Wang’s CD_p(m,K) Curvature Condition
- Wang’s CD_p(m,K) condition is a nonlinear generalization of the Bakry–Émery curvature-dimension criterion using the p-Laplacian, distinguishing its behavior from the classical p=2 case.
- It establishes a local curvature criterion via higher-order difference operators on graphs, indicating that nonnegative p-curvature holds in specific regimes such as path leaves and certain cycles, while failing in others.
- The failure of tensorization for p>2 prevents extending nonnegative curvature to Cartesian graph products, impacting applications in spectral theory and geometric analysis on discrete structures.
Wang’s condition is a nonlinear generalization of the Bakry–Émery curvature dimension condition to graph settings, defined via the -Laplacian operator for . For a weighted undirected graph , it yields a local curvature criterion involving higher-order difference operators. This framework extends the Bakry–Émery theory to nonlinear settings by replacing the standard Laplacian and carré du champ operators with their -analogs, revealing new geometric and analytic phenomena on discrete spaces, especially concerning the behavior of "nonnegative -curvature" and its stability under taking Cartesian products of graphs (Hu, 22 Jan 2026).
1. Definitions and Operator Structure
Given a graph with finite positive vertex measure and symmetric edge weights , consider and :
- -Laplacian:
- -gradient (first-order carré du champ):
with .
- Iterated carré du champ:
where uses the same direction .
These recover the classical Bakry–Émery operators when (i.e., , , and standard expressions).
2. The Condition
The condition at , for all positive , is defined by: The largest for which this holds is called the -Bakry–Émery curvature, denoted $\,_{p,x}G(m)$. The (“nonnegative -curvature”) condition simplifies to for all admissible .
For , this coincides with the Bakry–Émery curvature-dimension criterion. For , it yields a genuinely nonlinear curvature notion, creating new regimes in discrete curvature theory (Hu, 22 Jan 2026).
3. Explicit Evaluation on Graph Examples
Analytic expressions for have been worked out on paths, cycles, and star graphs. All computations use unweighted, unit-measure graphs:
- Path Graphs ():
- For the middle vertex in or , explicit calculations show that for , , so holds. For $1
), : nonnegative -curvature fails.
- At leaves of any path , it is shown for all that holds, due to nonnegativity of the associated function at its minimum.
- On general vertices in (), for , all analysis terms in the decomposition of are nonnegative.
- For the middle vertex in or , explicit calculations show that for , , so holds. For $1
- Cycles (, ):
- For most , the combinatorial structure local to a vertex is as in ; thus, holds for . For small cycles and , separate analysis confirms the result for and failure for $1
- For most , the combinatorial structure local to a vertex is as in ; thus, holds for . For small cycles and , separate analysis confirms the result for and failure for $1
- Star Graphs ():
- At a leaf , the curvature constant is given explicitly and becomes negative as soon as the number of leaves . For , nonnegative -curvature holds.
The following table summarizes the regimes for :
| Graph/Vertex | $1
| |
|---|---|---|
| Path (middle) | Curvature | Curvature |
| Path (leaf) | Curvature | Curvature |
| Star (leaf, ) | Curvature | Curvature may fail |
| Star (leaf, ) | Curvature | Curvature |
| Cycle () | Curvature | Curvature |
4. Failure of Tensorization for
For the classical Laplacian (), the product operator satisfies
enabling nonnegative curvature to "tensorize" under Cartesian products of graphs, which is essential for extending curvature lower bounds to graph products.
By contrast, for the -versions:
- and are additive under graph products: and similarly for .
- However, for with , explicit calculations (involving mixed-difference terms) yield counterexamples where
Consequently, the natural proof strategy establishing preservation of nonnegative curvature under product operations fails for . This marks a significant divergence from the Bakry–Émery theory (Hu, 22 Jan 2026).
5. Main Theoretical Consequences
- For , nonnegative -curvature, i.e., , is abundant in paths (all vertices), cycles (all ), and star graphs at leaves up to neighbors.
- In contrast, for $1
, , and display , and so fails dramatically.
- On star graphs, the explicit formula for the leaf-curvature constant exhibits sharp dependence on the number of leaves, becoming negative for large stars ().
- The failure of the tensorization property for prevents extension of nonnegative -curvature to Cartesian products by Bakry–Émery style arguments, even though and retain their additivity properties.
This structure highlights a dichotomy: the -Bakry–Émery framework recovers the classical situation, but for affords new geometric features, including strong dependence on graph structure, lack of tensorization, and a break-down of local nonnegativity in certain regimes.
6. Connections and Limitations
Wang’s condition recovers Bakry–Émery for and generalizes it to a truly nonlinear curvature notion for in discrete settings. The dichotomy in curvature behavior (between and $1
demonstrate that simple extension of classical results to these nonlinear contexts is not possible. This complexity underscores both the potential and the limitations of applying nonlinear curvature-dimension inequalities in network analysis, spectral theory, and metric geometry on graphs (Hu, 22 Jan 2026).