Many astrophysical plasmas and some laboratory plasmas are relativistic: either the thermal speed or the local bulk flow in some frame approaches the speed of light. Often, such plasmas are magnetized in the sense that the Larmor radius is smaller than any gradient scale length of interest. Conventionally, relativistic MHD is employed to treat relativistic, magnetized plasmas; however, MHD requires the collision time to be shorter than any other time scale in the system. Thus, MHD employs the thermodynamic equilibrium form of the stress tensor, neglecting pressure anisotropy and heat flow parallel to the magnetic field. We re-examine the closure question and find a more complete theory, which yields a more physical and self-consistent closure. Beginning with exact moments of the kinetic equation, we derive a closed set of Lorentz-covariant fluid equations for a magnetized plasma allowing for pressure and heat flow anisotropy. Basic predictions of the model, including its thermodynamics and the dispersion relation's dependence upon relativistic temperature, are examined. Further, the model is applied to two extant astrophysical problems. / text

A Direct Simulation Monte Carlo fluid dynamics code named FENIX has been employed to study gas flow-through properties of the inductively coupled plasma mass spectrometer (ICP-MS). Simulation data have been tested against the Navier-Stokes and heat equations in order to see if FENIX functions properly. The Navier-Stokes and heat equations have been constructed from simulation data and are compared term by term. This comparison shows that FENIX is able to correctly reproduce fluid dynamics throughout the ICP-MS simulation, with an exception immediately behind the ICP-MS sampler cone, where the continuum criterion for the Navier-Stokes equation is not met. Testing the data produced by Fenix also shows that this DSMC method correctly produces momentum and thermal boundary layer phenomenon as well. FENIX output data produce statistical fluctuations of about 2%. Limitations occur from fitting data near surfaces, incurring a relative error of about 5%, and fitting data to take second derivatives where fluid velocity gradients are steep, introducing a relative error of about 10%.