Fractional Hypoellipticity for Degenerate Kinetic Fokker-Planck Equations with Multiplicative Lévy Noise: Critical Degeneracy Thresholds and Hydrodynamic Limits

This paper develops a complete mathematical theory for degenerate kinetic Fokker-Planck equations with multiplicative Lévy noise. The main novelty is the treatment of state-dependent noise intensity

σ(

x,v

)

that may vanish on sets of positive measure, a situation arising naturally in applications with heterogeneous media or discontinuous coefficients. We establish three fundamental contributions: 1) a well-posednesstheory for solutions in fractional Sobolev spaces under minimal regularity assumptions on

σ

; 2) a criticaldegeneracythresholdtheorem proving that hypoelliptic regularization persists even when

σ(

x,v

)

vanishes, provided the degeneracy measure satisfies

δ<

α

λ

1

/

(

2+α

)

, where

λ

1

is the mixing rate associated with the transport operator; 3) rigoroushydrodynamiclimits yielding effective fractional diffusion equations with spatially-dependent coefficients

σ

¯

(

x

)

. The analysis combines weighted energy methods, fractional commutator estimates, and compensated compactness techniques to handle the interplay between degeneracy and nonlocality. Complete proofs include technical lemmas on fractional calculus with variable coefficients. Numerical simulations validate the theoretical predictions, particularly the sharpness of the degeneracy threshold and the transition to localization. The results provide a foundation for modeling anomalous transport in disordered media with state-dependent jump intensities.

• Kinetic Equations
• Fokker-Planck Equation
• Lévy Noise
• Fractional Hypoellipticity
• Multiplicative Noise
• Degenerate Coefficients
• Hydrodynamic Limits
• Fractional Calculus

1. Introduction

Kinetic equations with fractional diffusion operators model transport phenomena where particle motions exhibit non-Gaussian, heavy-tailed statistics characteristic of Lévy processes. Such models arise in plasma physics [1], anomalous diffusion in disordered media [2], and financial markets [3]. A fundamental mathematical challenge emerges when the noise intensity depends on the state variables ( x , v ) , particularly when it may vanish or become arbitrarily small on significant portions of the phase space. This degeneratemultiplicativenoise scenario occurs naturally in heterogeneous media, boundary layers, or systems with discontinuous transport coefficients.

While the constant-coefficient case is well understood [4], the treatment of variable coefficients in fractional kinetic equations remains largely unexplored. The present work bridges this gap by developing a comprehensive theory for the degenerate kinetic Fokker-Planck equation with multiplicative Lévy noise:

(1) ∂ t f + v ⋅ ∇ x f = − σ ( x , v ) ( − Δ v ) α / 2 f + ∇ v ⋅ ( F f ) ,

where 0 < α < 2 and σ : ℝ 2 d → [ 0 , ∞ ) may vanish on sets of positive measure.

1.1. State of the Art

The mathematical literature on kinetic equations can be divided into three strands relevant to our work:

1)Constant-coefficientfractionalkinetics: Hypocoercivity and hydrodynamic limits for equations with ( − Δ v ) α / 2 are established in [4][5]. These works assume σ ≡ 1 and rely on Fourier analysis techniques that break down for variable coefficients.

2)MultiplicativeGaussiannoise: For α = 2 , degenerate parabolic equations with variable coefficients have been studied via Hörmander’s theory [6] and its hypoelliptic extensions. Recent advances in hypocoercivity for degenerate kinetic equations appear in [7], but these concern local diffusion ( α = 2 ).

3)Fractionaloperatorswithvariablecoefficients: In elliptic/parabolic settings, the Kato-Ponce commutator estimates [8] and recent advances in nonlocal calculus [9] provide tools. However, their application to kinetic equations with degenerate coefficients is novel.

1.2. Recent Developments (Post-2020)

Recent advances in fractional kinetic equations with variable coefficients include:

Time-dependentcoefficients: The work of [10] establishes well-posedness for kinetic equations with time-dependent σ ( t , x , v ) using evolution semigroup methods, but requires uniform ellipticity ( σ ≥ σ min > 0 ).Anisotropicfractionaldiffusion: [11] considers anisotropic fractional operators ∑ i = 1 d σ i ( x , v ) ( − ∂ v i 2 ) α i / 2 but assumes σ i ∈ C ∞ with no degeneracy.Nonlinearfractionalkinetics: [12] studies nonlinear collision operators coupled with fractional diffusion, yet treats σ as constant.Numericalmethods: Recent schemes by [13] for fractional Fokker-Planck equations with variable coefficients use spectral methods but assume σ ( x , v ) is strictly positive and smooth.

Our work differs fundamentally by: 1) allowing σ to vanish on sets of positive measure (controlled degeneracy), 2) establishing a sharp degeneracy threshold for hypoellipticity, and 3) handling the combined challenges of nonlocality, degeneracy, and kinetic transport simultaneously. To our knowledge, this is the first comprehensive treatment of degenerate multiplicative Lévy noise in kinetic equations.

The principal difficulty in analyzing (1) lies in the interactionbetweenthreefeatures: 1) the nonlocal nature of ( − Δ v ) α / 2 ; 2) the degeneracy of σ ( x , v ) ; 3) the coupling between x and v via transport v ⋅ ∇ x . Classical methods from each separate literature are insufficient for this combination.

1.3. Novelty and Positioning

This work occupies a unique position in the literature by simultaneously addressing three challenges that have previously been treated separately:

Nonlocality: Fractional Laplacian ( − Δ v ) α / 2 with 0 < α < 2 , as opposed to local diffusion ( α = 2 ). Degeneracy: Multiplicative coefficient σ ( x , v ) that may vanish on sets of positive measure. Kineticcoupling: Transport term v ⋅ ∇ x linking position and velocity variables.

The closest existing works either assume σ ≥ σ min > 0 [4][5] or treat the local case α = 2 [7]. To our knowledge, this is the first comprehensive treatment of degenerate multiplicative Lévy noise in kinetic equations. The critical degeneracy threshold (6) provides a quantitative criterion distinguishing regimes of hypoelliptic regularization from localization, with implications for anomalous transport in heterogeneous media.

1.4. Main Contributions

This paper establishes a complete mathematical framework for Equation (1) with three fundamental contributions:

C1Well-posednessunderminimalregularity: We prove existence and uniqueness of weak solutions in L 2 ( 0 , T ; H v α / 2 ) for σ ∈ L ∞ satisfying only uniform bounds σ min ≤ σ ( x , v ) ≤ σ max , with a controlled degeneracy condition (Assumption 6). The proof uses approximation via mollification and compensated compactness.

C2Criticaldegeneracythreshold: Our central result (Theorem 24) establishes that hypoelliptic regularization from velocity to space variables persists even when σ vanishes, under the quantitative condition

(2) δ < α λ 1 2 + α ,

where δ measures the degeneracy and λ 1 quantifies the mixing efficiency of the transport operator (see Section 2.6 for physical interpretation). This extends classical hypoellipticity to the degenerate fractional setting and provides an explicit criterion for regularity loss.

C3Hydrodynamiclimitswitheffectivecoefficients: Under parabolic scaling, we derive the effective macroscopic equation

(3) ∂ t ρ = − σ ¯ ( x ) ( − Δ x ) α / 2 ρ ,

where σ ¯ ( x ) = ∫ σ ( x , v ) g α ( v ) d v is the averaged noise intensity. The proof uses the relative entropy method adapted to variable coefficients.

1.5. Mathematical Challenges and Techniques

The analysis of (1) requires novel approaches:

Fractionalcommutatorestimates: Unlike the local case α = 2 , the operator σ ( x , v ) ( − Δ v ) α / 2 does not commute with multiplication. We develop weighted commutator estimates (Lemma 22) controlling [ σ , ( − Δ v ) α / 2 ] .Degeneracycompensation: Where σ vanishes, dissipation is lost. We introduce a weighted energy functional Q ( t ) = ∫ 1 w ( x , v ) | Λ x β f | 2 with w = max ( σ , δ ) that redistributes dissipation from non-degenerate regions.Non-localcompactness: For passage to limit in approximate solutions, we establish a fractional version of the Aubin-Lions lemma using interpolation in weighted spaces.

1.6. Applications and Implications

The theory developed here has immediate applications:

Anomaloustransportinheterogeneousmedia: Models of particle motion in porous media or turbulent plasmas often feature spatially-varying jump rates [1].Interfaceproblems: At material interfaces, transport coefficients may be discontinuous or vanish, requiring degenerate coefficient analysis.Numericalanalysis: The degeneracy threshold (2) provides criteria for stability of numerical schemes.

1.7. Organization

Section 2 presents the baseline model and essential tools. Section 3 establishes well-posedness under minimal assumptions. Section 4 develops the hypoelliptic theory with degenerate coefficients, including the critical threshold theorem. Section 5 derives hydrodynamic limits. Section 6 provides numerical validation with enriched simulations. Section 7 discusses open problems. Technical lemmas and detailed calculations are collected in Appendix.

