Courant Sharp Metrics on Surfaces

• Courant sharp metrics are Riemannian metrics where selected eigenfunctions reach the maximal nodal domain count allowed by Courant’s theorem.
• Perturbation methods reveal that smoothing nodal critical points typically does not increase the domain count, highlighting constraints in nodal geometry.
• Metric surgery constructions show every closed surface can support Courant sharp metrics up to any finite spectral level, impacting eigenfunction analysis.

Searching arXiv for papers on Courant sharp metrics and related nodal-domain perturbation results. Courant sharp metrics are Riemannian metrics for which Courant’s upper bound on nodal-domain count is realized at prescribed low spectral levels. For a closed surface $(M,g)$, an eigenpair $(\lambda_k,u_k)$ is called Courant sharp when $\mu(u_k)=k$, where $\mu(u_k)$ denotes the number of nodal domains of the Laplace eigenfunction $u_k$; a metric $g$ is called Courant sharp up to level $K$ if for each $k=1,\dots,K$ there is a choice of eigenfunction $u_k$ with $\mu(u_k)=k$ (Mukherjee et al., 7 Jul 2025). The underlying problem—determining when an eigenvalue admits an eigenfunction with as many nodal domains as its spectral label—was motivated by the analysis of minimal spectral partitions and has been studied on flat tori, triangles, Möbius strips, Klein bottles, and cylinders (Bérard et al., 2020).

1. Definitions and spectral setting

On a closed Riemannian surface $(\lambda_k,u_k)$0, the Laplace–Beltrami operator $(\lambda_k,u_k)$1 on $(\lambda_k,u_k)$2 has a discrete non-negative spectrum, repeated according to multiplicity,

$(\lambda_k,u_k)$3

with $(\lambda_k,u_k)$4-normalized real eigenfunctions $(\lambda_k,u_k)$5 satisfying

$(\lambda_k,u_k)$6

The nodal set

$(\lambda_k,u_k)$7

is a piecewise smooth $(\lambda_k,u_k)$8-dimensional submanifold, and its complement splits into connected components called nodal domains. Writing $(\lambda_k,u_k)$9 for the number of nodal domains, Courant’s Nodal Domain Theorem asserts

$\mu(u_k)=k$0

for the indexing convention above (Mukherjee et al., 7 Jul 2025).

In the compact-surface convention

$\mu(u_k)=k$1

counting multiplicities, an eigenvalue $\mu(u_k)=k$2 is Courant-sharp if there exists at least one eigenfunction in the $\mu(u_k)=k$3-th eigenspace having exactly $\mu(u_k)=k$4 nodal domains. Two immediate consequences are standard: $\mu(u_k)=k$5 and $\mu(u_k)=k$6 are always Courant-sharp, and if $\mu(u_k)=k$7 is Courant-sharp then $\mu(u_k)=k$8 (Bérard et al., 2020).

The 2025 perturbative formulation extends the notion from individual eigenpairs to the ambient geometry. A metric is Courant sharp up to level $\mu(u_k)=k$9 when the first $\mu(u_k)$0 levels all admit a choice of eigenfunction saturating Courant’s bound. This places the emphasis on the metric as an organizing object for nodal geometry rather than on a single fixed eigenspace (Mukherjee et al., 7 Jul 2025).

2. Metric perturbations and local maximality of nodal counts

A central perturbative framework considers a smooth one-parameter family of metrics $\mu(u_k)$1 on a closed surface, with $\mu(u_k)$2, under the assumption that $\mu(u_k)$3 is simple so that one can choose eigenfunctions $\mu(u_k)$4 depending smoothly on $\mu(u_k)$5 and converging in $\mu(u_k)$6 to $\mu(u_k)$7. If $\mu(u_k)$8 is a nodal critical point of $\mu(u_k)$9 of vanishing order $u_k$0, so that locally

$u_k$1

then the local structure of the perturbed nodal set is constrained in three ways: the number of nodal critical points of $u_k$2 in $u_k$3 is at most $u_k$4; no nodal critical point in $u_k$5 has vanishing order $u_k$6; and if

$u_k$7

while $u_k$8 is the analogous count for $u_k$9, then

$g$0

where $g$1 is the number of connected components of $g$2 (Mukherjee et al., 7 Jul 2025).

