Fast ion research is necessary to understand confinement of high energy particles in fusion devices like DIII-D. Measuring the fast ion distribution is necessary to understand how well it is confined. The fast ion D-alpha (FIDA) diagnostic and neutral particle analyzer (NPA) are two diagnostics that measure part of the distribution. Upgrades to the FIDA diagnostic are necessary to resolve scattering issues from the unshifted D-alpha emission that contaminates other parts of the measured spectra. Effective filter performance is possible using the reflection of a narrow bandpass filter. The imaging neutral particle analyzer (INPA) was developed on DIII-D to provide greater spatial and energy resolution compared to traditional NPAs. Measured images are validated using simulations. Robust passive signals are also observed and found to come from fast ion populations traversing the edge. Fast ions can also interact with plasma waves or modes to transfer energy which can lead to losses. The energetic particle-induced geodesic acoustic mode (EGAM) is one plasma mode that is excited by fast ions. Various phenomena associated with EGAMs are characterized to assess the impact it has on fast ions.

Diagnosis of the fast ion population, which is necessary for heating but can drive dangerous instabilities, is crucial in achieving and maintaining a burning plasma regime. Present fast ion diagnostic methods tend to rely on delicate components unlikely to survive in a reactor environment, motivating exploration into alternative diagnostic methods. One candidate, the subject of this thesis, entails measurement of collective ion cyclotron emission (ICE) spectra via robust magnetic pickup loops integrated with the first wall, making it potentially compatible with ITER and other reactor-relevant devices [116, 69]. ICE, distinct from thermal electron cyclotron emission, manifests as narrowband peaks at harmonics of the ion cyclotron frequency $f_{ci}$ and is generally localized near the core or the edge of the plasma. These modes have been observed to be sensitive to the fast ion population, be they introduced by auxiliary heating systems or fusion products. However, the connection between observed ICE spectra and the fast ion population has not been sufficiently established, meriting diagnostic upgrades and dedicated experiments.

This work aims to characterize ICE mode structure and fine-scale phenomena in L- and H-mode plasmas on DIII-D using the ICE diagnostic, which was recently upgraded to include additional channels and consequently new measurement capabilities. In addition to basic frequency and amplitude information, measurements of mode polarization at the plasma edge, amplitude as measured at the centerpost and outer wall of the machine, and toroidal mode number are now possible and will be demonstrated through this work. Additionally, mode sensitivity to the vacuum region, plasma shape, and particularly the fast ion distribution is explored. The detailed mode structure measurements and phenomenology presented in this work can be used to test and validate theoretical models en route to developing a more complete physical understanding of ICE, so that it can be leveraged as a passive fast ion diagnostic in future burning plasmas.

All the information about a plasma species is encoded in its distribution function. While it would be helpful to measure the distribution function directly, it is only possible to measure its moments. If the form of the distribution function is not known a priori, it can be difficult to interpret diagnostic signals. This is particularly true in fast-ion physics, where diagnostics that nominally view the same thing, the fast-ion distribution function, give seemingly discordant measurements. The process of going from a fast-ion distribution to a measurement and the reverse process of going from a set of measurements to a fast-ion distribution are the main topics of this thesis.

Chapters 2-3 concern the modeling of fast-ion diagnostics. Here we derive functions that translate the information about a fast ion into measurable quantities, i.e. forward models. This is done for three diagnostics: the neutral particle analyzer (NPA), fast ion D-α (FIDA) spectroscopy, and neutron scintillators. Chapter 3 discusses the development of FIDASIM, the practical implementation of the forward models.

Chapter 4 deals with diagnostic velocity-space weight functions, an ansatz which is used to aid in the interpretation of experimental measurements and as an approximate forward model of the diagnostic. From the forward models discussed in Chapter 2, we derive weight functions in a full 6D generalized coordinate system, from which we also derive the velocity-space weight functions. Using an action-angle formulation, orbit-space weight functions, which can be used to exactly represent a diagnostic's forward model in a linear form, are derived.

Chapters 5-6 detail how to use weight functions to infer the fast-ion distribution function from experimental measurements. Orbit weight functions, in particular, facilitate the inference of the entire distribution function, using any fast-ion diagnostic that views the plasma. Benchmarks with synthetic data and a reconstruction of a classical distribution from experimental measurements, show that systematic errors and intrinsic biases in the inference methods are the main impediments to accurately inferring the fast-ion distribution function. However, experimental studies of the redistribution of fast ions by sawtooth crashes at ASDEX Upgrade demonstrate that the effects of systematic error and biases become less important when only considering relative changes in the distribution function.

Fusion energy research is working toward the goal of using fusion as a source of clean energy. Since energetic particles are the main source of heating in current and future devices, understanding fast ion transport and loss mechanisms is vital. Experimental research in this thesis uses two main fast ion diagnostics, the fast ion loss detector and the imaging neutral particle analyzer, to discern how energetic particles are transported in the presence of various magnetic perturbations. First, fast ion losses in the presence of externally applied, 3D magnetic fields are analyzed. Second, transport of confined particles and simultaneous losses are observed and discussed in the presence of a neoclassical tearing mode. Lastly, work on intermittent fast ion losses in the presence of multiple Alfven Eigenmodes is observed.

Spectra of the Balmer-alpha radiation of first-orbit fast ions after charge exchange

with edge neutrals have been measured in the DIII-D tokamak. Several collimated

optics systems view the edge region--while avoiding any active beams--and carry

light to a spectrometer tuned to the region of the 656.1 nm deuterium-alpha line.