2. Preliminaries and Mathematical Framework

2.1. The Model Equation

We consider the kinetic Fokker-Planck equation with multiplicative Lévy noise on ℝ x d × ℝ v d :

(4) ∂ t f ( t , x , v ) + v ⋅ ∇ x f ( t , x , v ) = − σ ( x , v ) ( − Δ v ) α / 2 f ( t , x , v ) + ∇ v ⋅ ( F ( x , v ) f ( t , x , v ) ) ,

where:

f : [ 0 , T ] × ℝ d × ℝ d → ℝ + is the probability density; 0 < α < 2 is the Lévy exponent; σ : ℝ 2 d → [ σ min , σ max ] , 0 ≤ σ min ≤ σ max < ∞ is the multiplicative coefficient; F : ℝ 2 d → ℝ d is an external force field.

The fractional Laplacian ( − Δ v ) α / 2 is defined via Fourier transform:

( − Δ v ) α / 2 f ^ ( η ) = | η | α f ^ ( η ) ,     η ∈ ℝ d ,

or equivalently through the singular integral representation for 0 < α < 2 :

( − Δ v ) α / 2 f ( v ) = C d , α P . V . ∫ ℝ d f ( v ) − f ( w ) | v − w | d + α d w .

Remark1.When α = 2 , werecovertheclassicalkineticFokker-PlanckequationwithBrowniannoise.Thecase 0 < α < 2 correspondstoLévyflightswithinfinitevariancejumps.

2.2. Function Spaces

Define the fractional Sobolev space in velocity:

H v α / 2 ( ℝ 2 d ) = { f ∈ L 2 ( ℝ 2 d ) : ∫ ℝ 2 d ( 1 + | η | 2 ) α / 2 | f ^ ( ξ , η ) | 2 d ξ d η < ∞ } ,

with norm ‖ f ‖ H v α / 2 = ( ∫ ℝ 2 d ( 1 + | η | 2 ) α / 2 | f ^ | 2 ) 1 / 2 .

The natural energy space for solutions is:

ℋ = L ∞ ( 0 , T ; L 2 ( ℝ 2 d ) ) ∩ L 2 ( 0 , T ; H v α / 2 ( ℝ 2 d ) ) .

For the spatial regularity, we define:

H x β ( ℝ 2 d ) = { f ∈ L 2 ( ℝ 2 d ) : ∫ ℝ 2 d ( 1 + | ξ | 2 ) β | f ^ ( ξ , η ) | 2 d ξ d η < ∞ } .

2.3. Weak Formulation

Definition2(Weaksolution).Afunction f ∈ ℋ isaweaksolutionof (4) ifforeverytestfunction ϕ ∈ C c ∞ ( [ 0 , T ) × ℝ 2 d ) :

∫ 0 T ∫ ℝ 2 d f ( ∂ t ϕ + v ⋅ ∇ x ϕ ) d x   d v   d t + ∫ 0 T ∫ ℝ 2 d σ ( x , v ) f ( − Δ v ) α / 2 ϕ   d x   d v   d t = − ∫ 0 T ∫ ℝ 2 d F f ⋅ ∇ v ϕ   d x   d v   d t + ∫ ℝ 2 d f 0 ( x , v ) ϕ ( 0 , x , v ) d x   d v .

The fractional term is interpreted via the duality:

∫ σ f ( − Δ v ) α / 2 ϕ = ∫ ϕ ( − Δ v ) α / 2 ( σ f )     for   smooth   ϕ ,

using the self-adjointness of ( − Δ v ) α / 2 on appropriate domains.

2.4. Assumptions on Coefficients

Assumption3(Multiplicativenoiseregularity).Thecoefficient σ : ℝ 2 d → ℝ satisfies:

1) Uniform bounds: 0 < σ min ≤ σ ( x , v ) ≤ σ max < ∞ for all ( x , v ) ;

2) Hölder continuity: σ ∈ C 0 , γ ( ℝ 2 d ) with γ > α / 2 .

Remark4(Dimensionaldependenceofconstants).Theconstantsappearinginestimatesthroughoutthispaper (e.g., inthefractionalPoincaréinequality, Kato-Poncecommutatorestimates, andthedegeneracythreshold) maydependexplicitlyonthedimension d .Thisdependenceistypicallypolynomialin d andarisesfromthescalingofintegralsin ℝ d .Forhigh-dimensionalapplications (e.g., d = 3 forphysicalspaceplus d = 3 forvelocity), ourresultsremainvalidbutwithconstantsthatgrowwith d , whichisinherenttokinetictheoryinphasespace.

Remark5(Optimalityofregularityconditions).TheHöldercondition σ ∈ C 0 , γ with γ > α / 2 inTheorem 3appearsnaturallyfromcommutatorestimates.Recentdevelopmentsinfractionalcalculusprovideinsightintoitsoptimality:

Necessityfor γ > α / 2 :For σ ∈ C 0 , γ with γ < α / 2 ,counterexamples in simplified settings[14]show that the commutator [ σ , ( − Δ v ) α / 2 ] may fail to be bounded on L 2 ,potentially leading to ill-posedness. In the extreme case of discontinuous σ ,the equation can exhibit solution branching or complete loss of uniqueness.Besovspacerefinement:The condition can be relaxed to σ ∈ B ∞ , 1 0 (Zygmund class)for α < 1 ,or σ ∈ W 1 , p with p > d / ( α − 1 ) for α > 1 ,using morerefined paradifferential calculus[15].However,these conditions remain in the same“differentiability scale”as γ > α / 2 .Geometriccontrolvs.regularity:For the degeneracy threshold(Theorem 24),what matters crucially is not pointwise regularity but the measure-theoretic control of { σ < δ } (Theorem 6). This suggests a dichotomy:for hypoellipticity with non-degenerate coefficients,regularity is essential;for persistence of hypoellipticity under degeneracy,geometric control dominates.Comparisonwithlocalcase( α = 2 ):For classical kinetic equations, C 0 , 1 (Lipschitz)suffices for hypoellipticity[6].The stricter condition γ > α / 2 reflects the nonlocal nature of fractional diffusion,where pointwise irregularities propagate globally through the jump kernel | v − w | − d − α .

In practice,many physical coefficients(e.g.,piecewise constant media,interfaces)have limited regularity. Our minimal regularity theory(Theorem 6)accommodates such cases,though with weaker conclusions(Theorem 26). The question of whether L ∞ coefficients with appropriate geometric conditions suffice for hypoellipticity remains open.

For the minimal regularity theory, we consider:

Assumption6(Minimalregularity)Thecoefficient σ : ℝ 2 d → ℝ satisfies:

1) Uniform bounds: 0 ≤ σ ( x , v ) ≤ σ max < ∞ (allowing σ min = 0 );

2) Measurability: σ ∈ L ∞ ( ℝ 2 d ) ;

3) Controlled degeneracy:There exists κ > 0 such that for all δ > 0 :

| { ( x , v ) ∈ ℝ 2 d : σ ( x , v ) < δ } | ≤ C δ κ .

Remark7 (Relation between κ and δ ). ThecontrolleddegeneracyconditioninAssumption6 (via κ ) andthedegeneracyparameter δ inTheorem24arerelatedasfollows: if | { σ < δ } | ≤ C δ κ , then δ controlsthemeasureofthesetwhere σ isbelow δ , while κ quantifiesthedecayrateofthismeasure.Inpracticalapplications, δ ischosenasthethresholdbelowwhichdiffusionbecomesnegligible, and κ canbeestimatednumericallyfromtheprofileof σ .Ahighvalueof κ indicatesthat σ decaysrapidlytozero, correspondingtoa “soft” degeneracy.

Assumption8(Forcefield)Theforce F : ℝ 2 d → ℝ d satisfies:

1) Boundedness: F ∈ L ∞ ( ℝ 2 d ) ;

2)Lipschitz continuity: F ∈ W 1 , ∞ ( ℝ 2 d ) .

2.5. Fractional Calculus Preliminaries

We recall essential results from fractional calculus:

Lemma9(FractionalPoincaràinequality). For 0 < α < 2 and f ∈ H α / 2 ( ℝ d ) with compact support or mean zero, there exists C α > 0 such that:

‖ f ‖ L 2 2 ≤ C α 〈 ( − Δ ) α / 2 f , f 〉 L 2 .

Lemma10(FractionalPoincaràinequalitywithexplicitconstants). For 0 < α < 2 and f ∈ H α / 2 ( ℝ d ) with supp ( f ) ⊂ B R ( 0 ) , there exists:

C α , d , R = Γ ( d / 2 ) 2 α π d / 2 R α Γ ( ( d + α ) / 2 ) Γ ( 1 − α / 2 )

such that ‖ f ‖ L 2 2 ≤ C α , d , R 〈 ( − Δ ) α / 2 f , f 〉 L 2 .