The corresponding global statement is a monotonicity theorem: under the same hypotheses, for each fixed $g$3 and small $g$4,

$g$5

If $g$6 has no nodal critical points, then for small $g$7 equality holds:

$g$8

The note summarizes this by stating that non-generic metrics, namely those with nodal critical points, locally maximize the nodal domain count of the perturbed eigenfunction branch, whereas generic metrics—being Morse functions without degenerate nodal zeros in the sense of Uhlenbeck—cannot gain nodal domains under small perturbations (Mukherjee et al., 7 Jul 2025).

A common misconception is that nodal complexity should increase under generic perturbation. The perturbative theorem gives the opposite local picture: the nodal-domain count is locally maximized at non-generic configurations, and smoothing away critical nodal degeneracies does not create additional nodal domains (Mukherjee et al., 7 Jul 2025).

3. Existence of Courant-sharp metrics on closed surfaces

The principal existence theorem states that for any positive integers $g$9 and $K$0, there exists a Riemannian metric $K$1 on the closed surface $K$2 of genus $K$3 such that the first $K$4 eigenvalues

$K$5

are simple and their normalized eigenfunctions satisfy

$K$6

Thus every closed surface admits metrics that are Courant sharp up to an arbitrary finite level (Mukherjee et al., 7 Jul 2025).

The construction proceeds by surgery. On the unit disc one uses an $K$7-invariant metric, referred to as Freitas’ metric, of fixed area whose first $K$8 Neumann eigenfunctions are simple and have exactly $K$9 nodal circles when $k=1,\dots,K$0. One then embeds a small geodesic disc in the target surface $k=1,\dots,K$1, replaces its metric by the rescaled Freitas metric, and shrinks the remainder of $k=1,\dots,K$2 by a factor $k=1,\dots,K$3. A continuous-spectrum argument shows that the first $k=1,\dots,K$4 global eigenvalues converge to those of the Freitas model, with $k=1,\dots,K$5 convergence of eigenfunctions on the inserted disc. By the global monotonicity theorem, the nodal domain counts cannot drop under this gluing, and each of the first $k=1,\dots,K$6 eigenfunctions on $k=1,\dots,K$7 ends up with exactly $k=1,\dots,K$8 domains (Mukherjee et al., 7 Jul 2025).

The same note records that this yields Courant sharp metrics on familiar closed surfaces, including the sphere $k=1,\dots,K$9, the torus $u_k$0, and higher genus surfaces. The construction is described as producing an explicit, piecewise-smooth, and then mollified metric on any genus-$u_k$1 surface realizing Courant sharpness up to arbitrary finite level (Mukherjee et al., 7 Jul 2025).

This result sharply contrasts with the fixed-geometry classification problem. On many explicit flat models, only the first one or two spectral levels are Courant-sharp. The existence theorem shows that this scarcity is not an intrinsic obstruction of topology alone; by varying the metric, one can realize Courant sharpness up to any prescribed finite depth (Mukherjee et al., 7 Jul 2025).

4. Localized perturbations and prescribed boundary nodal intersections

The perturbative machinery also yields a boundary prescription result for compact surfaces with boundary. If $u_k$2 is a compact surface of genus $u_k$3 with $u_k$4 boundary circles $u_k$5, and if $u_k$6 are given positive integers, then there exists a metric $u_k$7 and an index $u_k$8 such that the $u_k$9-th Neumann eigenfunction $\mu(u_k)=k$0 satisfies

$\mu(u_k)=k$1

Thus one can prescribe the even number of intersections of a Neumann eigenfunction with each boundary component (Mukherjee et al., 7 Jul 2025).

The construction again uses localized surgery. On each disc one picks an eigenfunction with $\mu(u_k)=k$2 intersections on $\mu(u_k)=k$3. A small generic perturbation makes the spectrum simple and preserves the boundary intersection pattern, since no critical intersections appear. Each perturbed disc is then attached to the $\mu(u_k)=k$4-th boundary component of $\mu(u_k)=k$5 by a thin neck of thickness $\mu(u_k)=k$6, while the remaining part of $\mu(u_k)=k$7 is scaled by $\mu(u_k)=k$8. By continuity of eigenvalues and eigenfunctions, together with the global monotonicity of nodal counts, the resulting Neumann eigenfunction on $\mu(u_k)=k$9 still has exactly $(\lambda_k,u_k)$00 intersections on $(\lambda_k,u_k)$01 (Mukherjee et al., 7 Jul 2025).