Viewing geometry and the high energy of the lost ions produce Doppler shifts, which

effectively separate the fast-ion contributions from the bright, cold edge light. Modulation of the fast-ion source allows for time-evolving background subtraction. A model

has been developed for the spectra of these first-orbit fast ions. The passive fast-ion

D-alpha simulation (P-FIDAsim) is a forward model consisting of an experimentally-

validated beam model, an ion orbit-following code, a collisional-radiative model, and

a synthetic spectrometer. Eighty-six experimental spectra were obtained using 6 different neutral beam fast-ion sources and 13 different viewing chords. Parameters such

as plasma current, toroidal field, electron density, plasma cross-sectional shape, and

number of x-points were varied. Uncalibrated experimental spectra have an overall

Spearman rank correlation coefficient with the shape of simulated spectra of 0.58 with

subsets of cases rising to a correlation of 0.80. A single set of calibrated spectra (shot 152817) was measured and is used to estimate the neutral density throughout the

cross-section of the tokamak. This is done by inverting the simulated spectra in order

to nd the best neutral density (in a least squares sense) required to best match the

experimental spectra. The resulting 2D neutral density shows the expected increase

toward each x-point. The average neutral density is found to be 3.3x10^5cm^-􀀀3 at

the magnetic axis, 2.3x10^8cm^-􀀀3 in the core, 8.1x10^9cm^-􀀀3 at the plasma boundary, and 1.1x10^11cm^-􀀀3 near the wall. A technique is developed which--after using

first-orbit light to calibrate the system--can quantify losses from a wider variety of

mechanisms. Fast-ion losses resulting from sawtooth crashes (shot 149941) is estimated to eject 1.2% of the fast-ion inventory, in good agreement with a 1.7% loss

estimate made by the TRANSP code.

This thesis presents advances in computational modelling, analysis and techniques that can be used to study mission critical topics in fusion. Energetic particles is an important field of research since information about the plasma state can be encoded in the fast-ion distribution function. Energetic ions can resonate with plasma waves in a fusion device, degrade plasma performance or confinement, and damage the inner walls of the vessel. There is a need in the community to better understand the fast-ion distribution function, improve plasma performance, and mitigate unwanted impacts from wave-particle interactions.

In chapter 2, the weight function for 3-MeV protons produced in d(d,p)t fusion reactions between a fast ion and a thermal deuteron is developed. The weight function W(X) is a diagnostic sensitivity to phase-space variables X that relates the measured signal C to the distribution function F(X) through the equation C = integral[ W(X) F(X) dX ]. The algorithm developed here accounts for the complications associated with the curved "sightline" trajectories of the escaping protons. Time-reversed orbits are initially calculated to get the effective solid angles and sightlines for the range of incident proton velocity vectors. Synthetic diagnostic code FIDASIM [Plasma Phys. Cont. Fusion 62 (2020) 105008] is upgraded to accept these inputs, then calculate the reactivity averaged over the thermal distribution of the "target" deuterons and the probability that a fast ion of specified energy and pitch has a gyroangle that is consistent with the kinematic equations along each of the sightlines. The outputs of FIDADSIM are verified using independent calculations on the Mega Amp Spherical Tokamak.

In chapter 3, a conceptual design to diagnose the lifetime of spin polarized fusion experiments using existing port geometries is developed at DIII-D. The cross sections for the D-T and D-3He fusion reactions are increased by as much as 50% if the fuel remains spin polarized parallel to the magnetic field in magnetically confined fusion experiments. The goal in this chapter is to assess the feasibility of lifetime measurements of spin polarization, in magnetic fusion relevant conditions, on the DIII-D tokamak using relative changes in charged fusion product (CFP) loss measurements that depend upon the differential fusion cross section. Relative measurements that capture changes in the escaping CFP pitch, poloidal, and energy distributions are studied in two realistic TRANSP calculated plasma scenarios (high Ti and beam-plasma).Ideal CFP detection, a realistic assessment of CFP signals and reduced chi-squared calculations show polarization lifetime measurements are feasible for the thermonuclear (high Ti) scenario.

In chapter 4, Machine Learning (ML) models are developed to automatically detect Alfvén eigenmodes (AE) and these models achieve high performance (True Positive Rate = 90% and False Positive Rate = 14%).ML-based models can be useful for real-time detection and control of AEs in steady-state plasma scenarios. These ML systems can be implemented into control algorithms that drive actuators for mitigation of unwanted AE impacts. Using labels created from a curated database [Heidbrink, et al., NF ‘20], Machine Learning-based systems are trained using single chord and crosspower spectrograms to predict the presence of 5 AEs (EAE, TAE, RSAE, BAE and LFM). The advantages of using the CO2 interferometer to detect AEs, and the results from a comparison between inputs (single chord and crosspower spectrograms) and another comparison between two different ML models (Reservoir Computing Network and Long Short-Term Memory Network) are covered here.

The transport of fast ions in magnetic flux ropes in a laboratory plasma is studied. Strong perturbing flux ropes (δE ~175 V/m, δB ~7 G) are generated by secondary cathode-anode pair at the upgraded LArge Plasma Device (LAPD). A 500-1000 eV lithium ion test beam is passed through the turbulent region and recollected by a gridded collimated analyzer, revealing enhanced fast ion broadening attributable to flux rope perturbations. The broadening is observed to be well in excess of Coulomb scattering levels. Monte Carlo simulation is performed with model electrostatic and magnetic fields, revealing negligible spreading as a result of the magnetic perturbations. Modeled electrostatic perturbations are observed to broaden the beam by 3.0 mm2 at the closest recollection plane, increasing as the transit time squared further downstream. Transport attributed to electrostatic fluctuations has been shown to decrease with energy while magnetic transport does not. Enhanced fast ion transport observed during the flux rope off phase is presently unexplained.