Sharp conditions for the BBM formula and asymptotics of heat content-type energies

Basic Information

• Paper ID: 2502.14655
• Title: Sharp conditions for the BBM formula and asymptotics of heat content-type energies
• Authors: Luca Gennaioli (University of Warwick), Giorgio Stefani (Università degli Studi di Padova)
• Classification: math.AP (Analysis of PDEs), math.FA (Functional Analysis)
• Submission Date: November 7, 2025 (arXiv v2)
• Paper Link: https://arxiv.org/abs/2502.14655

Abstract

This paper establishes necessary and sufficient conditions for families of non-negative measurable kernels $(ρ_t)_{t\in(0,1)}$ with $p \in [1,\infty)$, ensuring that the associated Bourgain-Brezis-Mironescu (BBM) energies $(\mathscr{F}_{t,p})_{t\in(0,1)}$ converge to a variant of the $p$-Dirichlet energy on $\mathbb{R}^N$ as $t\to 0^+$, in both pointwise and Γ-convergence senses. The authors provide sufficient conditions on $(ρ_t)_{t\in(0,1)}$ ensuring local compactness in $L^p(\mathbb{R}^N)$ of sequences with bounded BBM energy. Additionally, when the limiting $p$-energy is of non-local type, the paper provides sufficient conditions guaranteeing pointwise convergence, Γ-convergence, and compactness. Finally, these results are applied to asymptotic formulas for heat content-type energies, covering both local and non-local cases.

Research Background and Motivation

Core Problem

This paper investigates the asymptotic behavior of non-local functional energies, particularly BBM-type energies: $\mathscr{F}_{t,p}(u) = \int_{\mathbb{R}^N} \int_{\mathbb{R}^N} \frac{|u(x)-u(y)|^p}{|x-y|^p} ρ_t(x-y) dx dy$ whether they converge to the classical $p$-Dirichlet energy as the parameter $t \to 0^+$.

Research Significance

1. Theoretical Importance: The BBM formula establishes a bridge between non-local functionals and classical Sobolev spaces, serving as a fundamental tool in modern analysis
2. Applied Value: Extensive applications in image denoising, materials science (antisymmetric exchange interactions), convergence of non-local Ginzburg-Landau functionals, and related fields
3. Physical Background: Heat content describes the evolution of heat over time in a perfectly insulated container, with its asymptotic behavior encoding geometric information

Limitations of Existing Methods

1. Only Sufficient Conditions: Since the pioneering work of Bourgain-Brezis-Mironescu 20, the literature has primarily provided sufficient conditions for the BBM formula
2. Restrictions of Fourier Methods: Davoli et al. 34 obtained necessary and sufficient conditions for $p=2$ using Fourier transforms, but this method cannot be extended to $p \neq 2$
3. Radial Symmetry Assumptions: Many results (e.g., 38) apply only to radially symmetric kernel families, limiting applicability
4. Lack of Compactness Analysis: For non-radially symmetric kernel families, equicoercivity results are nearly absent

Research Motivation

1. Answer the conjecture posed in 34: provide necessary and sufficient conditions for the BBM formula for all $p \in [1,\infty)$
2. Extend necessary and sufficient conditions from pointwise convergence to Γ-convergence (a stronger variational convergence concept)
3. Establish compactness criteria for non-radially symmetric kernel families
4. Provide unified treatment of asymptotic behavior for heat kernels and fractional heat kernels

Core Contributions

The main contributions of this paper include:

1. Necessary and Sufficient Conditions Theorem (Theorem 1.1): For any $p \in [1,\infty)$, provides necessary and sufficient conditions for the BBM formula, completely resolving the conjecture in 34, and extending results from pointwise convergence to Γ-convergence
2. Coercivity Theorem (Theorem 1.2): Under the maximal rank assumption, proves coercivity of the functional family, ensuring that limits of convergent sequences belong to appropriate Sobolev spaces
3. General Compactness Criterion (Theorem 1.3): For special kernel families (4.4), provides sufficient conditions for local compactness without requiring radial symmetry
4. Non-local to Non-local Convergence (Theorems 1.4-1.5): When the limiting energy remains non-local, provides sufficient conditions for pointwise/Γ-convergence and compactness
5. Heat Kernel Applications:
   • Theorem 1.6: Asymptotic formula for classical heat kernels (with Γ-convergence and compactness)
   • Theorem 1.7: Complete analysis of fractional heat kernels, distinguishing three cases: $2s > p$, $2s = p$, $2s < p$
   • Theorem 6.2: Generalizes 2 to general heat-type kernels
6. Hilbert Space Framework (Theorem 1.8): For strongly continuous symmetric operator semigroups, provides unified treatment of heat content asymptotics

