Modulus of continuity for solutions of non-local heat equations
Andrews, Chen
We extend the method of modulus of continuity for solutions of parabolic equations--as used, for instance, to prove the Fundamental Gap Conjecture--to solutions of non-local heat equations on R^n and in dimension one with a non-local Neumann boundary condition. Specifically, we show that if a solution of a non-local heat equation has an initial modulus of continuity satisfying simple criteria, then this modulus of continuity is preserved at all subsequent times. In the process of trying to generalise our result in one dimension, we found a counterexample suggesting that a non-local analogue of the Payne-Weinberger inequality would depend on more than the diameter of a bounded (convex) domain.
academic
Modulus of continuity for solutions of non-local heat equations
This paper generalizes the modulus of continuity method for parabolic equations (such as the method used to prove the fundamental gap conjecture) to non-local heat equations. Specifically, the authors prove that if a solution to a non-local heat equation has an initial modulus of continuity satisfying simple criteria, then this modulus is preserved at all subsequent times. The research covers three cases: non-local heat equations on ℝⁿ, domain heat equations on one-dimensional bounded intervals, and non-local heat equations with nonlinear terms. In extending the one-dimensional results, the authors discover a counterexample showing that an analogue of the non-local Payne-Weinberger inequality will depend on multiple geometric properties of bounded convex domains, rather than merely on the diameter.
Success of Classical Theory: The modulus of continuity method has achieved remarkable success in classical parabolic equation theory, particularly playing a key role in proving important results such as the Fundamental Gap Conjecture.
Rise of Non-local Operators: Non-local operators such as the fractional Laplacian have widespread applications in physics, probability theory, and geometric analysis, making the corresponding non-local heat equations important objects of study.
Theoretical Gap: Although the modulus of continuity theory for classical heat equations is mature, the corresponding theory for non-local heat equations has not been systematically established.
Theoretical Completeness: Generalizing classical methods to the non-local setting enriches the theoretical framework of partial differential equations.
Potential Applications: Provides technical tools for proving important results such as non-local versions of the Payne-Weinberger inequality.
Geometric Insights: Reveals through counterexamples the essential differences between non-local and classical problems regarding geometric dependence.
Classical modulus of continuity methods rely on local properties (such as gradient estimates) and cannot be directly applied to non-local operators, requiring the development of new techniques to handle long-range interactions.
Modulus of Continuity Theorem on ℝⁿ (Theorem 2.3): Proves that on the full space ℝⁿ, the modulus of continuity of solutions to non-local heat equations is preserved during evolution, provided the initial modulus of continuity satisfies specific criteria.
Results for One-Dimensional Domain Heat Equations (Theorem 3.3): Establishes similar modulus of continuity preservation properties for domain heat equations on compact intervals I = -D/2, D/2.
Construction of High-Dimensional Counterexample (Section 3.2.1): Constructs an ingenious counterexample proving that the second eigenvalue of the domain fractional Laplacian cannot depend solely on the domain's diameter, contrasting with the classical Payne-Weinberger inequality.
Nonlinear Extension (Theorem 4.2): Extends the method to non-local heat equations with gradient-dependent nonlinear terms, demonstrating the flexibility of the technique.
Coupling-by-Reflection Technique: Develops the "coupling-by-reflection" technique to cleverly handle non-local integral terms, which is the core technical innovation of the proof.
Consider the non-local linear operator:
LΩu(x):=∫Ωρ(z−x)(u(z)−u(x))dz,x∈Ω
where:
Ω⊂Rn is an open set
ρ is a non-negative, non-increasing, rotationally symmetric kernel function
u belongs to an appropriate function space
Objective: Prove that if u(x,t) satisfies the non-local heat equation
ut(x,t)=LΩu(x,t)
and the initial data u0 has modulus of continuity ϕ(⋅,0), then ϕ(⋅,t) is a modulus of continuity for u(⋅,t) at all times t≥0, i.e.,
∣u(y,t)−u(x,t)∣≤2ϕ(2∣y−x∣,t)
This is the core technical innovation of the paper. Choose coordinates so that e1=∣y−x∣y−x, and define the reflection map:
Rr=r−2⟨r,e1⟩e1
Key Observation: Due to the rotational symmetry of the kernel ρ, we have ρ(r)=ρ(Rr).
Decompose the integration domain into three disjoint sets:
S={r:⟨r,e1⟩=2∣x−y∣} (middle hyperplane)
L={r:⟨r,e1⟩<2∣x−y∣} (left half-space)
R={r:⟨r,e1⟩>2∣x−y∣} (right half-space)
Through variable substitution and reflection symmetry, the two integrals can be merged onto the same domain, thereby effectively applying the maximum principle.
By integrating out the (n−1) dimensions, reduce the n-dimensional problem to a one-dimensional problem:
ρ~(w)=∫Rn−1ρ(w,p)dp
Require the odd extension ϕ~ of the modulus of continuity to satisfy the one-dimensional non-local heat equation:
ϕ~t(r,t)=∫Rρ~(w)(ϕ~(r+w,t)−ϕ~(r,t))dw
Clever Choice of Regularization Function: For the fractional Laplacian (−Δ)s, choose Ψ(x)=(1+∣x∣2)α/2, where 0<α<min(2s,2), ensuring:
K:=supx∈RnLRnΨ(x)<∞
Detailed Analysis of Boundary Cases: In the one-dimensional case (Theorem 3.3), systematically handle various geometric configurations near interval endpoints through translation and symmetrization techniques.
