Laplacian Comparison Theorems

• Laplacian Comparison Theorems are inequalities that link the weighted Laplacian of distance functions to model geometries under curvature–dimension conditions.
• They use the m–Bakry–Émery Ricci tensor to extend classical theorems like Myers’ and Bishop–Gromov comparisons to regimes where m ≤ 1.
• These results provide sharp control for diffusion processes, guaranteeing properties such as stochastic completeness and the Feller property on weighted manifolds.

A Laplacian comparison theorem provides an inequality relating the Laplacian (or more generally, a diffusion operator) of a distance function to a comparison quantity arising from an associated model geometry, typically under curvature–dimension conditions. On weighted complete Riemannian manifolds equipped with a symmetric diffusion operator $L=\Delta-\nabla\phi\cdot \nabla$ (where $\phi\in C^2(M)$ is a potential), recent work has established sharp Laplacian comparison results under a “curvature–dimension” $(\mathrm{CD}(K, m))$ condition for $m \leq 1$, a regime previously unexplored in this generality (Kuwae et al., 2020, Kuwae et al., 2021). These results leverage the $m$–Bakry–Émery Ricci tensor and generalize many classical geometric and analytic theorems, including weighted Myers' theorem, Bishop–Gromov volume comparison, stochastic completeness and the Feller property of the associated diffusion, as well as rigidity and splitting results, to the entire range $m \leq 1$.

1. Framework: Weighted Manifolds and the $m$–Bakry–Émery Ricci Tensor

Let $(M, g)$ be a complete $n$–dimensional Riemannian manifold, $\phi\in C^2(M)$, and consider the weighted measure $d\mu = e^{-\phi} dv_g$. The symmetric diffusion operator $L = \Delta - \nabla\phi\cdot\nabla$ has invariant measure $\mu$ and governs the associated diffusive semigroup. To encode Ricci curvature in this context with an “effective” dimension $m$, the $m$–Bakry–Émery Ricci tensor is defined by

$$\mathrm{Ric}_{m,n}(L) = \mathrm{Ric} + \nabla^2\phi - \frac{\nabla\phi\otimes \nabla\phi}{m - n},$$

with curvature–dimension condition (CD($K(x)$, $m$)) given by

$$\mathrm{Ric}_{m,n}(L)(x) \geq K(x)\quad\text{in the appropriate direction.}$$

For $m=n$, this reduces to the Bakry–Émery tensor, and for $m=\infty$ to the standard Bakry–Émery Ricci tensor. Prior literature focused on $m\geq n$ or $m=1$ (typically for gradient vector fields) [XDLi05, WeiWylie, Wylie:WarpedSplitting, WylieYeroshkin]. The regime $m<1$ is new and not previously addressed in the literature (Kuwae et al., 2020, Kuwae et al., 2021).

2. Statement of the Laplacian Comparison Theorem

For a reference point $p\in M$, let $r_p(x)$ be the Riemannian distance from $p$ to $x$. The main theorem states that under a $\mathrm{CD}(K, m)$–condition for $m \leq 1$,

$$\mathrm{Ric}_{m,n}(L)_x(\nabla r_p, \nabla r_p) \geq (n-m) \kappa(s_p(x)) e^{-4\phi(x)/(n-m)}$$

holds along geodesics away from the cut locus, where $K(x) = (n-m)\kappa(s_p(x))e^{-4\phi(x)/(n-m)}$ with $\kappa$ continuous, and

$$s(p, q) = \inf\left\{\int_0^{r_p(q)} e^{-2\phi(\gamma_t)/(n-m)}\,dt : \gamma \text{ is a unit-speed geodesic from }p\text{ to }q\right\}$$

is a re-parametrized distance function absorbing the potential $\phi$.

Define the model function $m_\kappa(s) = (n-m)\cot_\kappa(s)$, where $\cot_\kappa(s)$ solves the Riccati equation

$$-\frac{d}{ds}\cot_\kappa(s) = \kappa(s) + [\cot_\kappa(s)]^2, \quad \lim_{s\downarrow 0} s\cot_\kappa(s) = 1.$$

The theorem (Theorem 2.1 in (Kuwae et al., 2020)) gives the sharp estimate

$$(L r_p)(x) \leq e^{-2\phi(x)/(n-m)}\, m_\kappa(s_p(x))$$

for $x$ in the pre-cut-locus domain of $p$. Thus, the weighted Laplacian, scaled by the density, is controlled by the model function, with curvature and dimension replaced by their modified, “weighted” counterparts.

3. Consequences: Myers, Bishop–Gromov, and Splitting Theorems

Weighted Myers’ Theorem

Under the hypothesis that the comparison model’s “explosion time” $\delta_\kappa$ (smallest $s$ s.t. $s_\kappa(s) = 0$) is finite and that the re–parametrized distance $s(p, q)$ grows appropriately along diverging geodesics ($(φ,m)$–completeness), one deduces that $M$ is compact (Theorem 2.2). This generalizes Myers’ theorem to the weighted setting for $m \leq 1$.