Methodology Details

Task Definition

Core Task: Given a kernel family $(ρ_t)_{t \in (0,1)} \subset L^1_{loc}(\mathbb{R}^N)$ with $ρ_t \geq 0$ and $p \in [1,\infty)$, analyze the convergence of the functional $\mathscr{F}_{t,p}(u) = \int_{\mathbb{R}^N} \frac{\|u(\cdot + z) - u\|^p_{L^p}}{|z|^p} ρ_t(z) dz$ as $t \to 0^+$, including:

1. Whether it converges to the local $p$-Dirichlet energy variant $\mathscr{D}^μ_p(u) = \int_{S^{N-1}} \|σ \cdot Du\|^p_{L^p} dμ(σ)$
2. Convergence modes (pointwise, Γ-convergence)
3. Compactness (precompactness of sublevel sets)

Main Methodological Architecture

1. Proof Strategy for Necessary and Sufficient Conditions (Theorem 1.1)

Sufficiency Direction (A)⇒(B):

• Step 1: Prove a stronger result (Theorem 3.1): condition (A) ensures convergence to the generalized energy $\mathscr{G}^{μ,ν}_p(u) = \int_{S^{N-1}} \|σ \cdot Du\|^p_{L^p} dμ(σ) + \int_{\mathbb{R}^N \setminus \{0\}} \frac{\|u(\cdot+z)-u\|^p_{L^p}}{|z|^p} dν(z)$ where $μ \in \mathcal{M}^+(S^{N-1})$, $ν \in \mathcal{M}^+(\mathbb{R}^N)$
• Step 2: Utilize measure decomposition from Lemma 2.9:
  • $ν_k = ρ_{t_k} \mathcal{L}^N \stackrel{*}{\rightharpoonup} ν$ (weak* convergence)
  • $μ$ defined through spherical integral limits: $μ^δ_t(E) = \int_E \left(\int^δ_0 ρ_t(σr) r^{N-1} dr\right) d\mathcal{H}^{N-1}(σ)$
• Step 3: Domain decomposition estimates (equation 3.2): $\mathscr{F}_{t,p}(u; \mathbb{R}^N) = \mathscr{F}_{t,p}(u; B_δ) + \mathscr{F}_{t,p}(u; A_δ) + \mathscr{F}_{t,p}(u; B^c_{1/δ})$ where $A_δ = \{x: δ < |x| < 1/δ\}$
• Step 4: Handle each domain separately:
  • Small ball $B_δ$: utilize Lemma 2.2(i) inequality $\|u(\cdot+z)-u\|_{L^p} \leq \|\frac{z}{|z|} \cdot Du\|_{L^p}$
  • Annular region $A_δ$: utilize continuity $z \mapsto \frac{\|u(\cdot+z)-u\|^p_{L^p}}{|z|^p} \in C(\mathbb{R}^N \setminus \{0\})$
  • Exterior $B^c_{1/δ}$: controlled by condition (1.4) to be $o(1)$
• Step 5: Condition (A)'s second part $ν_k \stackrel{*}{\rightharpoonup} αδ_0$ ensures $ν = αδ_0$, thus $\mathscr{G}^{μ,αδ_0}_p = \mathscr{D}^μ_p$

Necessity Direction (B)⇒(A):

• Step 1: Derive uniform estimates from (B)'s pointwise convergence (Lemma 2.11): $\limsup_{t \to 0^+} \mathscr{F}_{t,p}(u) \leq C\|Du\|^p_{L^p}, \quad \forall u \in \text{Lip}_c(\mathbb{R}^N)$
• Step 2: Verify condition (1.4) through carefully constructed test functions:
  • For exterior integral: select $u_R(x) = \mathbb{1}_{B_{R/4}} * \text{(smooth cutoff)}$
  • For interior integral: select radial functions $v(x) = e^{|x|^2} * \text{(smooth cutoff)}$
• Step 3: Apply Theorem 3.1 to obtain convergence to $\mathscr{G}^{λ,ν}_p$ (for some $λ, ν$)
• Step 4: Scale transformation argument (Lemma 2.13): if $\mathscr{G}^{μ,ν}_p = \mathscr{G}^{μ,0}_p$ for all $u \in \text{Lip}_c(\mathbb{R}^N)$, then necessarily $ν = αδ_0$
  • Key: define $u_ε(x) = ε^{1-N/p} u(x/ε)$, then $\|σ \cdot Du_ε\|_{L^p} = \|σ \cdot Du\|_{L^p}$$\|u_ε(\cdot+z) - u_ε\|_{L^p} = ε^p \|u(\cdot + z/ε) - u\|_{L^p}$
  • Apply dominated convergence theorem to obtain $\lim_{ε \to 0^+} ε^p \int \frac{\|u(\cdot+z/ε)-u\|^p_{L^p}}{|z|^p} dν(z) = 0$

