---
title: Steklov-Neumann Eigenproblem Analysis
url: https://www.emergentmind.com/topics/steklov-neumann-eigenproblem
type: topic
---

# Steklov-Neumann Eigenproblem Analysis

The Steklov–Neumann eigenproblem denotes a family of boundary spectral problems in which the spectral parameter appears in a boundary condition of Steklov type on only part of the boundary, while the complementary part carries a homogeneous Neumann condition. In the weighted formulation,
$$
\begin{cases}
\Delta u=0 & \text{in }\Omega,\\
\partial_\nu u=\sigma \rho\,u & \text{on }\partial\Omega,
\end{cases}
$$
the choice $\rho\equiv 1$ on a distinguished boundary portion and $\rho\equiv 0$ on the complement produces the mixed Steklov–Neumann, or sloshing, problem. A second, closely related line of work interprets Steklov eigenvalues as limits of Neumann eigenvalues when interior mass concentrates in a thin boundary layer, so that Steklov spectra arise as critical boundary-concentration endpoints of Neumann families [1411.6567, 1410.0517].

## 1. Mixed boundary formulations

The standard mixed formulation is obtained by decomposing the boundary as
$$
\partial\Omega=F\cup B,\qquad F\cap B=\varnothing,
$$
and imposing
$$
\begin{cases}
\Delta u=0 & \text{in }\Omega,\\
\partial_\nu u=\sigma u & \text{on }F,\\
\partial_\nu u=0 & \text{on }B.
\end{cases}
$$
In the survey literature this is identified as a special case of the weighted Steklov problem, and specifically as the sloshing problem when $F$ is the free surface and $B$ the walls of the container [1411.6567].

A particularly important geometric realization occurs on doubly connected domains. If
$$
\partial\Omega=\Gamma_1\cup\Gamma_2,
$$
with $\Gamma_1$ the inner boundary and $\Gamma_2$ the outer one, the mixed Steklov–Neumann problem takes the form
$$
\begin{cases}
\Delta u = 0 & \text{in } \Omega,\\
\dfrac{\partial u}{\partial \nu}=0 & \text{on } \Gamma_1,\\
\dfrac{\partial u}{\partial \nu}=\mu u & \text{on } \Gamma_2.
\end{cases}
$$
Its weak formulation is
$$
\int_\Omega \nabla u\cdot \nabla v\, dV = \mu \int_{\Gamma_2} uv\, dS,\qquad \forall v\in H^1(\Omega),
$$
and the spectrum is discrete:
$$
0=\mu_0(\Omega)<\mu_1(\Omega)\le \mu_2(\Omega)\le \cdots \nearrow +\infty.
$$
The first nontrivial eigenvalue is characterized by
$$
\mu_1(\Omega) = \inf\left\{ \frac{\displaystyle\int_\Omega |\nabla u|^2\, dV} {\displaystyle\int_{\Gamma_2} u^2\, dS} :\; u\in H^1(\Omega),\  \int_{\Gamma_2}u\, dS=0 \right\},
$$
so orthogonality is imposed on the Steklov part of the boundary rather than in the bulk [2603.25448].

This mixed formulation should be distinguished from the pure Steklov problem on the full boundary, but it is not separate from weighted Steklov theory. In the framework
$$
\partial_\nu u=\sigma \rho u \quad \text{on }\partial\Omega,
$$
mixed Steklov–Neumann conditions are encoded by a weight that vanishes on the Neumann portion. This equivalence is one of the main reasons the Steklov–Neumann problem is often treated inside weighted Steklov spectral geometry rather than as an isolated boundary value problem [1411.6567].

## 2. Operator-theoretic and variational structure

For the classical Steklov problem on a compact Riemannian manifold with boundary $M=\partial\Omega$,
$$
\begin{cases}
\Delta u = 0 & \text{in } \Omega,\\
\partial_\nu u = \sigma u & \text{on } M,
\end{cases}
$$
the spectrum coincides with that of the Dirichlet-to-Neumann operator
$$
\Lambda f=\partial_\nu(Hf),
$$
where $Hf$ is the harmonic extension of $f$. In the smooth case, $\Lambda$ is a first-order elliptic pseudodifferential operator with the same principal symbol as $\sqrt{\Delta_M}$. This viewpoint extends directly to mixed problems by restricting the spectral boundary condition to a distinguished subset of the boundary [1411.6567].

