Weak laplacian comparison at the cut locus

Assume the Axiom of Countable Choice ${\mathrm{A}\mathrm{C}}_{\omega }$ as inherited through the distance, cut-locus and Laplace–Beltrami suppliers. Recorded orientation, not proved here. Let $(M,g)$ be a complete, connected, boundaryless Riemannian manifold of dimension $n\ge 2$ with $\mathrm{Ric}\ge (n-1)kg$, let $p\in M$, let $r:=d_g(p,\cdot )$ be the distance from $p$ and let $\mathrm{Cut}(p)$ be the cut locus of $p$ (Cut point and cut locus of a point). The pointwise Laplace–Beltrami comparison of Laplacian comparison for distance under a ricci lower bound holds at any point $q$ off $p$ and its cut locus, with the additional restriction $r(q)<\pi \mathrm{/}\sqrt{k}$ when $k>0$ imposed by the model comparison domain. Distance is smooth at every point off $p$ and its cut locus, independently of that model-radius restriction; its Hessian and Laplacian in the sense of Laplace–Beltrami operator as the trace of the Hessian are defined there (Distance from p is smooth off p and the cut locus); the estimate reads ${\mathrm{\Delta }}_{g}r(q)\le (n-1){\mathrm{ct}}_k(r(q))\mathrm{.}$

The following boundary facts are recorded for orientation, with their exact hypotheses, and are not established, used or reproduced in this library:

1. No pointwise statement at the cut locus. At a cut point $q$ the distance function need not be differentiable and the Hessian need not exist, so the inequality above has no pointwise meaning there. Any extension must relax the notion of Laplacian.

2. Barrier and support-function formulations. The standard extensions compare in the barrier sense at $q\mathrm{\ne }p$ with $r(q)<\pi \mathrm{/}\sqrt{k}$ when $k>0$: for every $\epsilon >0$ there is a smooth function ${\phi }_{\epsilon }$ on a neighbourhood of $q$ with ${\phi }_{\epsilon }\ge r$, ${\phi }_{\epsilon }(q)=r(q)$ and ${\mathrm{\Delta }}_g{\phi }_{\epsilon }(q)\le (n-1){\mathrm{ct}}_k(r(q))+\epsilon$. For a unit-speed minimizer $\gamma :[0,r(q)]\to M$ from $p$ to $q$, the standard supports are $r_{q,\eta }(x)=\eta +d(x,\gamma (\eta ))$ with $0<\eta <r(q)$ and $\eta \downarrow 0$. The small-shift restriction is essential; at $\eta =r(q)$ the shifted distance is based at $q$ and is not smooth there. Dai–Wei §1.3 gives the passage from barriers to viscosity inequalities and cites viscosity/distribution equivalence. Tangent-cone analysis of the cut locus is not a prerequisite of this route.

3. Required extra setup. The barrier formulation needs smooth upper supports at nonsmooth distance points. The distributional formulation needs the weak Laplacian pairing ${\int }_{\mathrm{\Omega }}r{\mathrm{\Delta }}_g\psi d{\mathrm{vol}}_g\le {\int }_{\mathrm{\Omega }}(n-1){\mathrm{ct}}_k(r)\psi d{\mathrm{vol}}_g$ for every nonnegative $\psi \in C_c^{\mathrm{\infty }}(\mathrm{\Omega })$, on $\mathrm{\Omega }\subset M\setminus \{p\}$ within the model-radius domain, and the elliptic equivalence theorem just cited. The cut locus having measure zero alone does not prove this inequality. These tools are cited above but not developed in this pair.

This remark is explicitly non-load-bearing: it is a boundary marker, not a proof supplier, and it must not be cited by any item of this pair or elsewhere as an input.

• The pointwise comparison actually proved in this pair is exactly the smooth statement above, with $q$ off $p$ and off the cut locus.
• The rate of volume growth beyond the cut locus is handled in this pair by polar integration, which may discard the null cut locus (Polar integration may discard the cut locus), not by any extension of the pointwise estimate.
• Cheng's maximal-diameter rigidity and the Toponogov hinge and triangle comparisons need upper comparison statements at points where a minimizing segment meets the cut locus; their proofs obtain the required upper support tests locally from their own routes — the ball-packing equality argument with cut-time continuity in the first case, and the local Alexandrov support inequality with first variation in the second — and none of them cites this remark. No item of the pair lists rem-weak-laplacian-comparison-at-the-cut-locus among its dependencies, and no claim about the Laplacian of $r$ at $\mathrm{Cut}(p)$ is made anywhere in the pair. The measure-theoretic statement $\mathrm{vol}(\mathrm{Cut}(p))=0$ (Cut locus of a point has riemannian volume zero) is a separate result and is used only there.