How two-dimensional are planet–disc interactions? I. Locally isothermal discs

Planet--disc interactions, despite being fundamentally three-dimensional, are often studied in the two-dimensional `thin-disk' approximation. The overall morphology of planet--disc interactions has ben shown to be similar in both 2D and 3D simulations, however, the ability of a 2D simulation to quantitatively match 3D results depends strongly on how the potential of the planet is handled. Typically, the 2D planetary potential is smoothed out using some `smoothing length', a free parameter, for which different values have been proposed, depending on the particular aspect of the interaction focused on. In this paper, we re-derive 2D Navier--Stokes in detail for planet--disc interactions to find better ways to represent the 2D gravitational force. We perform a large suite of 2D and 3D simulations to test these force prescriptions. We identify the parts of the interaction that are fundamentally 3D, and test how well our new force prescriptions, as well as traditional smoothed potentials, are able to match 3D simulations. Overall, we find that the optimal way to represent the planetary potential is the `Bessel-type potential', but that even in this case 2D simulations are unable to reproduce the correct scaling of the total torque with background gradients, and are at best able match the one-sided Lindblad torque and gap widths to level of 10 per cent. We find that analysis of observed gap structures based on standard 2D simulations may systematically underestimate planetary masses by a factor of two, and discuss the impacts of 3D effects on observations of velocity kinks.