The spectrum is discrete whenever the trace operator
$$
H^1(\Omega)\to L^2(\partial\Omega)
$$
is compact; the survey states that this holds, for instance, if $\Omega$ has Lipschitz boundary. For the pure Steklov problem the min–max principle is
$$
\sigma_k(\Omega,g)=\min_{E\in \mathcal E(k)}\ \sup_{0\neq u\in E} \frac{\int_\Omega |\nabla u|^2\,dA}{\int_M u^2\,dS},
$$
with orthogonality to constants on the boundary. In the mixed setting on doubly connected domains, the corresponding denominator is supported only on the Steklov boundary $\Gamma_2$ [1411.6567, 2603.25448].

A concrete model reduction occurs for the square. By diagonal symmetries, the Steklov problem on the square decomposes into four mixed Steklov problems on a right isosceles triangle; in each reduced problem the Steklov condition is imposed on the hypotenuse, and Dirichlet or Neumann conditions are imposed on the legs according to parity under reflection. One of these reductions is explicitly a sloshing, hence mixed Steklov–Neumann, problem. In the even–even class, the restrictions to the hypotenuse are eigenfunctions of the free beam equation
$$
\frac{d^4}{dx^4}f=\omega^4 f \qquad \text{on }(-1,1),
$$
with boundary conditions
$$
\frac{d^3}{dx^3}f=\frac{d^2}{dx^2}f=0 \qquad \text{at }x=-1,1,
$$
showing that mixed Steklov–Neumann problems can sometimes be reduced to one-dimensional self-adjoint spectral problems [1411.6567].

## 3. Neumann-to-Steklov limits

A distinct but mathematically adjacent meaning of the Steklov–Neumann theme arises from boundary mass concentration. Let
$$
\Omega_\varepsilon=\{x\in \Omega:\ d(x,\partial\Omega)<\varepsilon\},
$$
and define the concentrating density
$$
\rho_\varepsilon(x)= \begin{cases} \varepsilon, & x\in \Omega\setminus \overline{\Omega_\varepsilon},\\[2mm] \displaystyle \frac{M-\varepsilon |\Omega\setminus \Omega_\varepsilon|}{|\Omega_\varepsilon|}, & x\in \Omega_\varepsilon.
\end{cases}
$$
Then one studies the Neumann problem
$$
\begin{cases}
-\Delta u=\lambda \rho_\varepsilon u & \text{in }\Omega,\\
\partial_\nu u=0 & \text{on }\partial\Omega.
\end{cases}
$$
For each fixed $u\in H^1(\Omega)$, the corresponding Rayleigh quotient converges to the Steklov quotient, and for bounded $C^2$ domains one has
$$
\lim_{\varepsilon\to 0}\lambda_j(\varepsilon)=\lambda_j \qquad \forall j\in\mathbb N.
$$
Thus the Steklov spectrum is the spectral limit of Neumann problems with fixed total mass concentrating near the boundary [1410.0517, 1602.06078].

On the unit ball this limit is explicit. For constant boundary density $\rho=M/\sigma_N$, the Steklov eigenvalues are
$$
\lambda_l=\frac{l\,\sigma_N}{M}=\frac{lN\omega_N}{M},\qquad l\in\mathbb N,
$$
and the eigenfunctions are homogeneous harmonic polynomials of degree $l$. Separation of variables and Bessel-function analysis yield the first-order expansion
$$
\lambda_l(\varepsilon)= \lambda_l+\left(\frac{2l\lambda_l}{3}+\frac{2\lambda_l^2}{N(2l+N)}\right)\varepsilon+o(\varepsilon ),
$$
hence
$$
\lambda'_l(0)= \frac{2l\lambda_l}{3}+\frac{2\lambda_l^2}{N(2l+N)}>0 \qquad (l\ge 1).
$$
Therefore every positive branch is strictly increasing for sufficiently small $\varepsilon>0$, and the limiting Steklov eigenvalue is a local minimizer of the corresponding Neumann branch [1602.06078].

This asymptotic picture explains the description of Steklov eigenvalues as “critical Neumann eigenvalues.” They are not merely formal boundary analogues of Neumann spectra; they are realized as singular endpoint values of Neumann problems, and on the ball they sit at the bottom of nearby concentrating-mass branches [1410.0517].