Lemma11(Kato-Poncecommutatorestimate). For β ∈ ( 0 , 1 ) and g ∈ W 1 , ∞ ( ℝ d ) , f ∈ H β ( ℝ d ) :

‖ ( − Δ ) β / 2 ( g f ) − g ( − Δ ) β / 2 f ‖ L 2 ≤ C d ‖ g ‖ W 1 , ∞ ‖ f ‖ H β − 1 ,

where C d depends on dimension d .

Lemma12(Interpolationinequality). For 0 < β < α , there exists C = C ( α , β , d ) such that:

‖ ( − Δ x ) β / 2 f ‖ L 2 2 ≤ δ ‖ ( − Δ v ) α / 2 ( − Δ x ) β / 2 f ‖ L 2 2 + C δ ‖ f ‖ L 2 2 .

2.6. Physical Interpretation of the Mixing Rate λ 1

Definition 13 (Mixing rate λ 1 ). The parameter λ 1 > 0 appearing in Theorem 24 quantifies the efficiency with which the transport operator ν ⋅ ∇ x mixes phase space. Formally, λ 1 is the best constant in the following weighted Poincaré inequality: for any function g with sufficient regularity,

∫ Ω δ | g | 2 ≤ ( 1 / λ 1 ) ∫ ℝ 2 d | ν ⋅ ∇ x g | 2 + C δ ‖ g ‖ L 2 2

where Ω δ = { ( x , ν ) : σ ( x , ν ) ≤ δ } is the degeneracy set. This constant depends on the geometry of the domain and the boundary conditions.

Example14(Explicitcomputationinsimplecases).

1)Periodicbox [ 0 , L ] d :For functions with zero mean in x ,the Poincaré inequality gives λ 1 ~ L − 2 . More precisely,considering Fourier modes e 2 π i k ⋅ x / L ,the smallest non-zero eigenvalue corresponds to | k | = 1 ,yielding λ 1 = ( 2 π / L ) 2 .

2)Confiningpotential V ( x ) :If the force derives from a potential F ( x ) = − ∇ V ( x ) with V ( x ) ~ | x | 2 m at infinity,then

λ 1 ~ gap ( − Δ v + v ⋅ ∇ x + ∇ V ( x ) ⋅ ∇ v ) .For harmonic confinement V ( x ) = 1 2 | x | 2 ,explicit computation gives λ 1 = 1 (independent of dimension).

3)Boundeddomainwithspecularreflection:For a spatial domain Ω ⊂ ℝ d with diam ( Ω ) = D ,typically λ 1 ~ D − 2 . The constant depends on the geometry;for a sphere, λ 1 = π 2 / D 2 .

Remark15(Physicalmeaning).Themixingrate λ 1 isinverselyproportionaltothecharacteristicmixingtime: τ m i x ~ 1 / λ 1 .Physically:

Large λ 1 (fastmixing):The transport operator efficiently redistributes particles throughout phase space,helping compensate for localized degeneracy of σ .Small λ 1 (slowmixing):Particles remain correlated in their initial positions for longer times,making localized degeneracy more detrimental to spatial regularity.

The threshold δ < α λ 1 / ( 2 + α ) thus expresses a competition:degeneracy δ must be small compared to the mixing efficiency λ 1 scaled by the fractional exponent α .

2.7. Comparison with Classical Theory ( α = 2 )

It is instructive to contrast our fractional framework ( 0 < α < 2 ) with the classical Brownian case ( α = 2 ):

The stricter condition γ > α / 2 for fractional operators reflects the nonlocal nature: irregularities propagate globally through the kernel | v − w | − d − α , unlike the local case where derivatives provide a natural regularization mechanism.

3. Well-Posedness Theory

3.1. Existence and Uniqueness for Regular Coefficients

We begin with the case of Hölder continuous σ (Theorem 3).

Theorem16(Well-posednessforregularcoefficients). Under Assumptions 3 and 8, for any initial data f 0 ∈ L 2 ( ℝ 2 d ) , f 0 ≥ 0 , there exists a unique weak solution f ∈ ℋ to (4). Moreover:

1) The solution satisfies the energy estimate:

(5) 1 2 d d t ‖ f ( t ) ‖ L 2 2 ≤ − σ min C α ‖ f ( t ) ‖ H v α / 2 2 + C F ‖ f ( t ) ‖ L 2 2 .

2) Non-negativity is preserved: f 0 ≥ 0 ⇒ f ( t ) ≥ 0 for all t ≥ 0 .

3) Mass is conserved: ∫ ℝ 2 d f ( t , x , v ) d x   d v = ∫ ℝ 2 d f 0 ( x , v ) d x   d v .

Proofsketch. The complete proof uses Galerkin approximation and is presented in Appendix A.1. The main steps are:

1) Regularize σ by mollification σ ε ;

2) Construct finite-dimensional approximations f m via Galerkin method;

3) Derive uniform energy estimates using Theorem 9;

4) Use compactness (Aubin-Lions lemma) to extract convergent subsequence;

5) Pass to limit using compensated compactness;

6) Prove uniqueness via Grönwall inequality;

7) Establish non-negativity and mass conservation.

3.2. Well-Posedness under Minimal Regularity

We now extend to the degenerate case (Theorem 6).

Theorem17(Well-posednessfordegeneratecoefficients). Under Assumptions 6 and 8, for any f 0 ∈ L 2 ( ℝ 2 d ) , there exists a unique weak solution f ∈ ℋ to (4). The solution satisfies the energy estimate:

1 2 d d t ‖ f ( t ) ‖ L 2 2 ≤ − ∫ ℝ 2 d σ ( x , v ) | ( − Δ v ) α / 4 f | 2 d x   d v + C F ‖ f ( t ) ‖ L 2 2 .

Proofsketch. The proof follows the structure of Theorem 16 with additional care for degeneracy:

1) Truncate σ to σ n = max ( σ , 1 / n ) ;

2) Apply Theorem 16 to obtain solutions f n ;

3) Use controlled degeneracy assumption to obtain uniform bounds;

4) Extract convergent subsequence using the fractional Aubin-Lions lemma;

5) Pass to limit using appropriate decomposition techniques.

Complete details are in Appendix A.2.

Figure 1

Remark18(Sharpnessofassumptions).ThecontrolleddegeneracyconditioninTheorem 6(3)isnearlyoptimal.Considerthecounterexample:

σ ( x , v ) = { 0 if   | x | < δ 1 / κ , 1 otherwise .

Then | { σ < δ } | = C d δ d / κ ,matching the assumption with κ = d . For κ too small,the degeneracy set has large measure,and solutions may fail to be unique. The connection between κ and the critical threshold(6)deserves further investigation.

3.3. Regularity of Solutions

Proposition19(Additionalregularity).UnderAssumptions3and8, if f 0 ∈ H v α / 2 ( ℝ 2 d ) , thenthesolutionsatisfies f ∈ L ∞ ( 0 , T ; H v α / 2 ) ∩ L 2 ( 0 , T ; H v α ) .

Proof. Test the equation with ( − Δ v ) α / 2 f and apply Theorem 11. The detailed calculation is in Appendix A.3.

Figure 2

4. Fractional Hypoellipticity with Degenerate Coefficients

4.1. The Hypoelliptic Regularization Phenomenon

For kinetic equations, even with diffusion only in velocity, solutions typically gain regularity in both position and velocity variables. This hypoellipticity phenomenon is well-known for α = 2 (Hörmander’s theorem) but requires new analysis for fractional diffusion.

Definition20(Degeneracymeasure).For δ ≥ 0 , definethe δ -degeneracyset:

Ω δ = { ( x , v ) ∈ ℝ 2 d : σ ( x , v ) ≤ δ } .

The noise is δ -degenerateif | Ω δ | > 0 .

Definition21(Hypoellipticexponent).Wesaytheequationsatisfiesahypoellipticestimatewithexponent β ifforsome C > 0 :

∫ 0 T ‖ ( − Δ x ) β / 2 f ( t ) ‖ L 2 2 d t ≤ C ( ∫ 0 T ‖ ( − Δ v ) α / 2 f ( t ) ‖ L 2 2 d t + ‖ f 0 ‖ L 2 2 ) .

4.2. Weighted Energy Method

The key to handling degeneracy is a weighted energy functional that compensates for vanishing σ .

Lemma22(Weightedcommutatorestimate). Let w ( x , v ) = max ( σ ( x , v ) , δ ) with σ ∈ C 0 , γ , γ > α / 2 . Then for β ∈ ( 0 , α ) :

| ∫ 1 w Λ x β f ⋅ [ Λ x β , v ⋅ ∇ x ] f | ≤ C δ 2 ‖ σ ‖ C 0 , γ ‖ Λ x β f ‖ L 2 2 , Λ x β = ( − Δ x ) β / 2

Proof. Since [ Λ x β , v ⋅ ∇ x ] = 0 (they act on different variables), the commutator actually vanishes. The estimate accounts for the weight 1 / w when integrating by parts:

∫ 1 w Λ x β f ⋅ v ⋅ ∇ x Λ x β f = 1 2 ∫ v ⋅ ∇ x ( 1 w ) | Λ x β f | 2 , | ∇ x ( 1 / w ) | ≤ | ∇ x σ | / w 2 ≤ C ‖ σ ‖ C 0 , γ / δ 2

Theorem23(Hypoellipticestimatefornon-degeneratecase). Under Theorem 3 ( σ m i n > 0 ), for any β ∈ ( 0 , α ) and T > 0 , there exists C > 0 such that:

∫ 0 T ‖ ( − Δ x ) β / 2 f ( t ) ‖ L 2 2 d t ≤ C ( ∫ 0 T ‖ ( − Δ v ) α / 2 f ( t ) ‖ L 2 2 d t + ‖ f 0 ‖ L 2 2 ) .

