The finite element analysis of a decoupled \(T-\psi\) scheme for solving eddy-current problems (Q2378774)

The eddy current equations are obtained from Maxwell's equations by neglecting the displacement current. In this paper the author derives error estimates for a \(T-\psi\) scheme for solving eddy-current problems. In a \(T-\psi\) scheme one uses that the current is divergence free in the conducting region which implies that it can be written as the rotation of a vector field \(\mathbf{J}=\nabla\times T\). The magnetic field is then written as \(H=T-\nabla\psi\), where \(\psi\) is the magnetic scalar potential. In the non-conducting region the current is zero so there is no need for a vector potential there. After formulating the eddy current equations in terms of \(T\) and \(\psi\) they are transformed into weak form. The appropriate spaces for studying convergence and derive error estimates are the Sobolev spaces \(H^1({\mathbf {curl}},\Omega)\) and \(H^1(\Omega)\), for \(T\) and \(\psi\), respectively. The first theorem is an error estimate for the semi-discrete equations in time. Here the Euler backward time discretization is used. This is followed up by a fully discrete error estimate when the spatial discretization is made with piecewise linear elements. For \(T\) the space is piecewise linear vector elements intersected with \(H^1({\mathbf {curl}},\Omega)\). Provided that the time step \(\tau\) is sufficiently small, the proved error estimate for the electric field is \({\mathcal O}(\sqrt{h^2 +\tau^2})\) in \(L^2\) norm, where \(h\) is the mesh size.