Cheng-type Eigenvalue Comparison Theorems

• Cheng-type eigenvalue comparison theorems are techniques that compare Laplace eigenvalues on manifolds with model spaces under curvature bounds.
• They utilize one-dimensional Sturm–Liouville formulations and radial basis functions to derive sharp eigenvalue bounds in various geometric configurations.
• Recent extensions include nonlinear, weighted, and synthetic frameworks, broadening applications to Kähler geometry and non-smooth metric measure spaces.

Cheng-type eigenvalue comparison theorems are spectral comparison results in which eigenvalues of Laplace-type operators on a manifold or a domain are estimated against model eigenvalues determined by constant-curvature space forms, one-dimensional Sturm–Liouville problems, or synthetic curvature-dimension models. In the classical Riemannian setting, Cheng’s 1975 estimates control Laplace eigenvalues of closed manifolds by diameter or volume under a lower Ricci curvature bound; subsequent work extends the same comparison paradigm to geodesic balls, convex and non-convex domains, Robin and mixed boundary problems, weighted and nonlinear operators, differential forms, Kähler geometry, weak curvature hypotheses, and non-smooth metric measure spaces (Hassannezhad et al., 2015, Chen et al., 6 Jul 2026, Luca et al., 31 Jul 2025).

1. Classical formulation and model spaces

In the closed-manifold setting, Cheng’s comparison principle appears as an upper bound on Laplace eigenvalues under a lower Ricci curvature bound. If $(M^n,g)$ is closed and connected with

$\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$

then there exist constants $C_6(n)$ and $C_7(n)$, depending only on $n$, such that for every $k\ge 1$,

$$\lambda_k(M)\le \frac{(n-1)^2}{4}K + C_6(n)\Bigl(\frac{k}{d}\Bigr)^2, \qquad \lambda_k(M)\le \frac{(n-1)^2}{4}K + C_7(n)\Bigl(\frac{k}{V^{1/n}}\Bigr)^2,$$

where $d=\operatorname{diam}(M)$ and $V=\operatorname{Vol}_g(M)$ (Hassannezhad et al., 2015). In this formulation, the term $\frac{(n-1)^2}{4}K$ reflects the bottom of the spectrum on the simply connected space form of constant curvature $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$0, while the remaining term encodes the global size parameter.

A more local version compares the first Dirichlet eigenvalue of a geodesic ball with that of the corresponding ball in a model space. In the weighted framework of Chen–Ma–Mao, the model eigenvalue $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$1 is defined by the radial Sturm–Liouville problem

$\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$2

and if $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$3 and $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$4, the bound

$\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$5

is exactly Cheng’s original result for the Laplacian (Chen et al., 6 Jul 2026). The same model-ball viewpoint also underlies the $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$6-Laplacian comparison of Abolarinwa–Azami, where $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$7 under $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$8, with equality precisely in the isometric case (Abolarinwa et al., 2020).

The classical model spaces are geodesic balls in simply connected space forms. For $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$9, Cheng’s comparison reduces to Euclidean balls; for $C_6(n)$0, the hyperbolic radial ODE replaces the Euclidean Bessel description; and for $C_6(n)$1, the admissible radius is restricted by the model injectivity scale (Hassannezhad et al., 2015, Chen et al., 6 Jul 2026).

2. Comparison mechanisms

The standard proof architecture begins with a radial ground state on the model ball and transplants it to the target manifold. In the $C_6(n)$2-system, one takes the radial first eigenfunctions $C_6(n)$3 on the model ball $C_6(n)$4, defines $C_6(n)$5 and $C_6(n)$6, and inserts the pair into the Rayleigh-type variational characterization of $C_6(n)$7 (Abolarinwa et al., 2020). The key geometric input is Bishop’s volume comparison in polar coordinates,

$C_6(n)$8

together with monotonicity of the radial model eigenfunctions and an integration-by-parts argument that transfers the model ODE to the ambient manifold.