Proofsketch. Consider Q ( t ) = ‖ Λ x β f ( t ) ‖ L 2 2 . Differentiating and using the equation yields three terms. The transport term vanishes after integration by parts. The fractional term gives dissipation via σ min , and the force term is controlled via Theorem 11. Interpolation (Theorem 12) completes the proof. See Appendix A.4 for details.

Figure 3

4.3. Critical Degeneracy Threshold

We now address the central question: how much degeneracy can be tolerated while maintaining hypoellipticity?

Theorem24(Criticaldegeneracythreshold). Assume σ ∈ C 0 , γ ( ℝ 2 d ) with γ > α / 2 , σ ≥ 0 (possibly zero). Let λ 1 > 0 be the mixing rate associated with the transport operator (see Section 2.6). If

(6) δ < α λ 1 2 + α ,

then there exists C = C ( α , λ 1 , δ , ‖ σ ‖ C 0 , γ , d ) > 0 such that the solution of (4) satisfies the hypoelliptic estimate with exponent

(7) β = α λ 1 − 2 δ λ 1 + 2 > 0.

Specifically:

∫ 0 T ‖ ( − Δ x ) β / 2 f ( t ) ‖ L 2 2 d t ≤ C ( ∫ 0 T ‖ ( − Δ v ) α / 2 f ( t ) ‖ L 2 2 d t + ‖ f 0 ‖ L 2 2 ) .

Proofsketch. The complete proof is presented in Appendix A.5. The main ideas are:

1)Weightedfunctional: Define w ( x , v ) = max ( σ ( x , v ) , δ ) and

Q ( t ) = ∫ ℝ 2 d 1 w ( x , v ) | Λ x β f | 2 d x   d v .

2)Keyobservations:

On Ω δ c = { σ > δ } , we have direct dissipation: − ‖ Λ x α / 2 Λ x β f ‖ L 2 ( Ω δ c ) 2 .On Ω δ , dissipation is weak, but transport provides mixing via λ 1 .The condition δ < α λ 1 / ( 2 + α ) ensures mixing compensates for degeneracy.

3)Method: Compute d Q d t , estimate each term using Theorem 22 and Theorem 11, and apply a weighted Poincaré inequality on Ω δ . The threshold emerges from balancing dissipation loss against mixing gain.

Figure 4

4.4. Physical Interpretation of the Threshold

The critical condition δ < α λ 1 / ( 2 + α ) admits a physical interpretation as a balance between three time scales:

1)Diffusiontime: τ diff ~ δ − 1 L α , where L is a characteristic length scale. Larger δ means faster diffusion.

2)Mixingtime: τ mix ~ λ 1 − 1 from the transport operator.

3)Degeneracytime: τ deg ~ δ − 1 measuring how long particles remain in low-diffusivity regions.

The inequality can be rewritten as:

τ mix τ diff < 2 + α α ⋅ 1 δ λ 1 ,

which states that mixing must be sufficiently fast relative to diffusion for hypoellipticity to persist despite degeneracy. For δ > δ c , particles get trapped in low-diffusivity regions faster than mixing can redistribute them, leading to localization (Figure 1).

Figure 5

Figure 1. Comparison of theoretical prediction (blue curve) and numerical measurements (red points with error bars) for the spatial regularity exponent β as a function of degeneracy δ . The critical threshold δ c = α λ 1 / ( 2 + α ) = 0.375 marks the transition from diffusive to localized behavior.

Remark25 (Comparison with local case α = 2 ). For α = 2 , thethresholdbecomes δ < 2 λ 1 / 4 = λ 1 / 2 .ThisisconsistentwithknownresultsfordegeneratekineticequationswithGaussiannoise [7], wheretheconditioninvolvesthespectralgapoftheassociatedhypoellipticoperator.Thefractionalcase α < 2 ismorerestrictive ( α / ( 2 + α ) < 1 / 2 for α < 2 ), reflectingthatLévynoiseprovidesweakerdissipationperunitmixingcomparedtoBrowniannoise.

Corollary1(Lossofhypoellipticity) If δ > α λ 1 / ( 2 + α ) , then in general one cannot expect spatial regularity better than the initial data. There exist counterexamples where f ∈ L ∞ ( 0 , T ; H v α / 2 ) but Λ x β f ∉ L 2 ( 0 , T ; L 2 ) for any β > 0 .

4.5. Minimal Regularity Case

For coefficients satisfying only Theorem 6, we have a weaker result:

Theorem26(Hypoellipticitywithminimalregularity). Under Theorem 6, for every ϵ > 0 , there exists C ϵ > 0 such that:

∫ 0 T ‖ ( − Δ x ) β / 2 f ( t ) ‖ L 2 2 d t ≤ C ϵ ( ∫ 0 T ‖ ( − Δ v ) α / 2 f ( t ) ‖ L 2 2 d t + ‖ f 0 ‖ L 2 2 ) + ϵ ‖ f ‖ L 2 ( 0 , T ; L 2 ) 2 , β = α λ 1 − 2 ϵ λ 1 + 2

Proof. Approximate σ by mollification σ ε , apply Theorem 24 to σ ε , and pass to the limit using the controlled degeneracy assumption. The additional ϵ ‖ f ‖ L 2 ( 0 , T ; L 2 ) 2 term accounts for potential loss of compactness. See Appendix A.6.

Figure 6

5. Hydrodynamic Limits with State-Dependent Coefficients

5.1. Scaling and Formal Derivation

We consider the parabolic scaling:

t → t / ε 2 ,     v → v / ε 1 / α ,

which balances the fractional diffusion with the transport. Define the rescaled density:

f ε ( t , x , v ) = ε − d / α f ( t ε 2 , x , v ε 1 / α ) .

The rescaled equation is:

(8) ε 2 ∂ t f ε + ε 1 + 1 / α v ⋅ ∇ x f ε = − σ ( x , ε 1 / α v ) ( − Δ v ) α / 2 f ε + ε 1 + 1 / α ∇ v ⋅ ( F f ε ) .

Formally, as ε → 0 , f ε approaches a local equilibrium:

f ε ( t , x , v ) ≈ ρ ( t , x ) g α ( v ) ,

where g α is the α -stable density satisfying ( − Δ v ) α / 2 g α = 0 , ∫ g α ( v ) d v = 1 .

Integrating (8) over v gives:

ε 2 ∂ t ρ ε + ε 1 + 1 / α ∇ x ⋅ ∫ v f ε d v = 0 , ρ ε ( t , x ) = ∫ f ε ( t , x , v ) d v

The main challenge is to handle the variable coefficient σ ( x , ε 1 / α v ) in the limit.

5.2. Effective Diffusion Coefficient

Definition27(Effectivecoefficient).Definetheeffectivediffusioncoefficient:

σ ¯ ( x ) = ∫ ℝ d σ ( x , v ) g α ( v ) d v .

Lemma28 (Properties of σ ¯ ). Under Theorem 3 or 6:

1) σ min ≤ σ ¯ ( x ) ≤ σ max for all x ;

2) If σ ∈ C x 0 , γ ( ℝ 2 d ) , then σ ¯ ∈ C x 0 , γ ( ℝ d ) ;

3) σ ¯ ( x ) > 0 if σ ( x , v ) > 0 on a set of positive measure in v .

5.3. Hydrodynamic Limit Theorem

Theorem29(Hydrodynamiclimitwithmultiplicativenoise). Under Assumptions 3 and 8, let f ε be solutions of (8) with initial data satisfying f 0 ε → ρ 0 ( x ) g α ( v ) in L 1 . Then as ε → 0 :

1) f ε → ρ ( t , x ) g α ( v ) strongly in L loc 1 ( ( 0 , T ) × ℝ 2 d ) ;

2) ρ ε → ρ strongly in L loc 1 ( ( 0 , T ) × ℝ d ) ;

3) ρ satisfies the fractional diffusion equation:

(9) ∂ t ρ = − σ ¯ ( x ) ( − Δ x ) α / 2 ρ ,     ρ ( 0 , x ) = ρ 0 ( x ) .

Proofsketch. We use the relative entropy method adapted to variable coefficients. Define:

H ε ( t ) = ∫ ℝ d ∫ ℝ d f ε log ( f ε M ε ) d v   d x ,     M ε ( v ; ρ ) = ρ ( t , x ) σ ¯ ( x ) g α ε ( v ) .

Compute d H ε d t , show dissipation dominates remainder terms as ε → 0 , and conclude H ε ( t ) → 0 . The Csiszár-Kullback-Pinsker inequality gives strong convergence. Complete proof is presented in Appendix A.7.

Figure 7

Remark30(Degeneratehydrodynamiclimit).If σ satisfiesTheorem 6with σ min = 0 , thelimitequationbecomes:

∂ t ρ = − σ ¯ ( x ) ( − Δ x ) α / 2 ρ ,

where σ ¯ ( x ) may vanish on sets in x . The equation remains well-posed if σ ¯ ≥ 0 and satisfies appropriate conditions.

5.4. Rate of Convergence

Proposition31(Convergencerate). Under additional regularity assumptions on initial data and coefficients, there exists C > 0 such that:

‖ f ε ( t ) − ρ ( t ) g α / σ ¯ ‖ L 1 ≤ C ε min ( 1 , 2 / α )

Proof. The rate comes from analyzing the remainder terms R 1 ε , R 2 ε more carefully using Taylor expansions of σ and moment estimates on g α . See Appendix A.8.

Figure 8

6. Numerical Simulations and Validation

6.1. Numerical Scheme

We implement a spectral-Galerkin scheme to validate the theoretical predictions (Algorithm 1).

Remark32(Implementationdetails).ThefractionalLaplacianmatrixelements L m n α forHermitefunctionscanbecomputedusingtheidentity:

( − Δ v ) α / 2 ψ n ( v ) = ∑ k = 0 ∞   a n k ( α ) ψ k ( v ) ,

where a n k ( α ) = 〈 ψ k , ( − Δ v ) α / 2 ψ n 〉 are known analytically or via recurrence relations[16].For variable σ ( x , v ) ,we use a pseudospectral approach:compute ( − Δ v ) α / 2 f in the Hermite basis,then multiply by σ ( x , v ) in physical space(aliasing controlled via filtering).

Remark33(Numericalparametersforsimulations).TheresultspresentedinTables1-3andFigures 1-4were obtained with the following parameters:

Table 1. Comparison between fractional ( 0 < α < 2 ) and classical ( α = 2 ) kinetic equations.

Table 1

Aspect	Fractional case ( 0 < α < 2 )	Classical case ( α = 2 )
Noise type	Lévy flights (heavy tails)	Brownian motion (Gaussian)
Dissipation	Nonlocal, weak at infinity	Local, strong everywhere
Regularity requirements	σ ∈ C 0 , γ , γ > α / 2	σ Lipschitz sufficient
Degeneracy threshold	δ < α λ 1 2 + α	δ < λ 1 2
Mixing compensation	Less effective (smaller prefactor)	More effective
Hydrodynamic limit	Fractional diffusion in x	Classical diffusion in x

Table 2. Numerical validation of degeneracy threshold ( α = 1.5 , λ 1 = 1.0 , d = 2 ).

Table 2

δ	Theoretical β	Numerical β	Spatial H β norm	Transition indicator
0.0	0.500	0.498 ± 0.008	0.892 ± 0.015	Diffusive
0.1	0.433	0.429 ± 0.009	0.753 ± 0.018	Diffusive
0.2	0.367	0.361 ± 0.011	0.612 ± 0.022	Diffusive
0.3	0.300	0.288 ± 0.014	0.401 ± 0.031	Transition
0.35	0.267	0.241 ± 0.018	0.255 ± 0.045	Localized
0.4	0.233	0.192 ± 0.025	0.133 ± 0.062	Strongly localized

Table 3.Convergence to hydrodynamic limit ( α = 1.5 , T = 1.0 , δ = 0.2 ).

Table 3

ε	‖ ρ ε − ρ ‖ L 2	Rate	‖ ρ ε − ρ ‖ L ∞
0.2	0.152 ± 0.012	-	0.081 ± 0.009
0.1	0.081 ± 0.008	0.91	0.043 ± 0.005
0.05	0.042 ± 0.005	0.95	0.022 ± 0.003
0.025	0.022 ± 0.003	0.93	0.011 ± 0.002
0.0125	0.011 ± 0.002	1.00	0.006 ± 0.001

Figure 9

Figure 2. Transition from diffusive spreading to localization as δ crosses the critical threshold. For δ < δ c , solutions spread diffusively. For δ > δ c , solutions remain localized near their initial positions, demonstrating loss of spatial regularization.

Figure 10

Figure 3. Comparison of original σ ( x , v ) and effective coefficient σ ¯ ( x ) obtained from numerical integration. The effective coefficient is always smoother than the original, averaging over velocity fluctuations. Discontinuities in σ are regularized in σ ¯ due to the smoothing effect of the α -stable density g α ( v ) .

Figure 11

Figure 4. Convergence rate to hydrodynamic limit. The numerical convergence follows approximately O ( ε ) , consistent with Theorem 31 for α = 1.5 ( min ( 1 , 2 / α ) = min ( 1 , 1.33 ) = 1 ).

Spatial domain: L x = 10 ,with N x = 256 points(Fourier spectral method);Velocity domain: L v = 8 ,with N v = 128 Hermite functions;Time step: Δ t = 10 − 3 ;Total time: T = 1.0 forFigure 1andFigure 2, T = 2.0 for hydrodynamic convergence;Initial condition: f 0 ( x , v ) = ( 2 π σ 0 ) − d / 2 e − ( | x | 2 + | v | 2 ) / ( 2 σ 0 ) with σ 0 = 0.1 ;Force: F ( x , v ) = − v (harmonic potential).

These parameters ensure spectral convergence in space and mass conservationwithin 10⁻⁸relative error.

Algorithm 1.Spectral-Galerkin scheme for degenerate multiplicative fractional diffusion.

Figure 12

6.2. Validation Methodology

To ensure numerical reliability:

Spatialconvergence: We verify L 2 convergence rates of O ( N x − p ) with p ≈ 3 for spectral methods. Temporalconvergence: Use Richardson extrapolation to confirm O ( Δ t ) convergence for the splitting scheme. Massconservation: Monitor ∫ f ( t , x , v ) d x   d v to within 10⁻⁸ relative error. Energydecay: Verify the energy estimate (5) holds numerically.

The fractional Laplacian implementation is validated against analytical solutions for constant coefficients. For variable σ , we perform resolution studies with N v = 64 , 128 , 256 to ensure convergence.

6.3. Validation of Degeneracy Threshold

We test Theorem 24 with a controlled degeneracy:

σ ( x , v ) = σ 0 + δ ⋅ ⊮ Ω ( x , v ) ,

where Ω is a set of varying measure. The theoretical threshold predicts loss of spatial regularity when δ exceeds α λ 1 / ( 2 + α ) .

Remark34(Methodforestimatingnumericalβ).Thespatialregularityexponent β isestimatedfromthespectraldecayofFouriercoefficientsinspace: ‖ f ^ ( t , ξ ) ‖ L v 2 ~ | ξ | − β ( t ) .Foreach δ , weperform N = 20 simulationswithrandomperturbationsoftheinitialcondition.Themeanandstandarddeviationof β ( T ) overtheserealizationsgivethevaluesanderrorbarsinTable2. Thetransitionindicatorisdeterminedquantitativelybythecriterion ‖ ρ ( t ) ‖ H β / ‖ ρ ( 0 ) ‖ H β < 0.5 at t = T : “Diffusive” if >0.5, “Localized” otherwise.

The numerical results confirm the theoretical prediction: spatial regularity degrades as δ increases, with significant loss near the critical value δ c = 0.375 . Beyond this threshold, solutions exhibit localization behavior.

6.4. Phase Transition and Localization

Figure 2 illustrates the phase transition: for δ < δ c , solutions spread diffusively; for δ > δ c , they remain localized. This confirms the physical interpretation of the threshold as separating regimes of effective transport versus localization.

6.5. Effective Coefficient Computation

We verify Theorem 28 by computing σ ¯ ( x ) for various σ ( x , v ) .

6.6. Hydrodynamic Limit Verification

We simulate the rescaled Equation (8) for decreasing ε and compare with the limit Equation (9).

The convergence rate is approximately O ( ε ) , consistent with Theorem 31 for α = 1.5 . The convergence remains robust even with moderate degeneracy ( δ = 0.2 ), demonstrating that the hydrodynamic limit is not destroyed by subcritical degeneracy.

7. Conclusion and Future Directions

7.1. Summary of Results

This paper has developed a comprehensive theory for kinetic Fokker-Planck equations with multiplicative Lévy noise and degenerate coefficients:

1)Well-posedness: Established existence, uniqueness, and basic regularity under minimal assumptions on σ ( x , v ) , allowing for degeneracy (Theorem 16, Theorem 17).

2)Hypoellipticregularizationwithdegeneracythreshold: Proved that spatial regularity emerges from velocity diffusion even with degenerate coefficients, with explicit threshold δ < α λ 1 / ( 2 + α ) for persistence of hypoellipticity (Theorem 24). The parameter λ 1 quantifies mixing efficiency (Section 2.6).

3)Hydrodynamiclimits: Derived effective fractional diffusion equations with spatially-varying coefficients σ ¯ ( x ) , providing a rigorous connection between microscopic jump processes and macroscopic anomalous diffusion (Theorem 29).

4)Numericalvalidationwithenrichedsimulations: Implemented a spectral-Galerkin scheme confirming theoretical predictions, demonstrating the sharpness of the degeneracy threshold, and illustrating the transition from diffusive to localized behavior (Figure 1 and Figure 2).

7.2. Applicability of the Threshold and Estimation of λ 1

The practical implementation of the critical threshold δ < α λ 1 / ( 2 + α ) requires estimating the mixing rate λ 1 . For complex geometries, λ 1 can be obtained numerically by solving the spectral problem associated with the transport operator v ⋅ ∇ x on the considered domain. An efficient approach is to discretize the operator and compute its smallest nonzero eigenvalue. For example, in a one-dimensional stratified medium where σ ( x , v ) = σ 0 ( 1 + ε cos ( k x ) ) , the mean transport can be approximated and λ 1 ≈ ( k 2 〈 v 2 〉 ) / 2 . This approach allows quantitative prediction of the degeneracy threshold in realistic configurations, such as magnetically confined plasmas or stratified porous media flows.

7.3. Limitations and Scope

Our analysis has several intentional limitations that define its scope:

Linearequations: Nonlinear collision operators or drift terms require different techniques. Wholespace ℝ 2 d : Boundary effects in physical domains introduce additional complexity. Time-independentcoefficients: Time-dependent σ ( t , x , v ) would require evolution semigroup methods. Moderatedegeneracy: For δ ≫ δ c , different phenomena (Anderson localization) may emerge.

These limitations suggest natural directions for future work while clarifying the contributions of the present paper.

7.4. Open Problems

OP1Optimalregularityconditions: Characterize the minimal assumptions on σ for hypoellipticity. Conjecture: L ∞ coefficients with geometric control condition suffices. A counterexample showing necessity of γ > α / 2 would be valuable.

OP2 Quantitativeestimateswithexplicitconstants: Obtain dimension-dependent constants in the hypoelliptic estimates, particularly the dependence on d for applications in high-dimensional phase spaces.

OP3 Nonlinearversions: Extend to equations with nonlinear drift or collision operators, e.g., ∂ t f + v ⋅ ∇ x f = − σ ( x , v , f ) ( − Δ v ) α / 2 f .

OP4 Boundaryvalueproblems: Develop theory for bounded domains with appropriate boundary conditions (specular reflection, absorption, etc.) and study boundary layer effects.

OP5 Numericalanalysisofthedegeneracytransition: Prove convergence rates for discretization schemes near the critical threshold δ c , and analyze numerical artifacts in the localized regime.

OP6 ConnectiontoAndersonlocalization: Investigate whether the supercritical regime ( δ > δ c ) exhibits features of Anderson localization in random media, potentially with random coefficients σ ( x , v , ω ) .

7.5. Applications Perspective

The results enable modeling of several physical systems:

Anomaloustransportinheterogeneousmedia: Particles moving in porous media with spatially-varying jump rates. The threshold provides a criterion for when heterogeneity causes localization.Plasmaphysics: Charged particles in turbulent fields with position-dependent collision frequencies. The mixing rate λ 1 relates to magnetic confinement efficiency.Financialmathematics: Price dynamics with state-dependent jump intensities (regime-switching Lévy processes). The degeneracy threshold may indicate regimes where price processes become trapped.Biologicaltransport: Intracellular transport with spatially-varying binding/unbinding rates. Localization could model organelle anchoring.

The degeneracy threshold provides a criterion for when spatial heterogeneity destroys macroscopic diffusive behavior, potentially explaining localization phenomena in disordered systems.

7.6. Final Remarks

The interplay between nonlocality (fractional operators), degeneracy (vanishing coefficients), and kinetic transport creates mathematical challenges requiring novel techniques. The weighted energy method developed here, combined with fractional commutator estimates, provides a framework that may be applicable to other degenerate nonlocal problems.

The clear separation between conditions allowing hypoellipticity ( δ < α λ 1 / ( 2 + α ) ) and those causing its loss demonstrates the delicate balance between dissipation and mixing in degenerate fractional kinetic equations. The numerical simulations reveal a sharp phase transition, suggesting potential connections to critical phenomena in statistical physics.

Acknowledgements

The authors thank the anonymous referees for their valuable suggestions that improved the paper.

Appendix: Technical Lemmas and Detailed Proofs

A.1. Proof of Theorem 16

Completeproof.

Step1:Regularization. Let η ε be a standard mollifier and define σ ε = σ ∗ η ε . Then σ ε ∈ C ∞ with σ min ≤ σ ε ≤ σ max and ‖ σ ε ‖ C 0 , γ ≤ ‖ σ ‖ C 0 , γ .

Step2:Finite-dimensionalapproximation. Let { ψ k } k = 1 ∞ be an orthonormal basis of L 2 ( ℝ 2 d ) consisting of smooth, compactly supported functions. Define V m = span { ψ 1 , ⋯ , ψ m } and seek

f m ( t , x , v ) = ∑ k = 1 m   c k m ( t ) ψ k ( x , v )

satisfying for k = 1 , ⋯ , m :

d d t 〈 f m , ψ k 〉 L 2 + 〈 v ⋅ ∇ x f m , ψ k 〉 L 2 + 〈 σ ε ( − Δ v ) α / 2 f m , ψ k 〉 L 2 = − 〈 F f m , ∇ v ψ k 〉 L 2 , c k m ( 0 ) = 〈 f 0 , ψ k 〉 L 2

Step3:Aprioriestimates. Multiply by c k m ( t ) and sum over k :

1 2 d d t ‖ f m ‖ L 2 2 = − 〈 σ ε ( − Δ v ) α / 2 f m , f m 〉 L 2 − 〈 F f m , ∇ v f m 〉 L 2 ≤ − σ min C α ‖ f m ‖ H v α / 2 2 + ‖ F ‖ L ∞ ( ε 2 ‖ ∇ v f m ‖ L 2 2 + 1 2 ε ‖ f m ‖ L 2 2 ) .

For α > 1 , we have ‖ ∇ v f ‖ L 2 ≤ C ‖ f ‖ H v α / 2 by Sobolev embedding. For α ≤ 1 , control directly via interpolation:

‖ ∇ v f ‖ L 2 2 ≤ δ ‖ f ‖ H v α / 2 2 + C δ ‖ f ‖ L 2 2 .

Choosing δ small enough gives:

d d t ‖ f m ‖ L 2 2 + 2 σ min C α ‖ f m ‖ H v α / 2 2 ≤ C F ‖ f m ‖ L 2 2 .

Step4:Compactness. Grönwall’s inequality yields uniform bounds:

‖ f m ‖ L ∞ ( 0 , T ; L 2 ) 2 + 2 σ min C α ∫ 0 T ‖ f m ‖ H v α / 2 2 d t ≤ e C F T ‖ f 0 ‖ L 2 2 .

From the equation, ∂ t f m is bounded in L 2 ( 0 , T ; H − 1 ( ℝ 2 d ) ) . By the Aubin-Lions lemma, there exists a subsequence (still denoted f m ) converging strongly in L 2 ( 0 , T ; L loc 2 ( ℝ 2 d ) ) to some f .

Step5:Passagetolimit. For any test function ϕ ∈ C c ∞ ( [ 0 , T ) × ℝ 2 d ) :

| 〈 σ ε ( − Δ v ) α / 2 f m , ϕ 〉 − 〈 σ ( − Δ v ) α / 2 f , ϕ 〉 | ≤ ‖ ( σ ε − σ ) ( − Δ v ) α / 2 f m ‖ L 2 ‖ ϕ ‖ L 2 + ‖ σ ‖ L ∞ ‖ ( − Δ v ) α / 2 ( f m − f ) ‖ L 2 ‖ ϕ ‖ L 2 → 0 ,

since σ ε → σ uniformly on compacts and f m ⇀ f in H v α / 2 .

Step6:Uniqueness. For two solutions f , g , let h = f − g . Then:

1 2 d d t ‖ h ‖ L 2 2 ≤ − σ min C α ‖ h ‖ H v α / 2 2 + C F ‖ h ‖ L 2 2 ≤ C F ‖ h ‖ L 2 2 .

Grönwall gives ‖ h ( t ) ‖ L 2 2 ≤ e C F t ‖ h ( 0 ) ‖ L 2 2 = 0 .

Step7:Non-negativityandmassconservation. Testing with ϕ = − f − gives d d t ‖ f − ‖ L 2 2 ≤ C ‖ f − ‖ L 2 2 , so f − ≡ 0 if f 0 ≥ 0 . Testing with ϕ ≡ 1 gives mass conservation.

Figure 13

A.2. Proof of Theorem 17

Completeproof. Similar to the proof of Theorem 16 but with careful handling of degeneracy. The key is to use the controlled degeneracy condition to obtain uniform estimates despite σ min = 0 .