2. Compactness Criterion Proof (Theorem 1.3)

Core Idea: Construct Sobolev function approximations

• Step 1 (Proposition 4.4): For special kernel families (4.4), prove existence of $v_t \in S^p(\mathbb{R}^N)$ such that $\|v_t - u\|_{L^p} \leq C \mathscr{F}^K_{t,p}(u) β(t)^{-p}, \quad \|∇v_t\|_{L^p} \leq C \mathscr{F}^K_{t,p}(u)$
• Step 2: Construction method—convolution smoothing:
  • Select $φ \in \text{Lip}_c(\mathbb{R}^N)$ satisfying $φ \leq G * G$ and $|∇φ| \leq G * G$, where $G = \min\{K, 1\}$
  • Define $v_t = u * φ_t$, where $φ_t(x) = β(t)^N φ(β(t)x) / \|φ\|_{L^1}$
• Step 3: Utilize Lemma 4.5 (generalized Young inequality): $\int \|u(\cdot+z)-u\|^p_{L^p} (G*G)(z) dz \leq 2^p \|G\|_{L^1} \int \|u(\cdot+z)-u\|^p_{L^p} G(z) dz$
• Step 4: For bounded energy sequences $(u_k)$, construct $(v_k)$ bounded in $S^p(\mathbb{R}^N)$, thus locally compact

3. Non-local Convergence Treatment (Theorems 1.4-1.5)

Theorem 1.4 (Pointwise/Γ-convergence):

• Assumptions: $\frac{ρ_t(z)}{|z|^p} \leq Cκ(z), \quad \lim_{t \to 0^+} \frac{ρ_t(z)}{|z|^p} = κ(z) \text{ a.e.}$
• Proof: Direct application of dominated convergence theorem and Fatou's lemma

Theorem 1.5 (Compactness):

• Assumptions: For each $ε > 0$, there exists $δ > 0$ such that $\frac{ρ_t(z)}{|z|^p} \geq \frac{1}{εδ^N}, \quad \forall z \in B_δ, t \in (0,δ)$
• Proof: Utilize Fréchet-Kolmogorov compactness theorem, control smoothing error via Lemma 5.1

Technical Innovations

1. Unified Framework: Extend Fourier method from $p=2$ to general $p \in [1,\infty)$, avoiding dependence on Plancherel theorem
2. Measure Decomposition Technique (Lemma 2.9):
   • Cleverly decompose $ρ_t \mathcal{L}^N$ into spherical measure $μ$ and volume measure $ν$
   • Use countable set $I_0$ to avoid measure jumps at boundaries
3. Scale Transformation Argument (Lemma 2.13): Via scale invariance of $u_ε(x) = ε^{1-N/p} u(x/ε)$, prove $ν$ must concentrate at origin
4. Maximal Rank Condition (Definition 2.8):
   • Analogue of non-local ellipticity
   • Ensures $\text{span}(\text{supp} μ) = \mathbb{R}^N$, thus $\inf_{S^{N-1}} Θ_μ > 0$ (Lemma 2.1)
5. Domain Decomposition: Employ different estimation techniques at different scales, carefully control errors

Experimental Setup

This is a pure theoretical work containing no numerical experiments. All results are obtained through rigorous mathematical proofs.

Verification Methods

1. Verification on Known Examples:
   • Standard fractional kernels: $ρ_t(z) = \frac{1-t}{|z|^{N-(1-t)p}}$ (Remark 2.7)
   • Heat kernels and fractional heat kernels (Sections 6.1-6.2)
2. Counterexample Construction (Remark 3.2):
   • Construct kernel family $(ρ_t)$ satisfying (A) but requiring subsequence extraction for (B)
   • Definition: $ρ_t = ρ^{(1,1)}_t$ ($t = 1/k$, $k$ odd) or $ρ^{(N-1,2)}_t$ ($k$ even)
   • Result: Odd subsequence converges to $\|e_1 \cdot Du\|^p_{L^p}$, even subsequence to $\|e_N \cdot Du\|^p_{L^p}$

Experimental Results

Main Theoretical Results

1. Heat Kernel Asymptotic Formula (Theorem 1.6)

For classical heat kernel $h_t(x) = e^{-|x|^2/(4t)} / (4πt)^{N/2}$: $\lim_{t \to 0^+} t^{-p/2} \int_{\mathbb{R}^N} H_t(|u-u(x)|^p)(x) dx = \frac{2Γ(p)}{Γ(p/2)} \|Du\|^p_{L^p}$ where the constant is: