Viscous Hamilton-Jacobi Equation

• Viscous Hamilton-Jacobi equation is a nonlinear, nonlocal evolution equation that integrates Hamiltonian dynamics with translation-invariant Lévy diffusion.
• The analytical framework employs the Duhamel formula and fixed-point methods to establish well-posedness with robust Schauder and Lipschitz regularity estimates.
• Sharp heat kernel bounds and comparison principles underpin the global existence and uniqueness results, facilitating analysis of both local and nonlocal diffusive processes.

The viscous Hamilton-Jacobi equation is a class of nonlinear, nonlocal, and possibly anisotropic evolution equations central to the analysis of regularization phenomena, stochastic control, homogenization, and nonlocal dynamics. Its canonical form incorporates both a Hamiltonian nonlinearity and a translation-invariant Lévy-type diffusion operator, allowing the modeling of broad classes of local and nonlocal diffusive processes. Advanced regularity theory—Schauder and Lipschitz estimates—along with well-posedness and representation results, underpin its modern mathematical understanding and applications.

1. Equation Formulation and Analytical Framework

The general Cauchy problem for the viscous Hamilton–Jacobi equation is

$$\partial_t u(t,x) - \mathcal{L}u(t,x) + H(t,x,u(t,x), Du(t,x)) = f(t,x), \quad (t,x) \in (0,T) \times \mathbb{R}^d,$$

with initial data

$u(0,x) = u_0(x).$

The diffusion operator $\mathcal{L}$ is of Lévy type: $$\mathcal{L}u(x) = B \cdot Du(x) + \text{Div}(A Du(x)) + \int_{\mathbb{R}^d} [u(x+z) - u(x) - Du(x) \cdot z \mathbf{1}_{|z|<1}] \, \nu(dz),$$ where the Lévy triplet $(B,A,\nu)$ encodes drift, local diffusion, and jump (nonlocal) terms, with $A \geq 0$, and $\int (1 \wedge |z|^2)\nu(dz) < \infty$. The operator is called subcritical of order $\alpha_{\text{low}} \in (1,2]$ if its heat kernel enjoys certain $L^1$ bounds.

The Hamiltonian $H(t,x,u,p)$ is continuous and satisfies:

• (H0) Local Lipschitz in (x,p);
• (H1) Local boundedness up to second derivatives in (x,u,p);
• (H2) Controlled x-dependence;
• (H3) Monotonicity in u.

The source term $u(0,x) = u_0(x).$0 is in $u(0,x) = u_0(x).$1 and has uniform spatial Lipschitz bounds.

2. Mild Solution and Duhamel Representation

Existence and regularity theory are grounded in the Duhamel integral formula. For the heat kernel $u(0,x) = u_0(x).$2 of $u(0,x) = u_0(x).$3 (with Fourier symbol $u(0,x) = u_0(x).$4),

$u(0,x) = u_0(x).$5

and, for $u(0,x) = u_0(x).$6,

$u(0,x) = u_0(x).$7

This yields a fixed-point variational formulation enabling analysis and construction of solutions for general initial data and source terms.

3. Heat Kernel Bounds and Fractional Diffusion

Regularity theory relies on sharp heat kernel estimates: $u(0,x) = u_0(x).$8 with analogous bounds for the adjoint kernel $u(0,x) = u_0(x).$9. These facilitate the control of spatial derivatives and estimate solutions and their gradients in Hölder or Lipschitz spaces. Integral bounds over time intervals, such as

$\mathcal{L}$0

and

$\mathcal{L}$1

for $\mathcal{L}$2, are central to nonlinear analysis, especially for fractional and nonlocal diffusions.

4. Schauder and Lipschitz Regularity Theory

Schauder estimates (Theorem 4.1) establish that, for $\mathcal{L}$3, $\mathcal{L}$4, and suitable $\mathcal{L}$5,

$\mathcal{L}$6

for every $\mathcal{L}$7. The solution

$\mathcal{L}$8

demonstrates full classical regularity away from $\mathcal{L}$9.