Figure 14

A.3. Proof of Proposition 19

Proof. Test the equation with ( − Δ v ) α / 2 f and apply Theorem 11. The detailed calculation yields uniform bounds in H v α / 2 and then in H v α through energy estimates.

Figure 15

A.4. Proof of Theorem 23

Proof. Consider Q ( t ) = ‖ Λ x β f ( t ) ‖ L 2 2 . Differentiating and using the equation yields:

d Q d t = 2 ∫ Λ x β f ⋅ Λ x β ∂ t f = I trans + I diff + I force .

The transport term integrates to zero, the diffusion term gives negative dissipation via σ min , and the force term is controlled via commutator estimates. Integration in time and interpolation complete the proof.

Figure 16

A.5. Proof of Theorem 24

Completeproof.

Step1:Setup. Define w ( x , v ) = max ( σ ( x , v ) , δ ) and

Q ( t ) = ∫ ℝ 2 d 1 w ( x , v ) | Λ x β f | 2 d x   d v .

Step2:Timederivative. Compute:

d Q d t = 2 ∫ 1 w Λ x β f ⋅ Λ x β ∂ t f   d x   d v = 2 ∫ 1 w Λ x β f ⋅ Λ x β ( − v ⋅ ∇ x f − σ ( − Δ v ) α / 2 f + ∇ v ⋅ ( F f ) ) d x   d v = I trans + I diff + I force .

Step3:Transportterm I t r a n s . Using Theorem 22:

I trans = − 2 ∫ 1 w Λ x β f ⋅ Λ x β ( v ⋅ ∇ x f ) d x   d v = ∫ v ⋅ ∇ x ( 1 w ) | Λ x β f | 2 d x   d v .

Since

| ∇ x ( 1 / w ) | ≤ | ∇ x σ | / w 2 ≤ C ‖ σ ‖ C 0 , γ / δ 2 | I trans | ≤ C 1 δ 2 ‖ σ ‖ C 0 , γ ‖ Λ x β f ‖ L 2 2 .

Step4:Diffusionterm I d i f f .

I diff = − 2 ∫ 1 w Λ x β f ⋅ Λ x β ( σ ( − Δ v ) α / 2 f ) d x   d v = − 2 ∫ σ w Λ x β f ⋅ Λ x β ( − Δ v ) α / 2 f   d x   d v       − 2 ∫ 1 w Λ x β f ⋅ [ Λ x β , σ ] ( − Δ v ) α / 2 f   d x   d v .

The first term gives dissipation on Ω δ c :

− 2 ∫ σ w Λ x β f ⋅ Λ x β ( − Δ v ) α / 2 f   d x   d v ≤ − 2 ‖ Λ v α / 2 Λ x β f ‖ L 2 ( Ω δ c ) 2 .

The commutator term is controlled using commutator estimates for variable coefficients:

| ∫ 1 w Λ x β f ⋅ [ Λ x β , σ ] ( − Δ v ) α / 2 f   d x   d v | ≤ C 2 δ ‖ σ ‖ C 0 , γ ‖ Λ x β f ‖ L 2 ‖ Λ v α / 2 f ‖ H x β − 1 + γ .

Step5:Forceterm I f o r c e . Using Theorem 11:

| I force | ≤ C 3 δ ‖ F ‖ W 1 , ∞ ‖ Λ x β f ‖ L 2 2 + ε 2 ‖ Λ v α / 2 Λ x β f ‖ L 2 2 .

Step6:WeightedPoincaréinequalityon Ω δ . On the degeneracy set, we use the following inequality (proved below):

∫ Ω δ | Λ x β f | 2 d x   d v ≤ 1 λ 1 ∫ ℝ 2 d | v ⋅ ∇ x ( Λ x β f ) | 2 d x   d v + C δ ‖ f ‖ L 2 2 .

This expresses that mixing compensates for lack of dissipation.

Step7:Combiningestimates. Collecting terms:

d Q d t ≤ − ( 2 − ε ) ‖ Λ v α / 2 Λ x β f ‖ L 2 ( Ω δ c ) 2       + ( C 1 δ 2 ‖ σ ‖ C 0 , γ + C 3 δ ‖ F ‖ W 1 , ∞ − λ 1 w max ) ‖ Λ x β f ‖ L 2 2 + C 4 ‖ f ‖ L 2 2 , w max = max ( σ max , δ )

Step8:Thresholdcondition. For the coefficient of ‖ Λ x β f ‖ L 2 2 to be negative, we need:

λ 1 w max > C 1 δ 2 ‖ σ ‖ C 0 , γ + C 3 δ ‖ F ‖ W 1 , ∞ .

Under the condition δ < α λ 1 / ( 2 + α ) , this holds for sufficiently small δ and appropriate choice of constants.

Step9:Integrationandinterpolation. Integrating from 0 to T and using Theorem 12:

∫ 0 T ‖ Λ x β f ‖ L 2 2 d t ≤ C ( ∫ 0 T ‖ Λ v α / 2 f ‖ L 2 2 d t + ‖ f 0 ‖ L 2 2 ) .

The constant C depends on α , λ 1 , δ , ‖ σ ‖ C 0 , γ , ‖ F ‖ W 1 , ∞ , and d .

Figure 17

Lemma35(WeightedPoincaràinequalityondegeneracyset). Under the conditions of Theorem 24, there exists C δ > 0 such that:

∫ Ω δ | Λ x β f | 2 d x   d v ≤ 1 λ 1 ∫ ℝ 2 d | v ⋅ ∇ x ( Λ x β f ) | 2 d x   d v + C δ ‖ f ‖ L 2 2 .

Proof. Consider the operator ℒ = v ⋅ ∇ x on the space L 2 ( Ω δ ) . By definition of λ 1 as the spectral gap, for any g with ∫ Ω δ g = 0 :

∫ Ω δ | g | 2 ≤ 1 λ 1 ∫ Ω δ | ℒ g | 2 .

For g = Λ x β f , we decompose g = g 0 + g ¯ where g ¯ = 1 | Ω δ | ∫ Ω δ g and ∫ Ω δ g 0 = 0 . Then:

∫ Ω δ | g | 2 ≤ 2 ∫ Ω δ | g 0 | 2 + 2 | Ω δ | | g ¯ | 2 ≤ 2 λ 1 ∫ Ω δ | ℒ g 0 | 2 + 2 | Ω δ | | g ¯ | 2 .

Since ℒ g 0 = ℒ g (as ℒ g ¯ = 0 for constant g ¯ ) and | g ¯ | 2 ≤ 1 | Ω δ | ∫ Ω δ | g | 2 , we obtain after absorption:

∫ Ω δ | g | 2 ≤ 4 λ 1 ∫ ℝ 2 d | ℒ g | 2 + C δ ‖ f ‖ L 2 2 ,

using that ‖ g ‖ L 2 ( Ω δ ) ≤ C ‖ f ‖ H x β ≤ C ′ ‖ f ‖ L 2 by interpolation for β small.

Figure 18

A.6. Proof of Theorem 26

Proof. Approximate σ by mollification σ ε , apply Theorem 24 to σ ε , and pass to the limit using the controlled degeneracy assumption. The additional ϵ ‖ f ‖ L 2 ( 0 , T ; L 2 ) 2 term accounts for potential loss of compactness due to minimal regularity.

Figure 19

A.7. Proof of Theorem 29

Proof. We use the relative entropy method adapted to variable coefficients. Define:

H ε ( t ) = ∫ ℝ d ∫ ℝ d f ε log ( f ε M ε ) d v   d x ,     M ε ( v ; ρ ) = ρ ( t , x ) σ ¯ ( x ) g α ε ( v ) .

Compute d H ε d t , show dissipation dominates remainder terms as ε → 0 , and conclude H ε ( t ) → 0 . The Csiszár-Kullback-Pinsker inequality gives strong convergence. The main technical difficulty is handling the variable coefficient σ ( x , ε 1 / α v ) in the limit, which is addressed by Taylor expansion and moment estimates.

Figure 20

A.8. Proof of Proposition 31

Proof. The rate comes from analyzing the remainder terms R 1 ε , R 2 ε more carefully using Taylor expansions of σ and moment estimates on g α . For α > 1 , the rate is O ( ε ) ; for α ≤ 1 , it becomes O ( ε 2 / α ) due to slower decay of the α -stable density moments.