Control of Nonlinear Terms: In Theorem 4.2, successfully control the contribution of the nonlinear term q(∣Du∣) through Lipschitz conditions and gradient growth bounds:
∣DΨ(x)∣≤K∇Ψ(x)
For domain heat equations on I=[−D/2,D/2], the proof structure is similar but requires careful handling of boundary cases.
Key Technical Challenges:
When ∣x−y∣≥2δ (non-overlapping): directly apply reflection technique
When ∣x−y∣<2δ (overlapping): decompose the integration domain into L1=[y−x−δ,2y−x] and L2=[−δ,2y−x], utilizing χL2−χL1≥0 to apply the maximum principle
Boundary Treatment:
Symmetric configuration (x=−y): adjust integration interval to [−δ,2D−y]
Asymmetric configuration: translate so the midpoint is at the origin, separating symmetric and asymmetric parts
Consider the rectangular domain Ω=[0,L]×[−ϵ,ϵ], where 0<L≪ϵ, and take the trial function u(x,y)=y.
Calculation:
Zero mean condition: ∫Ωu=0
Energy estimate:
Ereg[u,u]=∫Ω∫Ω∣p−q∣2+2s(y1−y2)2dpdq≤4L2ϵ2
L2 norm: ∥u∥L22=32Lϵ3
Conclusion:
λ2≤∥u∥L22Ereg[u,u]≤ϵCL→0+,L→0+
This shows that as the rectangular domain degenerates to the line segment [−ϵ,ϵ], the second eigenvalue approaches 0, contradicting a lower bound depending solely on diameter.
Geometric Insight: This counterexample reveals the sensitivity of non-local operators to domain shape (rather than merely diameter), which is an essential distinction from the classical Payne-Weinberger inequality.
By choosing coordinates so that e1=∣y−x∣y−x, the reflection transformation has a concise form and preserves the kernel function:
ρ(r)=ρ(∣r∣)=ρ(∣Rr∣)=ρ(Rr)
This is the key symmetry that makes the entire proof possible.
The classical result states that the second eigenvalue of the Neumann Laplacian on convex domains satisfies:
λ2≥D2π2
depending only on the diameter D. This paper's counterexample shows that the non-local case does not possess this property.
Complete Theory on Full Space: On ℝⁿ, the modulus of continuity preservation for non-local heat equations is completely proved, providing a foundational tool for non-local parabolic equation theory.
Success of One-Dimensional Theory: On compact intervals, despite the presence of boundaries, the modulus of continuity property holds, with proof techniques requiring careful analysis of various geometric configurations.
Barriers to High-Dimensional Generalization: Through an ingenious counterexample, the paper reveals the essential difficulty of non-local problems in higher dimensions—geometric dependence transcends the single invariant of diameter.
Flexibility of the Method: The nonlinear extension demonstrates the adaptability of the technical framework.
Unresolved High-Dimensional Domain Problem: Conjecture 3.20 (modulus of continuity theorem on high-dimensional bounded convex domains) remains open, with the counterexample suggesting direct generalization is impossible.
Restrictions on Kernel Functions: The method relies on rotational symmetry and monotonicity of the kernel; more general kernels may require new techniques.
Existence of Regularization Function: Condition (d) requires the existence of Ψ satisfying specific growth conditions, which may not hold for general operators.
Optimality Questions: Are the conditions on the modulus of continuity optimal? Do weaker sufficient conditions exist?
New Methods for High-Dimensional Domain Problems: Develop theoretical frameworks not relying on the single invariant of diameter, potentially involving more complex geometric quantities such as domain shape and inradius.
Handling Non-symmetric Kernels: Explore the possibility of removing the rotational symmetry assumption.
Applications to Other Non-local Problems:
Non-local mean curvature flow
Fractional Yamabe problem
Non-local phase transition models
Numerical Verification: Conduct numerical simulations of counterexamples and boundary cases for additional insights.
Optimal Transport Connections: Explore connections with coupling methods in optimal transport theory.
AC10 B. Andrews and J. Clutterbuck, "Proof of the fundamental gap conjecture", 2010
AC13b B. Andrews and J. Clutterbuck, "Sharp modulus of continuity for parabolic equations on manifolds and lower bounds for the first eigenvalue", Anal. PDE, 2013
Overall Assessment: This is an excellent paper with strong technical content and significant theoretical contributions. The authors successfully generalize the classical modulus of continuity method to non-local heat equations, develop innovative technical tools (particularly the coupling-by-reflection technique), and reveal essential features of non-local problems through ingenious counterexamples. Although the high-dimensional bounded domain case remains incompletely resolved, the paper establishes a solid foundation for research in this direction and clearly indicates future research paths. This work has important value for both non-local partial differential equation theory and geometric analysis.