Witt Incomplete Cusp Edge Spaces

• Witt incomplete cusp edge spaces are singular geometric spaces defined by compact manifolds with boundary carrying incomplete cusp edge metrics and satisfying a Witt condition that eliminates middle-degree harmonic forms on the links.
• They are analyzed using stratified models, heat kernel constructions, and Dirac operator techniques to establish essential self-adjointness, discrete spectra, and precise index formulas.
• This framework connects geometric analysis with topological invariants by linking L²-harmonic forms to intersection cohomology and models such as the Weil–Petersson metric.

Witt incomplete cusp edge spaces are singular geometric spaces in which a compact manifold with boundary carries, on its interior, an incomplete cusp edge metric adapted to a boundary fibration, together with a Witt condition on the link geometry of the singular stratum. In a standard model, one has a fibration

$$Z \longrightarrow \partial M \xrightarrow{\ \phi\ } Y$$

and, near $\partial M$, a product-type incomplete cusp edge metric

$$g_0 = dx^2 + x^{2k} g_{\partial M/Y} + \phi^* g_Y,\qquad k\ge2,$$

so the fibres $Z_y$ collapse at rate $x^k$ while the base directions remain of fixed size. In the Hodge–de Rham setting, the Witt condition requires the vanishing of middle-degree harmonic forms on the links, equivalently $H^{\dim Z/2}(Z_y;\mathbb{R})=0$ when $\dim Z$ is even. This framework is motivated in particular by the Weil–Petersson metric near boundary divisors of moduli spaces, where the local model has $k=3$ (Liu, 4 Aug 2025, Gell-Redman et al., 2015, Hunsicker et al., 2012).

1. Geometric models and stratified structure

A smoothly stratified pseudomanifold $X$ is decomposed into strata

$X = \bigsqcup_i S_i,$

with singular strata modeled locally by cone bundles $\partial M$0, and admits a resolution $\partial M$1 by a manifold with corners whose boundary hypersurfaces fibred over the singular strata encode the link geometry (Hunsicker et al., 2012). In the incomplete cusp edge setting, one typically works directly with a compact manifold with boundary $\partial M$2, regards $\partial M$3 as the total space of a fibration $\partial M$4, and obtains the singular space by collapsing the fibres $\partial M$5 to points along the boundary (Liu, 4 Aug 2025).

Near $\partial M$6, with boundary defining function $\partial M$7, local base coordinates $\partial M$8 on $\partial M$9, and fibre coordinates $$g_0 = dx^2 + x^{2k} g_{\partial M/Y} + \phi^* g_Y,\qquad k\ge2,$$0 on $$g_0 = dx^2 + x^{2k} g_{\partial M/Y} + \phi^* g_Y,\qquad k\ge2,$$1, the product-type incomplete cusp edge metric is

$$g_0 = dx^2 + x^{2k} g_{\partial M/Y} + \phi^* g_Y,\qquad k\ge2,$$2

Thus

$$g_0 = dx^2 + x^{2k} g_{\partial M/Y} + \phi^* g_Y,\qquad k\ge2,$$3

The incomplete cusp edge tangent bundle is generated near $$g_0 = dx^2 + x^{2k} g_{\partial M/Y} + \phi^* g_Y,\qquad k\ge2,$$4 by

$$g_0 = dx^2 + x^{2k} g_{\partial M/Y} + \phi^* g_Y,\qquad k\ge2,$$5

and the dual one-forms $$g_0 = dx^2 + x^{2k} g_{\partial M/Y} + \phi^* g_Y,\qquad k\ge2,$$6 are smooth up to the boundary (Liu, 4 Aug 2025). Exact incomplete cusp edge metrics are allowed to differ from $$g_0 = dx^2 + x^{2k} g_{\partial M/Y} + \phi^* g_Y,\qquad k\ge2,$$7 by an $$g_0 = dx^2 + x^{2k} g_{\partial M/Y} + \phi^* g_Y,\qquad k\ge2,$$8 error on this tangent bundle.

