Search for: All records

The trapped residual magnetic flux during the cool-down due to the incomplete Meissner state is a significant source of radio frequency losses in superconducting radio frequency cavities. Here, we clearly correlate the niobium microstructure in elliptical cavity geometry and flux expulsion behavior. In particular, a traditionally fabricated Nb cavity half-cell from an annealed poly-crystalline Nb sheet after an 800 C heat treatment leads to a bi-modal microstructure that ties in with flux trapping and inefficient flux expulsion. This non-uniform microstructure is related to varying strain profiles along the cavity shape. A novel approach to prevent this non-uniform microstructure is presented by fabricating a 1.3 GHz single cell Nb cavity with a cold-worked sheet and subsequent heat treatment leading to better flux expulsion after 800 ^∘C/3 h. Microstructural evolution by electron backscattered diffraction-orientation imaging microscopy on cavity cutouts, and flux pinning behavior by dc-magnetization on coupon samples confirms a reduction in flux pinning centers with increased heat treatment temperature. The heat treatment temperature-dependent mechanical properties and thermal conductivity are reported. The significant impact of cold work in this study demonstrates clear evidence for the importance of the microstructure required for high-performance superconducting cavities with reduced losses caused by magnetic flux trapping. 

Incorrectly labeled examples, or label noise, is common in real-world computer vision datasets. While the impact of label noise on learning in deep neural networks has been studied in prior work, these studies have exclusively focused on homogeneous label noise, i.e., the degree of label noise is the same across all categories. However, in the real-world, label noise is often heterogeneous, with some categories being affected to a greater extent than others. Here, we address this gap in the literature. We hypothesized that heterogeneous label noise would only affect the classes that had label noise unless there was transfer from those classes to the classes without label noise. To test this hypothesis, we designed a series of computer vision studies using MNIST, CIFAR-10, CIFAR-100, and MS-COCO where we imposed heterogeneous label noise during the training of multi-class, multi-task, and multi-label systems. Our results provide evidence in support of our hypothesis: label noise only affects the class affected by it unless there is transfer.