🔢 mathematics

Estimates for harmonic functions near pseudo-corners

This paper investigates the blow-up of Sobolev norms for harmonic functions near pseudo-corners by employing inner expansions derived from generalized power series to establish explicit estimates for singularity coefficients and norms in terms of the approximation scale.

Original authors: Martin Costabel, Monique Dauge

Published 2026-07-22

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Martin Costabel, Monique Dauge

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are an architect designing a building. In the real world, walls usually meet at perfect 90-degree angles, or perhaps at gentle curves. But in the mathematical world of physics, we often study shapes with sharp, jagged points called "corners." When we try to solve equations that describe how heat spreads, how electricity flows, or how a drum vibrates on these shapes, things get tricky. The equations are usually very well-behaved on smooth surfaces, but near a sharp corner, the solution can start to act wild. It's like a calm river suddenly hitting a jagged rock; the water (or the mathematical value) might spike or become infinitely steep.

To understand these wild behaviors, mathematicians use a tool called "Sobolev norms." Think of a norm as a ruler that measures how "rough" or "smooth" a solution is. If the ruler says the number is small, the solution is smooth and predictable. If the number gets huge, the solution is getting messy. The big question is: What happens if we take a shape with a sharp corner and try to fix it by smoothing it out? Does the messiness disappear, or does it just get worse before it gets better? This is the puzzle of "singularities" in elliptic boundary value problems. It matters because real-world objects aren't perfect mathematical lines; they have tiny imperfections, and engineers need to know if those tiny imperfections will cause their calculations to blow up.

This paper, written by Martin Costabel and Monique Dauge, dives deep into exactly that scenario. They look at a specific type of sharp corner (a "pseudo-corner") and ask: If we replace that sharp point with a tiny, smooth curve (or a cluster of smaller corners), how does the "roughness" of the solution change as that curve gets smaller and smaller?

The authors prove that if the original sharp corner was "bad" enough (mathematically speaking, if the solution had a certain type of singularity there), then smoothing it out doesn't make the problem go away quietly. Instead, as the smoothing gets finer and finer (represented by a tiny number $\epsilon$ ), the "roughness" of the solution explodes. It doesn't just get a little bigger; it grows at a specific, predictable rate, like a snowball rolling down a hill. They show that the size of this explosion depends on the power of that tiny smoothing number $\epsilon$ .

Here is the twist they discovered: It matters how you measure the roughness. In the world of mathematics, there are different ways to define a "norm" (a way to measure size). The authors found that if you use the "natural" way to measure the roughness of these smoothed-out shapes, the number goes to infinity as the shape gets closer to the sharp corner. However, if you use a different, more abstract way of measuring (called an "interpolation norm"), the number stays calm and bounded. It's as if you have a storm that looks terrifying from one angle but looks perfectly calm from another. The paper proves that the "storm" is real in the natural measurement, and the calmness in the other measurement is just an illusion created by the way the ruler is constructed.

They also looked at a different scenario: instead of smoothing a corner into a curve, what if you replace one big sharp corner with two or three smaller sharp corners that are very close together? They found that the "roughness" still explodes, but the rate of the explosion changes depending on the angles of these new, smaller corners. The paper provides precise formulas for exactly how fast these numbers grow, showing that the geometry of the corner—whether it's a single sharp point or a cluster of them—dictates exactly how the mathematical solution behaves as it approaches the limit.

In short, the paper doesn't just say "it gets messy." It gives a precise recipe for how messy it gets, proving that the "blow-up" is inevitable if the original corner was singular enough, and that the choice of measurement tool can make the difference between seeing a manageable problem and a mathematical explosion. They didn't just guess this; they derived it using a powerful technique involving "converging series," which allowed them to write the solution as a sum of terms that get smaller and smaller, letting them spot the exact power of the explosion.

Technical Summary: Estimates for Harmonic Functions Near Pseudo-Corners

Problem Statement
This paper investigates the behavior of Sobolev norms for solutions to the Dirichlet problem for the Laplace operator ( $\Delta u = f$ ) when the domain $\Omega$ possesses a conical singularity (a corner) and is approximated by a family of more regular domains $\Omega_\varepsilon$ . The study focuses on two specific scenarios:

Rounding of a corner: A singular corner in a domain of arbitrary dimension $n$ is replaced by a smooth boundary within a small neighborhood of size $\varepsilon$ .
Approaching corners in 2D: A non-convex polygonal corner with opening angle $\omega$ is replaced by a cluster of $L$ smaller non-convex corners with opening angles $\varpi_\ell < \omega$ , separated by distances of order $\varepsilon$ .

The central question is how the Sobolev norms of the solutions $u_\varepsilon$ behave as $\varepsilon \to 0$ , particularly when the regularity index of the norm exceeds the threshold defined by the first singular exponent of the limiting corner. The authors aim to quantify the "blow-up" of these norms and the coefficients of singular functions in terms of powers of the approximation scale $\varepsilon$ .

Methodology
The primary technique employed is the use of convergent generalized power series expansions in the parameter $\varepsilon$ , building upon recent results established in [6]. Unlike previous asymptotic expansion techniques that often rely on formal series or specific matching conditions, this approach utilizes a rigorous "inner expansion" derived from a two-scale "cross-cutoff Ansatz."