A complementary viewpoint is provided by the $$g_0 = dx^2 + x^{2k} g_{\partial M/Y} + \phi^* g_Y,\qquad k\ge2,$$9–$Z_y$0 framework on fibred cusp spaces. There, a $Z_y$1–$Z_y$2 metric is $Z_y$3, and the case $Z_y$4 has local form

$Z_y$5

which is identified as the local model of a cusp-type singularity. The same survey states that incomplete cusp edge metrics with exponent $Z_y$6 are exactly incomplete $Z_y$7–$Z_y$8 metrics with $Z_y$9 in suitable boundary defining functions (Grieser et al., 21 Jul 2025).

For comparison, Hunsicker and Rochon study complete quasi iterated fibred cusp metrics on the regular stratum of a smoothly stratified space, with local model

$x^k$0

These metrics are complete, whereas incomplete cusp edge metrics place the singular stratum at finite distance. Their role is analytic: they provide complete cusp-type models whose $x^k$1 theory realizes intersection cohomology and clarifies what is specific to incompleteness and what is specific to the link geometry (Hunsicker et al., 2012).

2. The Witt condition and its analytic formulations

For an oriented stratified pseudomanifold $x^k$2, if a stratum has link $x^k$3 of even dimension, the Witt condition is

$x^k$4

Analytically, this excludes middle-degree $x^k$5 harmonic forms on the links, which is the mechanism behind uniqueness of the middle perversity, Poincaré duality for intersection cohomology, and essential self-adjointness of the de Rham operator in the cusp-type framework (Hunsicker et al., 2012).

On incomplete cusp edge spaces, the Hodge–de Rham boundary family is the vertical family of Dirac-type operators on the fibres $x^k$6. For the Hodge–de Rham operator $x^k$7, the induced vertical operator is $x^k$8, whose kernel is the de Rham cohomology of $x^k$9. The Witt condition is therefore

$H^{\dim Z/2}(Z_y;\mathbb{R})=0$0

or equivalently the middle-degree part of $H^{\dim Z/2}(Z_y;\mathbb{R})=0$1 vanishes (Liu, 4 Aug 2025). The same source emphasizes that this eliminates dangerously small eigenvalues of the vertical operator which would otherwise create additional self-adjointness phenomena and boundary contributions in index formulas.

In the non-iterated incomplete cusp edge Hodge theory of Mazzeo and Vertman, where the singular space has a single stratum of codimension $H^{\dim Z/2}(Z_y;\mathbb{R})=0$2 and link $H^{\dim Z/2}(Z_y;\mathbb{R})=0$3 of dimension $H^{\dim Z/2}(Z_y;\mathbb{R})=0$4, the condition is written as

$H^{\dim Z/2}(Z_y;\mathbb{R})=0$5

The paper further explains that the middle-degree fibre-harmonic modes are exactly the borderline modes for the radial ODE, so their absence removes the indicial obstructions to essential self-adjointness (Gell-Redman et al., 2015).

A related spectral form appears in the survey of fibred cusp spaces: for incomplete cusp metrics ($H^{\dim Z/2}(Z_y;\mathbb{R})=0$6), essential self-adjointness of Dirac operators is tied to the geometric Witt condition

$H^{\dim Z/2}(Z_y;\mathbb{R})=0$7

This is the spectral-gap form of the Witt condition for the vertical family and matches the role played by small eigenvalues in cusp edge analysis (Grieser et al., 21 Jul 2025).

3. Hodge Laplacians, $H^{\dim Z/2}(Z_y;\mathbb{R})=0$8-cohomology, and intersection cohomology

For an incomplete cusp edge metric

$H^{\dim Z/2}(Z_y;\mathbb{R})=0$9

Mazzeo and Vertman analyze the Hodge Laplacian

$\dim Z$0

Under the Witt condition and the order assumption

$\dim Z$1

they prove that $\dim Z$2 is essentially self-adjoint on $\dim Z$3, and that its unique self-adjoint extension has discrete spectrum (Gell-Redman et al., 2015). The proof uses a heat-kernel construction on a blown-up heat space and a precise description of the asymptotics at the singular set.