## 4. Annuli, holes, and symmetry

The mixed Steklov–Neumann problem is especially tractable on concentric annuli. For
$$
\Omega_0=B_{R_2}\setminus \overline{B_{R_1}},
$$
the eigenvalues are
$$
\mu_l(\Omega_0) = \frac{l(l+n-2)\left(\left(\frac{R_2}{R_1}\right)^{2l+n-2}-1\right)} {R_2\left((l+n-2)\left(\frac{R_2}{R_1}\right)^{2l+n-2}+l\right)}, \qquad l\ge 0,
$$
with eigenfunctions
$$
w_l^j(r,\theta) = \left( r^l+\frac{lR_1^{2l+n-2}}{(l+n-2)\,r^{l+n-2}} \right)Y_l^j(\theta),
$$
where $Y_l^j$ are spherical harmonics of degree $l$. The first nonzero eigenvalue corresponds to $l=1$, has multiplicity $n$, and the associated eigenfunctions can be written in Cartesian form [2603.25448].

Among eccentric annuli of the form
$$
\Omega_d = B_{R_2}(d)\setminus \overline{B_{R_1}},
$$
with fixed radii and $\overline{B_{R_1}}\subset B_{R_2}(d)$, the first nonzero mixed eigenvalue satisfies
$$
\mu_1(\Omega_d)\le \mu_1(\Omega_0),
$$
with equality only for the concentric configuration. This is a sharp extremal result for the first mixed Steklov–Neumann eigenvalue in that class [2603.25448].

When the inner spherical hole shrinks,
$$
\Omega_r=\Omega_{\mathrm{out}}\setminus \overline{B_r},
$$
one has
$$
\lim_{r\to 0}\mu_1(\Omega_r)=\sigma_1(\Omega_{\mathrm{out}}),
$$
where $\sigma_1(\Omega_{\mathrm{out}})$ is the first nontrivial Steklov eigenvalue of the unperforated outer domain. After harmonic extension into the hole, suitably normalized eigenfunctions converge strongly in $H^1(\Omega_{\mathrm{out}})$. The corresponding first mixed eigenfunction has exactly two nodal domains [2603.25448].

Higher mixed eigenvalues on domains with holes also admit annular comparison under symmetry assumptions. If
$$
\tilde{\Omega}=\Omega_{out}\setminus \overline{B_{R_1}}
$$
and $\Omega_{out}$ is connected, smooth, centered at the origin, and symmetric of order $4$, then for the concentric annulus
$$
\tilde{\Omega}_0=B_{R_2}\setminus \overline{B_{R_1}}
$$
of the same volume,
$$
\mu_k(\tilde{\Omega})\le \mu_k(\tilde{\Omega}_0)=\mu_1(\tilde{\Omega}_0), \qquad 1\le k\le n.
$$
The proof uses the $l=1$ annular eigenfunctions together with cancellation of mixed moments enforced by order-$4$ symmetry. The same work gives counterexamples with only order-$2$ symmetry, so the symmetry assumption is structurally essential rather than cosmetic [2412.17124].

## 5. Perturbation, optimization, and criticality

One optimization problem varies the boundary partition while keeping the geometry fixed. Starting from a decomposition
$$
\partial\Omega=\overline{\Gamma_S\cup\Gamma_N},
$$
one inserts a small interval $\gamma_\varepsilon\subset\Gamma_S$ of length $2\varepsilon$ and replaces it by Neumann boundary:
$$
(\Gamma_S,\Gamma_N)\mapsto (\Gamma_S\setminus \overline{\gamma_\varepsilon},\; \Gamma_N\cup \gamma_\varepsilon).
$$
If $\lambda_j^0$ is an eigenvalue of multiplicity $N_j$ and at least one eigenfunction in the eigenspace does not vanish at the center $c_\star$ of the inserted interval, then
$$
\lambda_j^\varepsilon-\lambda_j^0 = 2\varepsilon\,\lambda_j^0 \sum_{i=0}^{N_j-1}\big(u_{j+i}^0(c_\star)\big)^2 + O(\varepsilon^2).
$$
The associated Green’s function satisfies
$$
S_\varepsilon^\lambda(x_0,y) = S_0^\lambda(x_0,y) + 2\varepsilon\,\lambda\, S_0^\lambda(x_0,c_\star)\, S_0^\lambda(y,c_\star) + O(\varepsilon^2),
$$
which yields an algorithm for placing and enlarging Neumann pieces so that a prescribed parameter becomes close to a Steklov–Neumann eigenvalue and the Green’s function becomes large [1907.11147].

