Magnetic charges or why it is sometimes reasonable to model something mathematically that does not exist at all (Q2457883)

In the presence of currents and electric charges Maxwell equations suffer from a lack of symmetry because there are not discovered any magnetic charges (up to now?). On the other hand fictive magnetic charges can be useful in applications; an example is discussed impressively in the present paper with the aim to reduce errors in computational electromagnetics. Considering Maxwell's equations \[ (1)\quad{\vec E}_t - c^2 \nabla \times \vec B = -\frac{\vec j}{{\varepsilon}_0}, \qquad (2) \quad {\vec B}_t + \nabla \times \vec E = 0, \qquad (3) \quad \nabla \cdot \vec E = \frac{q}{{\varepsilon}_0}, \qquad (4) \quad \nabla \cdot \vec B = 0, \] the main idea is sketched in the following. The magnetic flux density \(\vec B\) can be computed from equation (2) for known electric field intensity \(\vec E\) and a given initial value for the magnetic flux density. Equation (4) is true for all times for an divergence-free initial value of the magnetic flux density because (4) can be deduced from (2) according the theorem that the divergence of the curl of any vector is identically zero. But, using numerical methods it can be that the condition of solenoidal magnetic flux density (4) is not fulfilled for all times because of discretisation and rounding errors. Thus, it is proposed to change the equations (2) and (4) to \[ {\vec B}_t + \nabla \times \vec E = - {\vec j}^M, \qquad \nabla \cdot \vec B = q^M \] with a fictive magnetic current density \({\vec j}^M\) and a fictive magnetic charge \(q^M\), that establish the above by the reviewer mentioned symmetry with the corresponding conservation relation of magnetic charge \(q^M_t + \nabla \cdot {\vec j}^M = 0\). This total magnetic charge can be written as \(q^M_t(\phi) + \nabla \cdot {\vec j}^M(\phi) = 0\) in order to reduce the number of variables in the ``symmetric'' Maxwell equations; that is necessary now. An appropriate ansatz like \(q^M(\phi) = \phi, \quad {\vec j}^M(\phi) = \phi \vec v\) (\(\vec v\) fictive velocity) or \(q^M(\phi) = \alpha \phi + \beta {\phi}_t, \quad {\vec j}^M(\phi) = - \nabla \phi\) (\(\alpha, \beta \) parameter) leads to equations which can be used such that the total magnetic charge converges to zero with time \(t\). Examples of such auxiliary equations are a linear advection equation, the heat conduction equation, the wave equation and the telegraph equation. The numerical efficiency of the different variants is discussed. The application of the method is outlined for the simulation of the Sun's surface by magnetohydrodynamic equations.