In the weighted Bakry–Émery setting, the same mechanism is recast in terms of the weighted measure $C_6(n)$9. Under

$C_7(n)$0

along minimizing radial geodesics, one has the generalized Bishop–Gromov estimate

$C_7(n)$1

and the radial Laplacian comparison

$C_7(n)$2

These inequalities make the transplanted radial function $C_7(n)$3 an admissible trial function whose weighted Rayleigh quotient is bounded above by the model eigenvalue (Chen et al., 6 Jul 2026).

For global spectral bounds, Cheng’s method is often combined with min–max splitting. In the Kato-Ricci setting, Carron–Rose heat-kernel estimates, volume doubling, and a Poincaré inequality imply a Dirichlet estimate on each ball $C_7(n)$4,

$C_7(n)$5

and since a manifold of diameter $C_7(n)$6 contains two disjoint balls of radius $C_7(n)$7, the Courant principle yields

$C_7(n)$8

(Rose et al., 2020). In the non-smooth setting, this proof pattern is replaced by localization: the measure is disintegrated along one-dimensional rays, the comparison is proved fiberwise, and the global inequality is recovered by reintegration (Luca et al., 31 Jul 2025). This suggests that the robust core of the Cheng method is not a specific curvature tensor inequality but the availability of a one-dimensional comparison structure together with a variational principle.

3. Nonlinear, weighted, and higher-rank extensions

The modern literature uses “Cheng-type” for a family of comparison principles that preserve the model-space logic while changing the operator, the boundary condition, or the geometric category.

Setting	Operator	Comparison statement
Geodesic balls	$C_7(n)$9-Laplacian system	$n$0 under $n$1
Geodesic balls and compact manifolds with boundary	Robin $n$2-Laplacian	For $n$3, model-ball or one-dimensional lower/upper bounds; inequalities reverse for $n$4
Weighted complete manifolds	Witten-Laplacian and weighted $n$5-Laplacian	$n$6 and $n$7
Closed manifolds	Hodge Laplacian on $n$8-forms	Uniform upper bounds under Ricci, injectivity-radius, and diameter bounds
Compact Kähler manifolds	Complex Laplacian $n$9	$k\ge 1$0 if $k\ge 1$1

For the $k\ge 1$2-Laplacian system, the Cheng comparison is sharp: equality holds if and only if the metric ball is isometric, via the exponential map, to the model space form ball (Abolarinwa et al., 2020). In the weighted case, Chen–Ma–Mao prove the analogous sharp inequalities for both the Witten-Laplacian and the weighted $k\ge 1$3-Laplacian under a lower bound on the $k\ge 1$4-Bakry–Émery Ricci tensor; they also identify the classical Cheng theorem as the specialization $k\ge 1$5, $k\ge 1$6 (Chen et al., 6 Jul 2026).

Boundary conditions substantially modify the comparison model. For the first Robin eigenvalue $k\ge 1$7 of the $k\ge 1$8-Laplacian, Li–Wang prove that on a geodesic ball $k\ge 1$9 with $$\lambda_k(M)\le \frac{(n-1)^2}{4}K + C_6(n)\Bigl(\frac{k}{d}\Bigr)^2, \qquad \lambda_k(M)\le \frac{(n-1)^2}{4}K + C_7(n)\Bigl(\frac{k}{V^{1/n}}\Bigr)^2,$$0,

$$\lambda_k(M)\le \frac{(n-1)^2}{4}K + C_6(n)\Bigl(\frac{k}{d}\Bigr)^2, \qquad \lambda_k(M)\le \frac{(n-1)^2}{4}K + C_7(n)\Bigl(\frac{k}{V^{1/n}}\Bigr)^2,$$1