A different optimization theory varies the boundary mass density in the weighted Steklov problem
$$
\partial_\nu u=\lambda \rho u \quad \text{on }\partial\Omega.
$$
For a finite cluster $F$ of eigenvalues, the elementary symmetric functions $\Lambda_{F,h}[\rho]$ are real-analytic on the noncollision set $\mathcal R[F]$, and their Fréchet derivatives are
$$
d\Lambda_{F,h}[\rho][\dot\rho] = -\sum_{k=1}^n c_k \sum_{l\in F_k} \int_{\partial\Omega} (\operatorname{Tr}u_l)^2 \,\dot\rho\, d\sigma.
$$
Under fixed total boundary mass
$$
M[\rho]=\int_{\partial\Omega}\rho\,d\sigma,
$$
critical densities are characterized by
$$
\sum_{k=1}^n c_k \sum_{l\in F_k} (\operatorname{Tr}u_l)^2 = c \qquad \text{a.e. on }\partial\Omega.
$$
On the unit ball, the constant density
$$
\rho=\frac{M}{|\partial\Omega|}
$$
is critical for all $\Lambda_{F,h}$ [1410.0517].

This last fact is one of the main structural differences between Steklov and standard Dirichlet–Neumann spectral optimization. The paper explicitly contrasts the Steklov case with a “maximum principle” for many other elliptic spectral optimization problems, where analogous critical density conditions are typically impossible. In the Steklov setting the condition lives on the boundary, and in highly symmetric domains such as the ball it can be satisfied. This is another sense in which Steklov eigenvalues behave as critical Neumann eigenvalues rather than as ordinary boundary counterparts of Dirichlet or Neumann spectra [1410.0517].

## 6. Computation and broader generalizations

For the classical Steklov operator on polygonal planar domains, conforming finite elements can produce certified lower bounds. The key abstract estimate is
$$
\lambda_k \ge \frac{\lambda_{k,h}}{1+M_h^2\lambda_{k,h}},
$$
where $\lambda_{k,h}$ is the conforming finite-element eigenvalue and $M_h$ is a computable projection-error constant obtained from an auxiliary nonhomogeneous Neumann problem and a hypercircle construction. In the local analysis of the trace constant, the paper also introduces a mixed Steklov–Neumann problem on a triangle, showing that even certification techniques for pure Steklov spectra naturally generate mixed Steklov–Neumann subproblems [2001.09820].

Boundary-only methods are also available. For smooth simply connected planar domains, the Dirichlet-to-Neumann operator can be written in terms of a generalized conjugation operator:
$$
\Lambda\gamma(t)=\rho(t)\,D E\gamma(t).
$$
After Nyström discretization of the boundary integral equation and Fourier differentiation, one obtains a dense algebraic eigenvalue problem
$$
Q\mathbf{x}=\lambda \mathbf{x},\qquad Q=PDE,
$$
or, in the notation of the paper,
$$
Q=PFWF^\ast E.
$$
This formulation treats interior and exterior Steklov problems in a unified way and reconstructs eigenfunctions by harmonic extension from their boundary traces [2604.05975].

The Neumann-to-Steklov principle extends far beyond the linear Laplacian. For admissible possibly irregular domains obtained from the unit ball by trace-compatible Sobolev homeomorphisms, the first nontrivial weighted $(p,q)$-Neumann eigenvalue with concentrating bulk weight $\gamma_a$ satisfies
$$
\Lambda^N_{p,q}(\gamma_a)\to \Lambda^{St}_{p,q}(\beta),
$$
and normalized minimizers converge, up to subsequences, strongly in $W^{1,p}(\Omega)$ to weighted Steklov minimizers. Equivalently, the best constants in weighted Poincaré inequalities converge to the best constants in weighted trace inequalities [2605.09759].

Beyond the Laplace setting, related Steklov-type boundary spectra appear for the modified Helmholtz equation, where polygonal corners numerically produce asymptotics
$$
\mu_k^{(p)} \simeq c_k \sqrt{p},
$$
and for Maxwell’s equations, where the boundary relation
$$
\nu\times\operatorname{curl}u=\lambda u \quad \text{on }\Gamma
$$
defines a compact NtD-based Steklov-type theory for tangential fields. Discrete analogues also exist on finite subgraphs of $\mathbb Z^n$, where the Dirichlet-to-Neumann operator defines a lattice Steklov spectrum [2310.19571, 2007.10765, 1902.05831]. These developments suggest that the Steklov–Neumann paradigm is best understood not as a single boundary value problem, but as a broad spectral mechanism linking boundary flux operators, mixed boundary partitions, and Neumann-to-boundary concentration limits.

Source: https://www.emergentmind.com/topics/steklov-neumann-eigenproblem