arXiv2025
We construct critical trajectories in kinetic geometry, i.e. curves in $\mathbb{R}^{1+2n}$ that are: tangential to the vector fields $\partial_t+v\cdot \nabla_x$ and $\nabla_v$, connecting any two given points, respecting the underlying kinetic scaling, and with the property, that the singularity of the $v$-tangent vector near the starting point equates the degeneracy of the dependency of the curve velocity in terms of the endpoint velocity. The construction is based on Newton's laws of motion, where the ansatz for the forcing of the kinetic trajectory is the superposition of functions combining the correct power scaling with desynchronised logarithmic oscillations. These critical trajectories provide a robust and versatile ''almost exponential map'' that allows to prove several functional analytic estimates. We introduce a notion of kinetic mollification and, as an application, deduce the kinetic Sobolev inequality with optimal exponent without relying on the fundamental solution. Moreover, we establish a universal estimate for the logarithm of positive supersolutions to the Kolmogorov equation with rough coefficients inspired by the work of Moser (1961, 1964) on elliptic and parabolic problems. Combining this estimate with De Giorgi-Moser iterations and a lemma due to Bombieri and Giusti, we give an alternative proof of the (weak) Harnack inequality for the Kolmogorov equation with rough coefficients, following the ideas of Moser (1971). Our result gives the optimal range of exponents in the weak Harnack inequality and the optimal (geometric) dependency of the Harnack constant on the bounds of the diffusion matrix.