., in beam physics, or meandering flux-rope models for solar energetic particle propagation. More generally, it illustrates the importance of identifying the system-specific sources of time-intermittent behavior when analyzing nondiffusive transport.

Dual Langmuir-probe array for 3D plasma studies in TORPEX.

We have designed and installed a new Langmuir-probe (LP) array diagnostic to determine basic three-dimensional (3D) features of plasmas in TORPEX. The diagnostic consists of two identical LP arrays, placed on opposite sides of the apparatus, which provide comprehensive coverage of the poloidal cross section at the two different toroidal locations. Cross correlation studies of signals from the arrays provide a basic way to extract 3D information from the plasmas, as experiments show. Moreover, the remarkable signal-to-noise performance of the front-end electronics allows us to follow a different approach in which we combine information from all probes in both arrays to reconstruct elementary 3D plasma structures at each acquisition time step. Then, through data analysis, we track the structures as they evolve in time. The LP arrays include a linear-motion mechanism that can displace radially the probes located on the low field side for experiments that require fine-tuning of the probe locations, and for operational compatibility with the recently installed in-vessel toroidal conductor.

X-Point Effect on Plasma Blob Dynamics.

Plasma blob dynamics on the high-field side in the proximity of a magnetic field null (X point) is investigated in TORPEX. A significant acceleration of the blobs towards the X point is observed. Close to the X point the blobs break apart. The E×B drifts associated with the blobs are measured, isolating the background drift component from the fluctuating contribution of the blob internal potential dipole. The time evolution of the latter is consistent with the fast blob dynamics. An analytical model based on charge conservation is derived for the potential dipole, including ion polarization, diamagnetic, and parallel currents. In the vicinity of the X point, a crucial role in determining the blob motion is played by the decrease of the poloidal magnetic field intensity. This variation increases the connection length that short circuits the potential dipole of the blob. Good quantitative agreement is found between the model and the experimental data in the initial accelerating phase of the blob dynamics.

Nondiffusive transport regimes for suprathermal ions in turbulent plasmas.

The understanding of the transport of suprathermal ions in the presence of turbulence is important for fusion plasmas in the burning regime that will characterize reactors, and for space plasmas to understand the physics of particle acceleration. Here, three-dimensional measurements of a suprathermal ion beam in the toroidal plasma device TORPEX are presented. These measurements demonstrate, in a turbulent plasma, the existence of subdiffusive and superdiffusive transport of suprathermal ions, depending on their energy. This result stems from the unprecedented combination of uniquely resolved measurements and first-principles numerical simulations that reveal the mechanisms responsible for the nondiffusive transport. The transport regime is determined by the interaction of the suprathermal ion orbits with the turbulent plasma dynamics, and is strongly affected by the ratio of the suprathermal ion energy to the background plasma temperature.

Time-Resolved Measurements of Suprathermal Ion Transport Induced by Intermittent Plasma Blob Filaments.

Suprathermal ion turbulent transport in magnetized plasmas is generally nondiffusive, ranging from subdiffusive to superdiffusive depending on the interplay of the turbulent structures and the suprathermal ion orbits. Here, we present time-resolved measurements of the cross-field suprathermal ion transport in a toroidal magnetized turbulent plasma. Measurements in the superdiffusive regime are characterized by a higher intermittency than in the subdiffusive regime. Using conditional averaging, we show that, when the transport is superdiffusive, suprathermal ions are transported by intermittent field-elongated turbulent structures that are radially propagating.

The new TORPEX in-vessel toroidal conductor for the generation of a poloidal magnetic field.

TORoidal Plasma EXperiment (TORPEX) is a Simple Magnetized Torus featuring open helical magnetic field lines obtained from the superposition of a small vertical component on the main toroidal field. This work introduces the experimental setup developed to include a poloidal magnetic field. The toroidal and poloidal fields generate a rotational transform, making the magnetic geometry of TORPEX closer to that of a tokamak. This upgrade opens the possibility to deal with closed and open flux surfaces, as well as with the transition region across the last closed flux surface. The main technical solutions are discussed together with the physical considerations at the basis of the system design. Selected examples of the magnetic configurations accessible with the set of magnetic field coils available on TORPEX are discussed, ranging from single-null X-points to magnetic snowflakes. The simplest magnetic configuration of quasi-circular concentric flux surfaces is tested experimentally. Measurements of the two-dimensional electron plasma density profiles and the particle confinement time are presented, together with the first steps towards the understanding of plasma production mechanisms.

Nondiffusive suprathermal ion transport in simple magnetized toroidal plasmas.

We investigate suprathermal ion dynamics in simple magnetized toroidal plasmas in the presence of electrostatic turbulence driven by the ideal interchange instability. Turbulent fields from fluid simulations are used in the nonrelativistic equation of ion motion to compute suprathermal tracer ion trajectories. Suprathermal ion dispersion starts with a brief ballistic phase, during which particles do not interact with the plasma, followed by a turbulence interaction phase. In this one simple system, we observe the entire spectrum of suprathermal ion dynamics, from subdiffusion to superdiffusion, depending on beam energy and turbulence amplitude. We estimate the duration of the ballistic phase and identify basic mechanisms during the interaction phase that determine the dependencies of the character of suprathermal ion dispersion upon the beam energy and turbulence fluctuation amplitude.

Convective cells and blob control in a simple magnetized plasma.

Blob control by creating convective cells using biased electrodes is demonstrated in simple magnetized toroidal plasmas. A two-dimensional array of electrodes is installed on a metal limiter to obtain different biasing schemes. Detailed two-dimensional measurements across the magnetic field reveal the formation of a convective cell, which shows a high degree of uniformity along the magnetic field. Depending on the biasing scheme, radial and vertical blob velocities can be varied significantly. A high level of cross-field currents limits the achievable potential variations to values well below the applied bias voltage. Furthermore, the strongest potential variations are not induced along the biased flux tube, but at a position shifted in the direction of plasma flows.