Maximizing critical currents in superconductors by optimization of normal inclusion properties (Q652388)

The paper investigates the pinning of vortices by embedding normal inclusions into a superconducting material with following determination of the critical current via an optimal control approach. The 2D time-dependent Ginzburg-Landau (TDGL) model of superconductors is used in the paper. After a suitable nondimensionalization, the dimensionless TDGL equations for the complex scalar-valued order parameter and real vector-valued magnetic potential are given together with the initial boundary conditions. As a result, the TDGL model for superconductivity which can account for normal inclusions is obtained. Then, by introducing a cost functional, the authors transform the problem to an optimization problem minimizing the cost functional with respect to some control parameters, that is, there are sought state variables and control variables such that the cost functional is minimized under the requirement that the variables satisfy the TDGL equations. The optimization problem stated is solved by using the gradient method, where the gradients of the cost functional with respect to the control variables are determined through sensitivity derivatives. By directly differentiating the state equations with respect to the control variable, the sensitivity equations for each control variable together with initial boundary conditions are obtained. Numerically, in order to discretise the TDGL equations, the authors use a finite element method in space and a backward differential multi-step method in time, and the resulting nonlinear system is solved by Newton's method. Vortex pinning and critical currents are studied numerically for different normal inclusion configurations. The computations are divided into two parts: (i) only the applied current is chosen as the control parameter to reduce the computational cost, and the effects on the critical current of the sample boundary and the number, size, shape, orientation and location of the normal inclusion sites are studied, all of which are fixed during the optimization procedure; (ii) various parameters are included for each inclusion site as control variables in addition to the current so as to determine the optimal location, shape, etc., of those inclusions. The finite element spaces are constructed in the calculations by using continuous piecewise quadratic polynomials. The time step is chosen in order to ensure that the computation is efficient and the accumulated errors from time discretisation are insignificant after a long time.