while the inequality reverses for $$\lambda_k(M)\le \frac{(n-1)^2}{4}K + C_6(n)\Bigl(\frac{k}{d}\Bigr)^2, \qquad \lambda_k(M)\le \frac{(n-1)^2}{4}K + C_7(n)\Bigl(\frac{k}{V^{1/n}}\Bigr)^2,$$2. For compact manifolds with smooth boundary, Ricci lower bound, boundary mean-curvature lower bound, and inradius $$\lambda_k(M)\le \frac{(n-1)^2}{4}K + C_6(n)\Bigl(\frac{k}{d}\Bigr)^2, \qquad \lambda_k(M)\le \frac{(n-1)^2}{4}K + C_7(n)\Bigl(\frac{k}{V^{1/n}}\Bigr)^2,$$3, the comparison is with a one-dimensional model problem involving

$$\lambda_k(M)\le \frac{(n-1)^2}{4}K + C_6(n)\Bigl(\frac{k}{d}\Bigr)^2, \qquad \lambda_k(M)\le \frac{(n-1)^2}{4}K + C_7(n)\Bigl(\frac{k}{V^{1/n}}\Bigr)^2,$$4

and the Dirichlet problem is recovered as $$\lambda_k(M)\le \frac{(n-1)^2}{4}K + C_6(n)\Bigl(\frac{k}{d}\Bigr)^2, \qquad \lambda_k(M)\le \frac{(n-1)^2}{4}K + C_7(n)\Bigl(\frac{k}{V^{1/n}}\Bigr)^2,$$5 (Li et al., 2020). In a related mixed-boundary direction, the $$\lambda_k(M)\le \frac{(n-1)^2}{4}K + C_6(n)\Bigl(\frac{k}{d}\Bigr)^2, \qquad \lambda_k(M)\le \frac{(n-1)^2}{4}K + C_7(n)\Bigl(\frac{k}{V^{1/n}}\Bigr)^2,$$6-Laplacian on domains with an interior hole admits Faber–Krahn-type inequalities and Cheng-type comparison theorems on manifolds, together with comparison results for inner Dirichlet and outer Neumann data in minimal submanifolds in Euclidean space (Wang, 2014).

The label also extends beyond scalar second-order operators. Bhattacharya–Maity establish Cheng-type upper bounds for the Hodge Laplacian on differential forms under

$$\lambda_k(M)\le \frac{(n-1)^2}{4}K + C_6(n)\Bigl(\frac{k}{d}\Bigr)^2, \qquad \lambda_k(M)\le \frac{(n-1)^2}{4}K + C_7(n)\Bigl(\frac{k}{V^{1/n}}\Bigr)^2,$$7

replacing the sectional-curvature hypotheses used by Dodziuk and Lott with Ricci lower bounds plus harmonic-coordinate control (Bhattacharya et al., 11 Mar 2026). In Kähler geometry, Tam–Yu prove the lower bound

$$\lambda_k(M)\le \frac{(n-1)^2}{4}K + C_6(n)\Bigl(\frac{k}{d}\Bigr)^2, \qquad \lambda_k(M)\le \frac{(n-1)^2}{4}K + C_7(n)\Bigl(\frac{k}{V^{1/n}}\Bigr)^2,$$8

for the first nonzero eigenvalue of the complex Laplacian under $$\lambda_k(M)\le \frac{(n-1)^2}{4}K + C_6(n)\Bigl(\frac{k}{d}\Bigr)^2, \qquad \lambda_k(M)\le \frac{(n-1)^2}{4}K + C_7(n)\Bigl(\frac{k}{V^{1/n}}\Bigr)^2,$$9, with rigidity to a Kähler–Einstein metric in the equality case and, under positive holomorphic bisectional curvature, rigidity to $d=\operatorname{diam}(M)$0 with the Fubini–Study metric (Tam et al., 2010). For higher-order elliptic operators, the poly-Laplacian literature includes sharp upper bounds for sums of the first $d=\operatorname{diam}(M)$1 Dirichlet eigenvalues for arbitrary order, presented as an extension of Cheng–Wei and improved when $d=\operatorname{diam}(M)$2 and $d=\operatorname{diam}(M)$3 is large enough (Zeng, 2013).