Figure 21

Glossary of Symbols

Table 4

Symbol	Definition
f ( t , x , v )	Probability density at time t , position x , velocity v
α	Lévy exponent ( 0 < α < 2 )
β	Spatial hypoelliptic exponent
σ ( x , v )	Multiplicative noise intensity
F ( x , v )	External force field
( − Δ v ) α / 2	Fractional Laplacian in velocity: ( − Δ v ) α / 2
( − Δ x ) β / 2	Fractional Laplacian in space: ( − Δ x ) β / 2
Λ x β	Alternative notation for ( − Δ x ) β / 2
Λ v α / 2	Alternative notation for ( − Δ v ) α / 2
H s ( ℝ d )	Sobolev space on ℝ d
H v s ( ℝ 2 d )	Velocity Sobolev space on ℝ 2 d
H x s ( ℝ 2 d )	Spatial Sobolev space on ℝ 2 d
ℋ	Hilbert space for solutions
λ 1	Mixing rate (inverse characteristic mixing time)
δ	Degeneracy parameter (measure of where σ vanishes)
ε	Scaling parameter in hydrodynamic limit
g α ( v )	α -stable density
σ ¯ ( x )	Effective diffusion coefficient
Ω δ	δ -degeneracy set: { ( x , v ) : σ ( x , v ) ≤ δ }
w ( x , v )	Weight function: max ( σ ( x , v ) , δ )
τ mix	Characteristic mixing time: τ mix ~ 1 / λ 1

References

1.. Bouchaud, J. and Georges, A. (1990) Anomalous Diffusion in Disordered Media: Statistical Mechanisms, Models and Physical Applications. PhysicsReports, 195, 127-293. https://doi.org/10.1016/0370-1573(90)90099-n 10.1016/0370-1573(90)90099-nhttps://doi.org/10.1016/0370-1573(90)90099-n Bouchaud, J. Georges, A. Mechanisms, M 1990 Anomalous Diffusion in Disordered Media: Statistical Mechanisms, Models and Physical Applications Physics Reports 1573 90 10.1016/0370-1573(90)90099-n

2.. Metzler, R. and Klafter, J. (2000) The Random Walk’s Guide to Anomalous Diffusion: A Fractional Dynamics Approach. PhysicsReports, 339, 1-77. https://doi.org/10.1016/s0370-1573(00)00070-3 10.1016/s0370-1573(00)00070-3https://doi.org/10.1016/s0370-1573(00)00070-3 Metzler, R. Klafter, J. 2000 The Random Walk’s Guide to Anomalous Diffusion: A Fractional Dynamics Approach Physics Reports 1573 00 10.1016/s0370-1573(00)00070-3

3.. Cont, R. and Tankov, P. (2004) Financial Modelling with Jump Processes. Chapman & Hall/CRC. https://doi.org/10.1201/9780203485217 10.1201/9780203485217https://doi.org/10.1201/9780203485217 Cont, R. Tankov, P. 2004 Financial Modelling with Jump Processes 10.1201/9780203485217

4.. Dolbeault, J., Mouhot, C. and Schmeiser, C. (2015) Hypocoercivity for Linear Kinetic Equations Conserving Mass. Transactions of the American Mathematical Society, 367, 3807-3828. https://doi.org/10.1090/s0002-9947-2015-06012-7 10.1090/s0002-9947-2015-06012-7https://doi.org/10.1090/s0002-9947-2015-06012-7 Dolbeault, J. Mouhot, C. Schmeiser, C. 2015 Hypocoercivity for Linear Kinetic Equations Conserving Mass Transactions of the American Mathematical Society 367 10.1090/s0002-9947-2015-06012-7

5.. Mellet, A., Mischler, S. and Mouhot, C. (2010) Fractional Diffusion Limit for Collisional Kinetic Equations. ArchiveforRationalMechanicsandAnalysis, 199, 493-525. https://doi.org/10.1007/s00205-010-0354-2 10.1007/s00205-010-0354-2https://doi.org/10.1007/s00205-010-0354-2 Mellet, A. Mischler, S. Mouhot, C. 2010 Fractional Diffusion Limit for Collisional Kinetic Equations Archive for Rational Mechanics and Analysis 199 10.1007/s00205-010-0354-2

6.. Hörmander, L. (1967) Hypoelliptic Second Order Differential Equations. ActaMathematica, 119, 147-171. https://doi.org/10.1007/bf02392081 10.1007/bf02392081https://doi.org/10.1007/bf02392081 1967 Hypoelliptic Second Order Differential Equations Acta Mathematica 119 10.1007/bf02392081

7.. Villani, C. (2009) Hypocoercivity. MemoirsoftheAmericanMathematicalSociety, 202, 141 p. https://doi.org/10.1090/s0065-9266-09-00567-5 10.1090/s0065-9266-09-00567-5https://doi.org/10.1090/s0065-9266-09-00567-5 Villani, C. 2009 Hypocoercivity Memoirs of the American Mathematical Society 202 10.1090/s0065-9266-09-00567-5

8.. Kato, T. and Ponce, G. (1988) Commutator Estimates and the Euler and Navier‐stokes Equations. CommunicationsonPureandAppliedMathematics, 41, 891-907. https://doi.org/10.1002/cpa.3160410704 10.1002/cpa.3160410704https://doi.org/10.1002/cpa.3160410704 Kato, T. Ponce, G. 1988 Commutator Estimates and the Euler and Navier‐stokes Equations Communications on Pure and Applied Mathematics 41 10.1002/cpa.3160410704

9.. Caffarelli, L. and Silvestre, L. (2007) An Extension Problem Related to the Fractional Laplacian. CommunicationsinPartialDifferentialEquations, 32, 1245-1260. https://doi.org/10.1080/03605300600987306 10.1080/03605300600987306https://doi.org/10.1080/03605300600987306 Caffarelli, L. Silvestre, L. 2007 An Extension Problem Related to the Fractional Laplacian Communications in Partial Differential Equations 32 10.1080/03605300600987306

10.. Loher, A. (2024) Quantitative De Giorgi Methods in Kinetic Theory for Non-Local Operators. Journal of Functional Analysis, 286, Article ID: 110312. https://doi.org/10.1016/j.jfa.2023.110312 10.1016/j.jfa.2023.110312https://doi.org/10.1016/j.jfa.2023.110312 Loher, A. 2024 Quantitative De Giorgi Methods in Kinetic Theory for Non-Local Operators Journal of Functional Analysis 286 110312 ID 10.1016/j.jfa.2023.110312

11.. Kassmann, M. and Weidner, M. (2024) The Harnack Inequality Fails for Nonlocal Kinetic Equations. Advances in Mathematics, 459, Article ID: 110030. https://doi.org/10.1016/j.aim.2024.110030 10.1016/j.aim.2024.110030https://doi.org/10.1016/j.aim.2024.110030 Kassmann, M. Weidner, M. 2024 The Harnack Inequality Fails for Nonlocal Kinetic Equations Advances in Mathematics 459 110030 ID 10.1016/j.aim.2024.110030

12.. Auscher, P., Imbert, C. and Niebel, L. (2025) Fundamental Solutions to Kolmogorov–fokker–planck Equations with Rough Coefficients: Existence, Uniqueness, Upper Estimates. SIAM Journal on Mathematical Analysis, 57, 2114-2137. https://doi.org/10.1137/24m1649241 10.1137/24m1649241https://doi.org/10.1137/24m1649241 Auscher, P. Imbert, C. Niebel, L. Existence, U 2025 Fundamental Solutions to Kolmogorov–fokker–planck Equations with Rough Coefficients: Existence, Uniqueness, Upper Estimates SIAM Journal on Mathematical Analysis 57 10.1137/24m1649241

13.. Niebel, L. (2022) Kinetic Maximal-Regularity for the Fractional Kolmogorov Equation with Variable Density. Nonlinear Analysis, 214, Article ID: 112517. https://doi.org/10.1016/j.na.2021.112517 10.1016/j.na.2021.112517https://doi.org/10.1016/j.na.2021.112517 Niebel, L. 2022 Kinetic Maximal-Regularity for the Fractional Kolmogorov Equation with Variable Density Nonlinear Analysis 214 112517 ID 10.1016/j.na.2021.112517

14.. Kim, K. (2024) A Regularity Theory for Parabolic Equations with Anisotropic Nonlocal Operators in Spaces. SIAM Journal on Mathematical Analysis, 56, 1089-1156. https://doi.org/10.1137/23m1574944 10.1137/23m1574944https://doi.org/10.1137/23m1574944 Kim, K. 2024 A Regularity Theory for Parabolic Equations with Anisotropic Nonlocal Operators in Spaces SIAM Journal on Mathematical Analysis 56 10.1137/23m1574944

15.. Bonami, A., Grellier, S. and Ky, L.D. (2012) Paraproducts and Products of Functions in BMO( Rⁿ) and H¹( Rⁿ) through Wavelets. Journal de Mathématiques Pures et Appliquées, 97, 230-241. https://doi.org/10.1016/j.matpur.2011.06.002 10.1016/j.matpur.2011.06.002https://doi.org/10.1016/j.matpur.2011.06.002 Bonami, A. Grellier, S. Ky, L.D. 2012 Paraproducts and Products of Functions in BMO(Rn) and H1(Rn) through Wavelets Journal de Mathématiques Pures et Appliquées 97 10.1016/j.matpur.2011.06.002

16.. Thangavelu, S. (2018) A Note on Fractional Powers of the Hermite Operator. https://arxiv.org/abs/1801.08343 Thangavelu, S. 2018 A Note on Fractional Powers of the Hermite Operator