Title: Kinematic analysis of the Large Magellanic Cloud using Gaia DR3

Abstract: Context: The high quality of the Gaia mission data is allowing to study the internal kinematics of the Large Magellanic Cloud (LMC) in unprecedented detail, providing insights on the non-axisymmetric structure of its disc. Aims: To define and validate an improved selection strategy to distinguish the LMC stars from the Milky Way foreground. To check the possible biases that assumed parameters or sample contamination from the Milky Way can introduce in the analysis of the internal kinematics of the LMC using Gaia data. Methods: Our selection is based on a supervised Neural Network classifier using as much as of the Gaia DR3 data as possible. We select three samples of candidate LMC stars with different degrees of completeness and purity; we validate them using different test samples and we compare them with the Gaia Collaboration paper sample. We analyse the resulting velocity profiles and maps, and we check how these results change when using also the line-of-sight velocities, available for a subset of stars. Results: The contamination in the samples from Milky Way stars affects basically the results for the outskirts of the LMC, and the absence of line-of-sight velocities does not bias the results for the kinematics in the inner disc. For the first time, we perform a kinematic analysis of the LMC using samples with the full three dimensional velocity information from Gaia DR3. Conclusions: The dynamics in the inner disc is mainly bar dominated; the kinematics on the spiral arm over-density seem to be dominated by an inward motion and a rotation faster than that of the disc in the piece of the arm attached to the bar; contamination of MW stars seem to dominate the outer parts of the disc and mainly affects old evolutionary phases; uncertainties in the assumed disc morphological parameters and line-of-sight velocity of the LMC can in some cases have significant effects. [ABRIDGED]