Overcoming the demagnetization of superconducting linear Halbach array

Combining trapped-field superconducting magnets with mutually orthogonal magnetization directions in a linear Halbach array configuration offers the prospect of generating both high fields and large field gradients. For a Halbach array made of 3 magnetized bulk large grain melt-textured YBa2Cu3O7-δ superconductors of given size and critical current density, it was previously shown [1] that the main limitation of the field generated by the array arises from the modification of the current density distribution within the superconductors during the assembly process.
In this work, two methods are explored to mitigate the impact of such demagnetization on the field generated by the array. In the first method, an additional bulk superconductor is placed above the central one, magnetized simultaneously with it and maintained in place while approaching the left and right peripheral samples to form the array. The additional sample is then removed from the structure, which induces a re-organisation of the supercurrents in the central superconductor. In the second method, the geometric shape of the peripheral samples is modified to reduce the field experienced by the sides of the central superconductor. Starting from cubic superconductors, the samples are cut at 45° with respect their c-axis, which gives samples having a square cross-section parallel to the ab-planes and a triangular cross-section parallel to the c-axis. Experiments with such arrays are carried out at 77 K with bulk Y1Ba2Cu3O7-δ superconductors of 13 mm side.
The spatial distribution of B ⃗ above the assembly is measured with a recently developed cryogenic 3-axis Hall sensor [2] and is compared to analytical calculations and finite-element simulations. This set of results makes it possible to understand in detail the advantage brought by the two methods mentioned above. A noticeable increase of the flux density generated above the array is found for both methods (~5% with samples used in experiments). For the first method, finite-element simulations show that this increase gets higher when using a taller additional sample, which allows to almost recover the full potential of the array.