This boundary prescription problem is not identical to Courant sharpness, but the two are linked by the same perturbative principle: localized degeneracies and controlled gluing can be used to stabilize or preserve nodal topology. A plausible implication is that the notion of a Courant-sharp metric is part of a broader nodal-data engineering program on surfaces, in which both domain counts and boundary intersection data are controlled through metric design.

5. Explicit model geometries and complete classifications

Earlier work in the data concentrates on determining, for a fixed geometry, which eigenvalues are Courant-sharp. The cumulative picture is that complete classifications are often possible on flat or highly symmetric spaces, but the resulting list is usually very short.

Setting	Courant-sharp classification	Source
Closed surface of genus $(\lambda_k,u_k)$02	There exists a metric $(\lambda_k,u_k)$03 such that the first $(\lambda_k,u_k)$04 eigenvalues are simple and $(\lambda_k,u_k)$05 for $(\lambda_k,u_k)$06	(Mukherjee et al., 7 Jul 2025)
Flat Klein bottles $(\lambda_k,u_k)$07, $(\lambda_k,u_k)$08	Only $(\lambda_k,u_k)$09 and $(\lambda_k,u_k)$10 are Courant-sharp	(Bérard et al., 2020)
Flat cylinders $(\lambda_k,u_k)$11, $(\lambda_k,u_k)$12	Only the first and second Dirichlet eigenvalues are Courant-sharp	(Bérard et al., 2020)
Square Möbius strip $(\lambda_k,u_k)$13	Only the Dirichlet eigenvalues $(\lambda_k,u_k)$14 and $(\lambda_k,u_k)$15 are Courant-sharp	(Bérard et al., 2020)
Flat torus $(\lambda_k,u_k)$16	Only $(\lambda_k,u_k)$17 and $(\lambda_k,u_k)$18 are Courant-sharp; equivalently only the labels $(\lambda_k,u_k)$19	(Léna, 2015)
Three-dimensional square torus $(\lambda_k,u_k)$20	Only $(\lambda_k,u_k)$21 and $(\lambda_k,u_k)$22 are Courant-sharp	(Léna, 2015)
Equilateral torus and triangles	Equilateral torus: only first and second; equilateral triangle: first, second, and fourth; right-angled isosceles and hemiequilateral triangles: only first and second	(Bérard et al., 2015)

For the flat Klein bottle associated with the square torus, the distinct eigenvalues are exactly $(\lambda_k,u_k)$23 with $(\lambda_k,u_k)$24, together with the further condition “$(\lambda_k,u_k)$25 even if $(\lambda_k,u_k)$26.” The paper proves that on each of the flat Klein bottles $(\lambda_k,u_k)$27 arising from square fundamental domains, the only Courant-sharp eigenvalues are $(\lambda_k,u_k)$28 and $(\lambda_k,u_k)$29. It also treats flat cylinders

$(\lambda_k,u_k)$30

showing that the only Courant-sharp Dirichlet eigenvalues are again the first and second (Bérard et al., 2020).

The Möbius-strip result is analogous. For the square Möbius strip $(\lambda_k,u_k)$31, the Dirichlet spectrum consists of

$(\lambda_k,u_k)$32

and among all Dirichlet eigenvalues $(\lambda_k,u_k)$33, the only Courant-sharp ones are

$(\lambda_k,u_k)$34

The same paper emphasizes that $(\lambda_k,u_k)$35 has multiplicity $(\lambda_k,u_k)$36 and supports a two-parameter family of second eigensolutions with peculiar nodal patterns, including closed nodal loops, higher-order boundary zeros, and a nodal domain homeomorphic to a Möbius strip, while still yielding exactly two nodal domains (Bérard et al., 2020).