Weighted Bishop–Gromov Volume Comparison

Using the Laplacian comparison and transport equations for the volume density, one obtains (Theorem 2.3)

$$\frac{\mu(A(p, r_a, r_b))}{\mu(A(p, r_0, r_a))} \leq \frac{\int_{r_0}^{r_b} s_\kappa(s)^{n-m} ds}{\int_{r_0}^{r_a} s_\kappa(s)^{n-m} ds}$$

for suitable annuli $A(p, r_0, r_1)$, with the volume measure computed in terms of the model function, reflecting both the geometry and the effect of the density.

Ambrose–Myers and Cheeger–Gromoll Splitting

The same methodology gives sufficient conditions for compactness via the integral of the weighted Ricci curvature along geodesics (Ambrose–Myers), and, under a nonnegative weighted curvature condition, splitting results akin to Cheeger–Gromoll (Kuwae et al., 2020, Kuwae et al., 2021). Specifically, in the presence of a line and nonnegative $m$–Bakry–Émery Ricci curvature for $m < 1$, isometric splitting as a warped product occurs.

4. Analytical Consequences: Diffusion Processes

Because $L$ generates a symmetric diffusion process with invariant measure $\mu$, the Laplacian comparison directly affects stochastic completeness and long–term behavior. Applying Grigor’yan’s volume growth criteria and the derived Laplacian bounds, one proves that the heat semigroup $P_t = e^{tL}$ is conservative (i.e., $P_t1 = 1$ for all $t > 0$) and that the Feller property holds (namely, $P_t$ maps $C_0(M)$ into itself) under suitable integrability conditions on the lower curvature bound (Kuwae et al., 2020). These results guarantee that $L$–diffusion does not exhibit explosion and maintains good regularity at infinity for $m \leq 1$.

5. New Regime $m < 1$: Extension and Novelty

Most prior work established Laplacian comparison, volume, and diameter estimates only for $m \geq n$ (and gradient vector fields) [Lot, Qian, XDLi05, WeiWylie] or $m=1$ [Wylie:WarpedSplitting, WylieYeroshkin]. The analysis for $m < 1$ is new (Kuwae et al., 2020, Kuwae et al., 2021). This extension required technical innovations to control the effect of the effective dimension and density in the weighted Laplacian and the Riccati–type comparison. In particular, the reduction to a model ODE (with explicit solutions for $m_\kappa(s)$ and the associated Riccati equation) and the re–parametrization of distance were critical for the derivation of sharp estimates and for making the extension to $m < 1$ both natural and optimal.

6. Summary Table: Major Theorems and Model Quantities

Theorem/Classical Analog	$m \leq 1$ version / Statement	Model Expression
Laplacian Comparison	$$(L r_p)(x) \leq e^{-2\phi(x)/(n-m)} m_\kappa(s_p(x))$$	$m_\kappa(s)=(n-m)\cot_\kappa(s)$
Myers’ Theorem	$\mathrm{diam}_\mu(M) \leq \delta_\kappa$ (if model explodes)	$\delta_\kappa = \inf\{s: s_\kappa(s)=0\}$
Bishop–Gromov Volume	Volume ratio bounded by model integrals	$\int s_\kappa(s)^{n-m} ds$
Cheeger–Gromoll Splitting	Splitting if existence of a line + $\mathrm{CD}(0,m)$	Warped product form
Stochastic Completeness / Feller	$P_t 1 = 1;~P_t: C_0(M)\to C_0(M)$ holds under integrability	Via volume growth

7. Methodological Innovations

The main technical advance is the formulation of comparison via a re–parametrized distance incorporating the density and a Laplacian bound involving the solution of a Riccati ODE adapted to the effective dimension $m$ and curvature function $\kappa$. This allows for a unified argument across all $m \leq 1$, and in particular for the difficult regime $m<1$, which is inaccessible by classical techniques. The connection to diffusion processes via these Laplacian bounds demonstrates the breadth of applicability, controlling both geometric structure and analytic properties.

8. Significance and Open Directions

The extension of Laplacian comparison theorems to $m \leq 1$ unifies and generalizes a wide spectrum of comparison results in geometric analysis. It opens the paper of weighted geometric inequalities and stochastic analysis in settings where the potential enforces an effective dimension “smaller” than the geometric dimension, a setting arising in synthetic Ricci curvature bounds and generalized diffusion theory. Open questions concern the sharpness of these bounds in various geometric classes, and the precise role of non-symmetric operators for more general vector fields or measure-theoretic ambiguities, which are not directly addressed by the standard Bakry–Émery framework.

---

These developments lay the foundation for further exploration of geometry and analysis on weighted manifolds, extending classical comparison theory to settings with weak or even negative effective dimension and providing a comprehensive suite of analytic and geometric tools for the paper of Laplacian–type operators in the presence of general measure densities.