# Normalized units for the Vlasov-Maxwell system¶

The Gkyl design philosophy involves the implementation of unit-full systems of equations, i.e., Gkyl simulations can be run with real parameters for direct comparison with experiments, with universal constants specified by using values provided by the National Institute of Standards and Technology. For example, in the absence of collisions, the Vlasov-Maxwell system of equations in S.I. units is as follows,

However, one may not always wish to run simulations with the unit-full system. Instead, one can consider a normalized set of equations. A natural choice for the normalization of the Vlasov-Maxwell system of equations would redefine all the relevant quantities as follows,

where

are the speed of light, electron plasma frequency, and the electron skin depth respectively. Note that the charge normalization means that in a proton-electron plasma, \(\tilde{q}_s = \pm 1\), and the density normalization is such that in a quasi-neutral plasma, the initial density of each species is 1.0. We can also check that the electric and magnetic field normalizations are reasonable by making sure that the normalization has the correct units for the electric and magnetic fields in S.I. units,

With these normalizations, the Vlasov-Maxwell system of equations then becomes,

This system of equations has the obvious advantage that universal constants, such as \(\epsilon_0\), are eliminated. In doing so, one does not need to worry about the propagation of round off error from, for example, the accumulation of the current to the electric field in the Ampere-Maxwell law, \(E^{n+1} = E^{n} + \delta t \mathbf{J}/\epsilon_0\) becomed \(E^{n+1} = E^{n} + \delta t \tilde{\mathbf{J}}\). Given Gkyl’s unit-full representation, a simple way to force the Vlasov-Maxwell solver to “use” these units is to specify the following parameters be equal to 1.0,

With the above parameters set to 1.0, then Vlasov-Maxwell simulations require only a few parameters to be completely determined. In a proton-electron plasma, these are, the proton-to-electron mass ratio, \(m_p/m_e\), the proton-to-electron temperature ratio, \(T_p/T_e\), the ratio of some characteristic velocity, such as the electron Alfv'en speed, to the speed of light, \(v_{A_e}/c\), and the plasma beta of either the protons or the electrons, \(\beta\). It is often convenient with this normalized system to use the combination of the ratio of the electron Alfv'en speed to the speed of light and the plasma beta to derive the temperature in normalized units, like so,

assuming the plasma is quasineutral and thus, \(n_0 = 1.0\) for both the protons and electrons. The proton beta and proton temperature then follow from the specified proton-to-electron temperature ratio. It is recommended that the user initialize Maxwellian distribution functions using this derived temperature, so as to avoid the ambiguity of the user’s definition of the thermal velocity,

Whether the user ultimately elects to use \(v_{th_s} = \sqrt{2 T_s/m_s}\) or \(v_{th_s} = \sqrt{T_s/m_s}\) is of no consequence to the initialization of the simulation, and likely only to manifest in the user’s specification of the velocity space extents.