Key methodological steps include:

Decomposition: The solution $u_\varepsilon$ is represented as a sum of a regular part and a singular part involving scaled canonical functions $K_j^+$ defined on the pattern domain $P$ (the shape used to round the corner).
Operator Analysis: The problem is reformulated using a block matrix operator $M[\varepsilon]$ acting on the pair of functions defined on the original domain and the pattern domain. The inverse of this operator is expanded as a Neumann series, yielding a convergent series of operators with exponents in a monoid $E_\infty$ generated by the singular exponents $\lambda_j^\pm$ .
Inner Expansion: The authors utilize an intrinsic "inner expansion" (Theorem 3.2) which expresses the solution in the boundary layer $\varepsilon Q$ as a series involving the coefficients $c_j(u_0)$ of the limiting solution $u_0$ and the canonical functions $K_j^+$ .
Norm Estimation: By analyzing the scaling properties of these series terms under the change of variables $x \mapsto x/\varepsilon$ , the authors derive lower and upper bounds for the Sobolev seminorms.

Key Contributions and Results

Blow-up in Smooth Pseudo-Corners (Section 4):
- For a domain where a corner is rounded by a smooth pattern, the authors prove that if the limiting solution $u_0$ has a non-zero coefficient $c_1(u_0)$ for the first singular function, the Sobolev seminorm $|u_\varepsilon|_{W^{m,p}(\varepsilon Q)}$ blows up as $\varepsilon \to 0$ provided the regularity index satisfies $m - n/p > \lambda_1^+$ .
- Theorem 4.1 provides an explicit lower bound:
  $|u_\varepsilon|_{W^{m,p}(\varepsilon Q)} \geq \varepsilon^{\lambda_1^+ + n/p - m} \left( |c_1(u_0)| |K_1^+|_{W^{m,p}(Q)} - \gamma_0 \varepsilon^{\lambda'} \|f\|_{L^2} \right)$
  where $\lambda_1^+$ is the smallest singular exponent. This confirms that the blow-up rate is exactly $\varepsilon^{\lambda_1^+ + n/p - m}$ .
- The result extends to fractional order Sobolev spaces $W^{s,p}$ .
Blow-up in Polygons with Approaching Corners (Section 5):
- In the 2D case where a large non-convex corner (angle $\omega$ ) is replaced by smaller non-convex corners (angles $\varpi_\ell$ ), the authors analyze the $H^s$ regularity.
- Theorem 5.2 establishes that if the regularity $s$ lies in the interval $(\pi/\omega + 1, \pi/\varpi + 1)$ , the Sobolev seminorms of the solution blow up as $\varepsilon \to 0$ . The rate is determined by the difference between the singular exponent of the large corner and the regularity index.
- Theorem 5.5 further decomposes the solution into a regular part and singular parts associated with the new small corners. It shows that both the regular part's norm and the coefficients of the new singular functions blow up as $\varepsilon \to 0$ if $c_1(u_0) \neq 0$ .
Operator Norms and Interpolation (Section 5.3):
- The paper highlights a crucial distinction between different definitions of solution operators in fractional Sobolev spaces.
- When considering the standard operator $T_\sigma: H^{-1+\sigma} \to H^{1+\sigma}$ , the operator norm blows up as $\varepsilon \to 0$ for certain ranges of $\sigma$ .
- However, if one defines the operator via interpolation of the spaces where the solution is known to be regular (the "restricted" source space $A^\heartsuit_\sigma$ ), the operator norm remains uniformly bounded.
- Significance: This demonstrates that while the norms in the source and target spaces are equivalent for a fixed domain, the constants in these equivalences depend on the domain geometry in a way that is not uniform with respect to $\varepsilon$ . Specifically, the interpolation norm can remain bounded while the "natural" norm explodes.
Inverse Estimate (Theorem 5.6):
- Conversely, if the solutions $u_\varepsilon$ are assumed to have $H^2$ regularity (i.e., the right-hand side $f$ is chosen such that singularities are suppressed in the approximating domains), the coefficient $c_1(u_0)$ of the singular function in the limiting domain is shown to be bounded by $C \varepsilon^{1-\pi/\omega} \|f\|_{L^2}$ . This implies that for the limiting solution to have $H^2$ regularity, the forcing term must vanish at a specific rate relative to $\varepsilon$ .

Significance and Claims
The authors claim that their work provides "transparent and easy to understand examples" for discussing potential theory on Lipschitz domains. The primary significance lies in:

Precision: Providing explicit, simple estimates for the blow-up of Sobolev norms and singularity coefficients using convergent series, rather than just asymptotic orders.
Norm Dependence: Illustrating that the regularity of the solution operator depends not only on the corner angles (the Lipschitz constant) but also on the distance between corners (the parameter $\varepsilon$ ).
Interpolation Theory: Demonstrating that in families of domains approaching a singularity, the equivalence between different definitions of fractional Sobolev norms (e.g., interpolation norms vs. natural norms) is non-uniform. This clarifies why operator norms may appear bounded in one framework but unbounded in another, resolving apparent contradictions in the literature regarding regularity thresholds.

The paper restricts its scope to the Dirichlet problem for the Laplacian to maintain simplicity and precision, noting that while more general elliptic problems with variable coefficients can be treated via asymptotic expansions (as in [15]), the convergent expansion method offers superior explicitness for this specific case.

More like this