4. Weak curvature hypotheses and alternative geometric controls

A major line of development weakens the pointwise Ricci lower bound while retaining Cheng-type spectral control. Under a Kato condition on the negative part $d=\operatorname{diam}(M)$4 of the Ricci curvature,

$d=\operatorname{diam}(M)$5

Rose proves that on a closed manifold the first nonzero eigenvalue satisfies

$d=\operatorname{diam}(M)$6

The proof uses Li–Yau type gradient estimates, Harnack inequalities, two-sided heat-kernel bounds, volume doubling, and a Poincaré inequality derived from the small Kato constant (Rose et al., 2020). In the boundary case, the same paper introduces the Neumann Kato constant $d=\operatorname{diam}(M)$7 and obtains, under additional boundary geometry assumptions, a Cheng-type lower bound for the first nonzero Neumann eigenvalue $d=\operatorname{diam}(M)$8.

A different weakening replaces Ricci control by scalar curvature. Munăteanu–Wang prove that if a complete noncompact $d=\operatorname{diam}(M)$9-manifold satisfies

$V=\operatorname{Vol}_g(M)$0

has finitely many ends, finite first Betti number, a lower Ricci bound, and a mild volume non-collapse condition, then the bottom of the spectrum obeys the sharp upper bound

$V=\operatorname{Vol}_g(M)$1

This contrasts with the classical Ricci-based estimate $V=\operatorname{Vol}_g(M)$2 in dimension $V=\operatorname{Vol}_g(M)$3, and the rigidity question for the equality case remains open (Munteanu et al., 2021).

Integral curvature assumptions yield another variant. Kwong’s Theorem 17 assumes a one-sided integral bound on $V=\operatorname{Vol}_g(M)$4 along all geodesics emanating from the center and obtains a model-ball comparison for the first Dirichlet eigenvalue, formulated there as

$V=\operatorname{Vol}_g(M)$5

with equality if and only if the ball is isometric to the model space (Kwong, 2019). The breadth of these formulations suggests that “Cheng-type” now denotes a comparison strategy rather than a single fixed inequality: depending on the operator, the boundary condition, and the curvature input, the conclusion may be an upper bound, a lower bound, or a model-space rigidity statement.

5. Synthetic curvature and the non-smooth extension

In the non-smooth category, De Luca, De Ponti, Mondino, and Tomasiello extend Cheng’s comparison to essentially non-branching $V=\operatorname{Vol}_g(M)$6 spaces. Writing

$V=\operatorname{Vol}_g(M)$7

they define the one-dimensional model eigenvalue $V=\operatorname{Vol}_g(M)$8 by the weighted Dirichlet problem on $V=\operatorname{Vol}_g(M)$9, and prove the sharp estimate

$\frac{(n-1)^2}{4}K$0

for every metric ball in an essentially non-branching $\frac{(n-1)^2}{4}K$1 space (Luca et al., 31 Jul 2025). When the space is a smooth $\frac{(n-1)^2}{4}K$2-manifold with $\frac{(n-1)^2}{4}K$3, this recovers the classical Cheng upper bound.

The proof is driven by the localization technique. The ball is decomposed into one-dimensional rays associated with the distance function $\frac{(n-1)^2}{4}K$4, the ambient measure is disintegrated into conditional measures $\frac{(n-1)^2}{4}K$5 satisfying one-dimensional curvature-dimension inequalities, and the model eigenfunction $\frac{(n-1)^2}{4}K$6 is tested fiberwise before being reintegrated globally (Luca et al., 31 Jul 2025). This replaces smooth polar-coordinate comparison by a synthetic one-dimensional convexity principle.

The corresponding rigidity theorem on $\frac{(n-1)^2}{4}K$7 spaces is highly structured. If equality holds in the Dirichlet comparison, then exactly one of three mutually exclusive configurations occurs: $\frac{(n-1)^2}{4}K$8 is a single point, exactly two points, or at least three points, leading respectively to a one-dimensional weighted model, a local weighted interval model, or an $\frac{(n-1)^2}{4}K$9-cone structure (Luca et al., 31 Jul 2025).

