The study and computation of nonlinear charged-particle transfer maps is fundamental to understanding single-particle beam dynamics in accelerator devices. Transfer maps for individual elements of the beamline can in general depend sensitively on nonlinear fringe-field and high-multipole effects. The inclusion of these effects requires a detailed and realistic model of the interior and fringe magnetic fields, including knowledge of high spatial derivatives. Current methods for computing such maps often rely on idealized models of beamline elements. This Dissertation describes the development and implementation of a collection of techniques for computing realistic (as opposed to idealized) charged-particle transfer maps for general beamline elements, together with corresponding estimates of numerical error.

Each of these techniques makes use of 3-dimensional measured or numerical field data on a grid as provided, for example, by various 3-dimensional finite element field codes. The required high derivatives of the corresponding vector potential A, required to compute transfer maps, cannot be reliably computed directly from this data by numerical differentiation due to numerical noise whose effect becomes progressively worse with the order of derivative desired. The effect of this noise, and its amplification by numerical differentiation, can be overcome by fitting on a bounding surface far from the axis and then interpolating inward using the Maxwell equations. The key ingredients are the use of surface data and the smoothing property of the inverse Laplacian operator.

We explore the advantages of map computation using realistic field data on surfaces of various geometry. Maps obtained using these techniques can then be used to compute realistically all derived linear and nonlinear properties of both single pass and circular machines. Although the methods of this Dissertation have been applied primarily to magnetic beamline elements, they can also be applied to electric and radio-frequency beamline elements.