Answers are grounded in this paper and cite their sources.
In the formulation with two magnetic vector potentials, the inaccuracy of the Biot-Savart integration for the source term causes the divergence of the conjugate gradient method. The ill-posed equations are regularized by subtracting a rotational field from the numerically integrated field. The convergence is drastically improved. Even when the current continuity is not satisfied strictly, we can get reasonable results. Also, the response to a small disturbance can be calculated by the regularization, even when the modeling of source currents is fairly rough.
No takes yet. Share an insight, caveat, or question.
A. Kameari (2004) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: