Browsing by Subject "Accelerator"
Now showing 1 - 3 of 3
- Results Per Page
- Sort Options
Item Coherent optical diagnostics of laser-wakefield-accelerated electron bunches(2022-11-28) LaBerge, Maxwell; Downer, Michael Coffin; Lumpkin, Alex; Keto, John; Breizman, Boris; Ditmire, Todd; Schramm, UlrichThis dissertation presents a comprehensive study of coherent optical transition radiation (COTR) as a diagnostic for electron bunches from laser wakefield accelerators (LWFAs). In Chapter 2, I provide an intuitive description of the LWFA process, derive scaling laws, and describe injection regimes used in the experimental results presented later. In chapter 3, I present the theory of COTR with relevant approximations for the context of the experimental work shown later and in chapter 4 I present a COTR-imaging based algorithm to reconstruct portions of electron beams. Chapters 5, 6, 7, and 8 are published or soon to be published works using COTR diagnostics. The first section of each of these chapters is a "context of contribution," which summarizes my input as well as input from fellow coauthors. Chapter 5 presents a COTR interferometry diagnostic used to determine the emittance of a portion of an LWFA accelerated electron beam. It also sheds light on the level of microbunching present in beams from this acceleration method. Chapter 6 develops how multi-spectral COTR imaging allows for better characterization of an electron bunch. It goes on to present two-dimensional reconstructions of the coherent portion of the beam at several wavelengths for a single shot. Finally, it gives evidence that the source of the radiation is the charge population in quasi-monoenergetic peak of the electron spectrum as opposed to the low-energy background electrons or electrons accelerated by laser-foil interactions. Chapter 7 presents the differences between coherent and incoherent OTR for relativistic beams and demonstrates the high level (few percent) of microbunching from laser wakefield accelerated electron beams. This level of microbunching is shown to be injection regime dependent. Finally, chapter 7 further develops on how to glean spatial electron bunch information from multi-spectral COTR images, building up to a reproduction of features seen in particle in cell simulations as well as a candidate three-dimensional reconstruction of the coherent portion of an electron bunch. Chapter 8 is the culmination of the multi-spectral COTR imaging work. Here, I present analysis of COTR patterns from three different injection regimes, starting with phenomenology across many shots and then going in-depth to discuss what three-dimensional features account for the experimentally observed COTR patterns. Finally, chapter 9 provides a summary of this work and the outlook for future experiments.Item Improved understanding of the nature of photon scatter at a high energy radiographic imaging facility(2018-08-16) Morneau, Rachel; Haas, Derek Anderson, 1981-; Charlton, William; Landsberger, Sheldon; Byers, MargaretThe U.S. Stockpile Stewardship Program was designed to sustain and evaluate the nuclear weapons stockpile while foregoing underground nuclear tests and requires complex computer calculations. The Dual Axis Radiographic Hydrodynamic Test (DARHT) facility at Los Alamos National Laboratory performs hydrodynamic tests which mimic a nuclear weapon implosion. These implosions happen very quickly and involve very large areal masses, so high energy X-rays are necessary to successfully penetrate the hydrotest that in turn produce radiographs which are numerically analyzed using model fitting and tomographic reconstruction techniques to find material edges and density distributions. One of the areas that can be improved in these computational models is the modeling of scatter at the DARHT facility. The large areal masses present at the radiographic facility cause large amounts of scattered photons to be produced when the X-rays interact with the material, up to 50% of the direct signal in simulations and up to 200% of the direct signal in experiments. With such large amounts of scatter, it is imperative to model the scatter accurately in order to reconstruct density fields. Using static characterization objects, an improved understanding of the scatter field, particularly the Compton scatter field, is developed which can be applied to experimental data. This research will investigate aspects of scatter at the DARHT accelerator using MCNP, a particle transport code ideal for modeling complex systems. Detailed MCNP calculations provide scatter fluxes at specified locations in the DARHT beam line. These calculations will then be used to form a physicsbased reduced-order model of the scatter field. This model employs a kernel that can be convolved with the direct transmission to represent a component of scatter correlated to the direct signal, and an additive object/environmental scatter field that is uncorrelated. The physics-based model of the scatter field provides several benefits, the first eliminating the need for optimization of arbitrary high-order polynomials to simulate the scatter field in the density reconstruction as was done historically. It diminishes the need for continued MCNP calculations for minor changes in the DARHT configuration; it also prevents MCNP calculations from needing to be optimized in conjunction with the density reconstruction, which requires significant computational power and time.Item Scale-CNN : a tool for generating scalable high-throughput CNN inference accelerators on FPGAs(2021-04-30) Rauch, Daniel Levi; John, Lizy KurianIn the past decade, research has shown that CNN inference can be considerably sped up via dedicated hardware accelerators. However, most existing accelerators have limited performance by only working on a single inference at a time and/or relying on slow off-chip memory accesses for hidden layers. These limitations stem from the high memory requirements of CNN inference, which can be 10s of Mb even for small networks with reduction techniques. Despite this, as Moore's law has continued to scale, this level of on-chip memory is now attainable. We propose Scale-CNN, a tool for generating multiple Pareto optimal design points for high-throughput CNN inference accelerators on Xilinx Ultrascale+ FPGAs. The Scale-CNN architecture dedicates separate hardware resources for each layer and stores all feature maps and weights on-chip, enabling a high-throughput network pipeline where each layer works on a different inference simultaneously with no off-chip memory accesses. Using Scale-CNN, we generate several accelerator IPs for Tiny Darknet on the smallest Virtex Ultrascale+ FPGA (XCVU3P) that range from 1.7 to 56.7 inferences per second utilizing 22% to 66% of FPGA resources.