The same work proves that $\dim Z$4-harmonic forms are polyhomogeneous conormal and satisfy the growth estimate

$\dim Z$5

near the singular stratum. The local $\dim Z$6 calculations on neighborhoods $\dim Z$7 match the local middle-perversity intersection cohomology groups, leading to the Hodge theorem

$\dim Z$8

where $\dim Z$9 is the stratified space obtained by collapsing the boundary fibres. When $k=3$0, so that $k=3$1 is homeomorphic to a smooth manifold, this reduces to ordinary de Rham cohomology (Gell-Redman et al., 2015).

The same paper also derives a small-time heat trace expansion. If

$k=3$2

then, as $k=3$3,

$k=3$4

so the cusp edge contributes a distinct boundary asymptotic governed by $k=3$5 and the base dimension $k=3$6 (Gell-Redman et al., 2015).

A complete counterpart is given by quasi iterated fibred cusp metrics. For a perversity $k=3$7, Hunsicker and Rochon choose

$k=3$8

and prove

$k=3$9

In the Witt case, the weight is trivial: $X$0 This gives a complete hyperbolic-cusp model realizing the same middle-perversity theory that appears analytically for incomplete cusp edge metrics (Hunsicker et al., 2012).

4. Dirac operators, heat kernels, Fredholm theory, and signature

For Dirac-type operators on incomplete cusp edge spaces, Gell-Redman and Piazza introduce ice Clifford modules $X$1 and associated operators

$X$2

Near the boundary, for a product-type metric and compatible connection, the local model is

$X$3

where the dominating vertical term is the Dirac operator on the fibres $X$4 (Liu, 4 Aug 2025).

Under the assumption that the boundary family $X$5 is invertible, they prove that $X$6 and $X$7 are essentially self-adjoint on $X$8, that their spectrum is purely discrete with Weyl asymptotics, and that eigenfunctions are smooth and vanish to infinite order at $X$9 (Liu, 4 Aug 2025). They further construct a Green’s operator in a cusp edge pseudodifferential calculus and identify the natural domain as

$X = \bigsqcup_i S_i,$0

so that

$X = \bigsqcup_i S_i,$1

is Fredholm.

The heat kernel of $X = \bigsqcup_i S_i,$2 is constructed on a blown-up heat space $X = \bigsqcup_i S_i,$3 with faces encoding the simultaneous limits $X = \bigsqcup_i S_i,$4, $X = \bigsqcup_i S_i,$5, and approach to the diagonal. The kernel is polyhomogeneous, and Getzler rescaling at both the temporal face and the cusp-edge front face yields a small-time supertrace expansion with interior coefficients reproducing the usual local index density and boundary coefficients expressed through the Bismut–Cheeger $X = \bigsqcup_i S_i,$6-form of the vertical family (Liu, 4 Aug 2025).

For a general Dirac-type operator with invertible boundary family, the resulting index formula is

$X = \bigsqcup_i S_i,$7

For the spin Dirac operator, this simplifies to

$X = \bigsqcup_i S_i,$8

and the paper notes that there is no additional transgression term of the type that appears in the incomplete edge case (Liu, 4 Aug 2025).

The Hodge–de Rham operator is the exceptional case in which the vertical family is never invertible, because harmonic forms on the fibres always produce kernel. In the Witt situation, however, one can combine the Hodge heat-kernel construction of Gell‑Redman–Swoboda with refined boundary Getzler rescaling. This yields the signature theorem: if $X = \bigsqcup_i S_i,$9 is a Witt incomplete cusp edge space with $\partial M$00, then

$\partial M$01

The same analysis shows that there is no extra contribution from the face $\partial M$02; a second Getzler rescaling forces any such supertrace contribution to vanish (Liu, 4 Aug 2025).

5. Weil–Petersson geometry and two-dimensional cusp-edge models

A particularly explicit two-dimensional model is provided by “Weil–Petersson type” metrics on punctured surfaces. Near each puncture, the metric is asymptotically modeled on the surface of revolution obtained by rotating

$\partial M$03

about the $\partial M$04-axis. If $\partial M$05 denotes the distance to the cusp, then the Gaussian curvature satisfies

$\partial M$06

and the local metric is consistent with

$\partial M$07

The level sets $\partial M$08 are circles of length behaving like $\partial M$09, and radial curves reach the cusp in finite length, so this is an incomplete cusp edge geometry in dimension two (Gadre et al., 2017).