The toral examples illustrate the role of multiplicity. On $(\lambda_k,u_k)$37, the eigenvalues are

$(\lambda_k,u_k)$38

and the only Courant-sharp ones are the zero mode and the first nontrivial mode. Because the first nonzero eigenvalue has multiplicity $(\lambda_k,u_k)$39, this is equivalently stated as: the only Courant-sharp eigenvalues are $(\lambda_k,u_k)$40 with $(\lambda_k,u_k)$41 (Léna, 2015). On $(\lambda_k,u_k)$42, the only Courant-sharp eigenvalues are $(\lambda_k,u_k)$43 and $(\lambda_k,u_k)$44, where the latter has multiplicity $(\lambda_k,u_k)$45 and occupies the labels $(\lambda_k,u_k)$46 through $(\lambda_k,u_k)$47 (Léna, 2015).

6. Proof mechanisms, structural obstructions, and open directions

Across the explicit model spaces, the dominant proof architecture is Pleijel’s method. In the flat Klein-bottle and cylinder setting, the method combines a Faber–Krahn–type lower bound on the first Dirichlet eigenvalue of any sub-domain $(\lambda_k,u_k)$48 of small area,

$(\lambda_k,u_k)$49

with a lower bound on the Weyl counting function

$(\lambda_k,u_k)$50

From these inequalities one deduces that any Courant-sharp $(\lambda_k,u_k)$51 must satisfy a strict inequality of the form

$(\lambda_k,u_k)$52

and hence that $(\lambda_k,u_k)$53 is bounded above by an explicit constant. Only finitely many eigenvalues then need to be inspected (Bérard et al., 2020).

The remaining candidates are typically eliminated by multiplicity and symmetry considerations. On the Klein bottle, explicit identification under the involution describes exactly which torus eigenfunctions descend, which in turn determines multiplicities. In low-dimensional eigenspaces one can then count nodal domains directly. For example, for $(\lambda_k,u_k)$54, the eigenspace is $(\lambda_k,u_k)$55-dimensional, spanned by $(\lambda_k,u_k)$56 and $(\lambda_k,u_k)$57, and every nonzero linear combination $(\lambda_k,u_k)$58 has only $(\lambda_k,u_k)$59 nodal domains, so $(\lambda_k,u_k)$60 is not Courant-sharp. For $(\lambda_k,u_k)$61, a Sturm-type or barrier argument à la Stern shows that one cannot reach $(\lambda_k,u_k)$62 nodal domains. Analogous explicit descriptions on $(\lambda_k,u_k)$63 show that $(\lambda_k,u_k)$64 and $(\lambda_k,u_k)$65 likewise fail to be sharp (Bérard et al., 2020).

Multiplicity blocks impose a general obstruction. The Möbius-strip analysis states explicitly that if $(\lambda_k,u_k)$66 has multiplicity $(\lambda_k,u_k)$67, then at most the first label in the cluster can possibly be sharp; all subsequent labels in the repeated block fail automatically by Courant’s upper bound on the number of domains (Bérard et al., 2020). This same phenomenon underlies the torus examples, where a single eigenspace may occupy several consecutive labels even though the maximal nodal count realized inside that eigenspace remains small.

The perturbative theory adds a complementary obstruction: generic metrics do not create new nodal domains under small deformation. Theorem 2 states that $(\lambda_k,u_k)$68 for small $(\lambda_k,u_k)$69, with equality when $(\lambda_k,u_k)$70 has no nodal critical points. In this sense, non-generic metrics locally maximize nodal counts, and Courant-sharp metrics are tied to controlled nongenericity rather than generic behavior (Mukherjee et al., 7 Jul 2025).

The open directions recorded in the 2025 note are: extending to higher dimensions, where nodal sets are co-dimension-$(\lambda_k,u_k)$71 hypersurfaces; quantifying how “far” a generic metric is from being Courant sharp, for example by an upper bound on the largest $(\lambda_k,u_k)$72 for which $(\lambda_k,u_k)$73 can occur; studying Courant sharpness under other boundary conditions, including Dirichlet and mixed; and investigating the interplay of quantum ergodicity and Courant sharpness in the negative-curvature setting (Mukherjee et al., 7 Jul 2025). These questions delimit the present scope of the theory: complete classifications are available for several highly structured geometries, while the metric-design perspective shows that arbitrary finite-depth sharpness is nevertheless realizable on every closed surface.