The same paper derives two further consequences. First, if the compact space has diameter $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$00, then for every $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$01 the $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$02-th Neumann eigenvalue satisfies

$\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$03

Second, for a non-compact $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$04 space with $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$05 and $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$06,

$\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$07

A physical application identifies spin-2 Kaluza–Klein masses with Neumann eigenvalues of a weighted Laplacian and concludes

$\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$08

for warped compactifications satisfying the reduced energy condition (Luca et al., 31 Jul 2025).

6. Rigidity, sharpness, and current directions

Rigidity is a persistent feature of Cheng-type comparison. In the $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$09-Laplacian system, equality forces the geodesic ball to be isometric to the model ball (Abolarinwa et al., 2020). For the Witten-Laplacian and weighted $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$10-Laplacian, equality holds if and only if $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$11 is isometric to the model ball in $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$12, with $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$13 constant along each radius (Chen et al., 6 Jul 2026). For the Robin $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$14-Laplacian, equality in the inradius comparison characterizes the canonical $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$15-model spaces (Li et al., 2020). In the Kähler setting, equality in $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$16 implies Kähler–Einstein geometry, and under positive holomorphic bisectional curvature it characterizes $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$17 with the Fubini–Study metric (Tam et al., 2010). In the synthetic setting, equality yields a one-dimensional or conical structure (Luca et al., 31 Jul 2025).

Sharpness is more nuanced. For Cheng’s closed-manifold diameter estimate, the quadratic growth in $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$18 is sharp for thin flat tori, whereas the volume form is asymptotically sharp in the sense of Weyl (Hassannezhad et al., 2015). For the Hodge Laplacian, Bhattacharya–Maity emphasize that the Cheng-type upper bound has the same $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$19-power as Weyl’s law, so it is sharp in order of growth (Bhattacharya et al., 11 Mar 2026). In the scalar-curvature comparison of Munăteanu–Wang, $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$20 attains $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$21, but the rigidity question under only the scalar bound is still open, and higher-dimensional analogues fail in general (Munteanu et al., 2021).

A common misconception is that Cheng-type theorems concern only the original Dirichlet comparison for geodesic balls under pointwise Ricci lower bounds. Current usage is materially broader: it includes closed-manifold diameter and volume bounds, weighted and nonlinear analogues, Robin and mixed boundary problems, lower bounds such as the Kähler estimate $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$22, bottom-spectrum estimates under scalar curvature, and synthetic $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$23 statements (Hassannezhad et al., 2015, Tam et al., 2010, Li et al., 2020, Luca et al., 31 Jul 2025).

A related branch of the literature replaces direct model comparison by universal inequalities in the Cheng–Yang tradition. For divergence-form operators, including the Laplacian and the square Cheng–Yau operator, one obtains gap estimates of the form

$\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$24

with the Laplacian case coinciding with the result of Chen–Zheng–Yang in the order of the eigenvalues (Silva et al., 2024). For the drifting Laplacian on bounded domains, Chen–Gomes–Miranda extend Cheng–Yang average bounds and second-Yang type inequalities by incorporating Weyl asymptotics and extrinsic geometry (Gomes et al., 2016). On pinched Cartan–Hadamard manifolds, drifted Cheng–Yau operators satisfy universal inequalities derived from a Bochner-type formula and Rauch comparison (Fonseca et al., 2021). In hyperbolic space, Cheng’s conjectured curvature-shifted inequality

$\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$25

has been verified up to a factor $\mathrm{Ricci}\ge -(n-1)K,\qquad K\ge 0,$26 for two special classes of bounded domains, while the full conjecture for arbitrary domains remains open (Luo, 22 Apr 2026).

Taken together, these developments show that the Cheng comparison paradigm has evolved from a model-ball estimate for the scalar Laplacian into a general spectral methodology. Its modern forms may be driven by Bishop–Gromov comparison, Bakry–Émery geometry, heat-kernel and Kato techniques, Green’s-function identities under scalar curvature, localization on one-dimensional needles, or harmonic-coordinate control for differential forms. What remains invariant is the guiding principle: global or local geometric control is converted into a comparison with a tractable model spectrum.