Global Lipschitz gradient bounds (Theorem 3.2): If $$\mathcal{L}u(x) = B \cdot Du(x) + \text{Div}(A Du(x)) + \int_{\mathbb{R}^d} [u(x+z) - u(x) - Du(x) \cdot z \mathbf{1}_{|z|<1}] \, \nu(dz),$$0,

$$\mathcal{L}u(x) = B \cdot Du(x) + \text{Div}(A Du(x)) + \int_{\mathbb{R}^d} [u(x+z) - u(x) - Du(x) \cdot z \mathbf{1}_{|z|<1}] \, \nu(dz),$$1

establishing propagation and control of spatial Lipschitz constants.

5. Existence, Uniqueness, and Comparison Principle

Existence of smooth solutions proceeds via a fixed-point argument in the function space $$\mathcal{L}u(x) = B \cdot Du(x) + \text{Div}(A Du(x)) + \int_{\mathbb{R}^d} [u(x+z) - u(x) - Du(x) \cdot z \mathbf{1}_{|z|<1}] \, \nu(dz),$$2, $$\mathcal{L}u(x) = B \cdot Du(x) + \text{Div}(A Du(x)) + \int_{\mathbb{R}^d} [u(x+z) - u(x) - Du(x) \cdot z \mathbf{1}_{|z|<1}] \, \nu(dz),$$3:

• Short-time existence employs a Banach contraction in the Duhamel map.
• Regularity upgrade ensures mild solutions are classical due to the time/spatial regularity.
• Global existence uses iteration and the global gradient bound.
• Uniqueness is enforced via a comparison principle, provided (H0), (H2), (H3), and (F0) are satisfied.

6. Scope: General Diffusion Operators

The developed theory extends to all translation-invariant Lévy operators whose kernels satisfy the above $$\mathcal{L}u(x) = B \cdot Du(x) + \text{Div}(A Du(x)) + \int_{\mathbb{R}^d} [u(x+z) - u(x) - Du(x) \cdot z \mathbf{1}_{|z|<1}] \, \nu(dz),$$4 bounds, including:

• Uniformly elliptic local diffusion ($$\mathcal{L}u(x) = B \cdot Du(x) + \text{Div}(A Du(x)) + \int_{\mathbb{R}^d} [u(x+z) - u(x) - Du(x) \cdot z \mathbf{1}_{|z|<1}] \, \nu(dz),$$5, e.g., $$\mathcal{L}u(x) = B \cdot Du(x) + \text{Div}(A Du(x)) + \int_{\mathbb{R}^d} [u(x+z) - u(x) - Du(x) \cdot z \mathbf{1}_{|z|<1}] \, \nu(dz),$$6 with $$\mathcal{L}u(x) = B \cdot Du(x) + \text{Div}(A Du(x)) + \int_{\mathbb{R}^d} [u(x+z) - u(x) - Du(x) \cdot z \mathbf{1}_{|z|<1}] \, \nu(dz),$$7);
• Fractional Laplacian-type ($$\mathcal{L}u(x) = B \cdot Du(x) + \text{Div}(A Du(x)) + \int_{\mathbb{R}^d} [u(x+z) - u(x) - Du(x) \cdot z \mathbf{1}_{|z|<1}] \, \nu(dz),$$8 for $$\mathcal{L}u(x) = B \cdot Du(x) + \text{Div}(A Du(x)) + \int_{\mathbb{R}^d} [u(x+z) - u(x) - Du(x) \cdot z \mathbf{1}_{|z|<1}] \, \nu(dz),$$9, $(B,A,\nu)$0);
• Strongly anisotropic sums $(B,A,\nu)$1, $(B,A,\nu)$2, $(B,A,\nu)$3;
• Spectrally one-sided Riesz-Feller operators on $(B,A,\nu)$4 ($(B,A,\nu)$5);
• CGMY-type models in finance;
• Arbitrary finite or infinite sums of the above.

This furnishes a robust Schauder theory and well-posedness results for viscous Hamilton-Jacobi equations driven by general local, nonlocal, or mixed Lévy diffusions.

---

These findings are comprehensively surveyed in "Towards a Schauder theory for fractional viscous Hamilton–Jacobi equations" (Jakobsen et al., 2024). The methodological pillars—Duhamel formula, heat kernel bounds, and comparison principles—allow the propagation of regularity and the extension of the analytical framework to strongly anisotropic, nonsymmetric, and spectrally one-sided diffusion models, solidifying the modern theory of viscous HJ equations in fractional and Lévy environments.