Why, how and when MHD turbulence at low $Rm$ becomes three-dimensional

MHD turbulence at low Magnetic Reynolds number is experimentally investigated by studying a liquid metal flow in a cubic domain. We focus on the mechanisms that determine whether the flow is quasi-2D, 3D or in any intermediate state. To this end, forcing is applied by injecting a DC current $I$ through one wall of the cube only, to drive vortices spinning along the magnetic field. Depending on the intensity of the externally applied magnetic field, these vortices extend part or all of the way through the cube. Driving the flow in this way allows us to precisely control not only the forcing intensity but also its dimensionality. A comparison with the theoretical analysis of this configuration singles out the influences of the walls and of the forcing on the flow dimensionality, which is characterised in several ways. First, when inertia drives three-dimensionality, the velocity near the wall where current is injected scales as $U_b\sim I^{2/3}$. Second, when the distance $l_z$ over which momentum diffuses under the action of the Lorentz force reaches the channel width $h$, the velocity near the opposite wall $U_t$ follows a similar law with a correction factor $(1-h/l_z)$. When $l_z<h$, by contrast, the opposite wall has less influence on the flow and $U_t\sim I^{1/2}$. The central role played by the ratio $l_z/h$ is confirmed by experimentally verifying Sommeria & Moreau (1982)'s scaling $l_z\sim N^{1/2}$ ($N$ is the interaction parameter) and finally, the nature of the three-dimensionality is further clarified by distinguishing weak and strong three-dimensionalities. It is found that both vanish only asymptotically in the limit $N\rightarrow\infty$. This provides evidence that because of the no-slip walls, 1) the transition between quasi-2D and 3D turbulence does not result from a global instability of the flow, and 2) it doesn't occur simultaneously at all scales.