This model has concrete dynamical consequences. Writing $\partial M$10, $\partial M$11 for the compatible almost complex structure, and

$\partial M$12

one obtains a quasi-Clairaut relation: for geodesic segments contained in a sufficiently small cusp neighborhood,

$\partial M$13

with $\partial M$14. Thus $\partial M$15 is almost constant along the excursion (Gadre et al., 2017).

The same paper proves strict convexity of the cusp distance along nonradial geodesics and the finite residence time estimate

$\partial M$16

For Liouville-almost every unit tangent vector $\partial M$17, the minimal distance to a fixed cusp up to time $\partial M$18 satisfies

$\partial M$19

so maximal cusp depth grows on polynomial, rather than logarithmic, scales (Gadre et al., 2017).

These two-dimensional models are directly tied to Weil–Petersson geometry. The modular surface $\partial M$20 has a cusp asymptotically modeled on $\partial M$21, and the higher-dimensional Weil–Petersson metric near a divisor has the form

$\partial M$22

which corresponds to $\partial M$23 with $\partial M$24 (Gadre et al., 2017, Liu, 4 Aug 2025). The two-dimensional cusp-excursion analysis therefore supplies an explicit local model for the same collapse rate that appears in the higher-dimensional incomplete cusp edge picture.

6. Signature classes, intersection spaces, and torsion-type extensions

The topological signature package for Witt spaces is supplied by Friedman and McClure. For a global $\partial M$25-Witt map $\partial M$26, they construct a symmetric Poincaré complex from the middle-perversity intersection chain complex on the induced cover and define

$\partial M$27

This class is compatible with Mishchenko’s symmetric signature on manifolds, is additive under disjoint union, multiplicative under products, invariant under stratified homotopy equivalence, and bordism invariant. In rational $\partial M$28-theory it agrees with the analytic signature index class of Albin–Leichtnam–Mazzeo–Piazza (Friedman et al., 2011). For Witt incomplete cusp edge spaces, this provides the topological signature class to which the analytic $\partial M$29-signature is compared.

A complementary topological model is given by Wrazidlo’s study of depth-one Witt spaces. If the rational Hurewicz map on each truncation cone is surjective, then the middle intersection space $\partial M$30 admits a Klimczak completion

$\partial M$31

to a rational Poincaré duality space. In dimension $\partial M$32, the signature of $\partial M$33 agrees with the Goresky–MacPherson intersection homology signature: $\partial M$34 This supplies a rational Poincaré-duality avatar of the same signature captured analytically by the Witt incomplete cusp edge signature operator (Wrazidlo, 2019).

Analytic torsion has been worked out most fully in the simpler cusp setting. On manifolds with cusps and unimodular Witt representations, Albin, Rochon, and Sher establish a Cheeger–Müller theorem using renormalized traces, model operators $\partial M$35 and $\partial M$36, and degeneration to a manifold with fibreed cusps. They relate analytic torsion to intersection $\partial M$37-torsion of the compactified stratified space, with explicit correction terms determined by the link (Albin et al., 2014). This is a depth-one, non-fibered model rather than a full cusp-edge theorem, but it isolates the same structural ingredients—renormalized heat traces, vertical and horizontal model operators, and Witt vanishing on the link—that govern torsion questions in the incomplete cusp edge setting.

Taken together, these results show that Witt incomplete cusp edge spaces admit a coherent analytic and topological theory. The geometry is controlled by anisotropic fibre collapse; the Witt condition removes the middle-degree link obstructions; $\partial M$38-harmonic forms realize middle-perversity intersection cohomology; Dirac and signature operators are subject to heat-kernel and index formulas with Bismut–Cheeger correction terms; and the analytic signature is matched by homotopy-invariant classes in $\partial M$39-theory and by rational Poincaré-duality models (Liu, 4 Aug 2025, Hunsicker et al., 2012, Friedman et al., 2011).