Massively-Parallel Magnet Design from a Web Browser

Next-generation light sources will rely on closely arranged sequences of high-field magnets to achieve ambitious beam parameters and high brightness. Codes available for 3D modeling and design of magnetic systems are inadequate for capturing the overlapping fields from these dense arrays of magnets. Moreover, these codes suffer from poor parallel performance, leading to long turn-around times that significantly hamper the process of design and optimization. Significant improvements to existing software are needed to address these shortcomings. We will enhance a well-known, open-source 3D magnetostatics code to operate on massively parallel architectures using state-of-the-art mathematical libraries. We will build a browser-based graphical user interface atop our newly developed Python interface. By the end of Phase II, scientists will be simulating and optimizing—on a time-scale of minutes rather than hours—the high-resolution designs needed for detailed analyses of next-generation magnetic devices. We prototyped a Python interface that can be used in either a script or online notebook. We parallelized that interface to enable rapid field evaluation. And we profiled the code and established a plan for parallelizing the core algorithms at the C++ level. We will complete the Python interface and implement a graphical user interface that makes the code easy to use, facilitates optimization, and provides an advanced geometry interface with 3D visualization. We will also parallelize compute-intensive parts of the code and enhance the underlying physics engine. We will use the new software to simulate new magnets for next-generation light sources. Commercial Applications and Other Benefits: The proposed software will significantly lower the computational cost of magnet design and optimization, and will lead to fundamental improvements in this field. We will use the expertise we develop to provide R&D consulting services to the global high-field and permanent magnet industry, which has a market size in excess of $14B. Our browser- based, open-source scientific application framework differentiates us from competitors offering expensive, difficult-to-learn magnetostatic design modules.