Abstract:We find explicit upper bounds for the density of marginals of continuous diffusions where we assume that the diffusion coefficient is constant and the drift is solely assumed to be progressively measurable and locally bounded. In one dimension we extend our result to the case that the diffusion coefficient is a locally Lipschitz-continuous function of the state. Our approach is based on a comparison to a suitable doubly reflected Brownian motion whose density is known in a series representation.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is to provide explicit local upper bounds for the density of Itô - processes with irregular drift terms. Specifically, the authors focus on whether, under certain conditions, the density functions of the solutions of these processes can be effectively estimated, and give specific mathematical expressions to describe these upper bounds.
### Background and Problem Statement of the Paper
1. **Research Background**:
- The article discusses stochastic differential equations (SDEs) and the properties of their solutions, especially the probability density functions of these solutions.
- For multi - dimensional SDEs driven by Brownian motion, people are usually concerned about what conditions the coefficients should satisfy so that the density function of the solution is absolutely continuous with respect to the Lebesgue measure, and hope to find specific upper bounds of the density function.
2. **Existing Methods**:
- Traditional methods rely on regularity assumptions of SDEs coefficients, such as Malliavin calculus, Dirichlet forms, etc.
- Baños & Krühner [BK16] proposed a method to identify the worst - case SDE through control theory techniques, thereby obtaining the optimal estimation of density bounds.
3. **Innovations of This Paper**:
- The goal of this paper is to provide explicit local density upper bounds under weaker assumptions (e.g., locally bounded drift terms and constant diffusion coefficients).
- Compared with previous works, this paper does not require the drift term to be globally bounded or have specific regularity, nor is it restricted to Markov processes.
### Main Results
- **Theorem 2.2**: For a d - dimensional Itô process X with a locally bounded drift and a constant diffusion coefficient, within an open set U, when time t > 0, the density ρ_t(x) of X(t) is locally bounded.
Expressed by the formula:
\[
\rho_t(x):=\limsup_{\epsilon \to 0}\frac{P(||X(t) - x|| < \epsilon)}{\text{vol}(B_{\epsilon,\infty}(0))}, \quad x\in U
\]
and ρ_t is a version of the density of X(t) on U.
- **Proposition 2.3**: When the diffusion coefficient is the identity matrix and the drift term is bounded within the open set U, a specific formula for the density upper bound is provided:
\[
\rho_t(x)\leq d\prod_{j = 1}^d\left(\frac{C\exp(- 2Cl)}{1-\exp(-2Cl)}+\phi(z_j)\sqrt{t}+C\Phi(z_j)+e^{Ca_j - C^2t/2}\left(\frac{3 + a_jC}{ltC^2}\right)^2\right)
\]
where \(a_j=\min\{l,|X_j(0)-x_j|\}\), \(z_j = \sqrt{t}C - a_j/\sqrt{t}\), \(j = 1,\ldots,d\).
### Methodology
- The authors connect the density of the Itô process with the transition density of the double - reflected Brownian motion, and use the known series expansion form of the latter to derive the explicit upper bound of the former.
- Using the Itô - Tanaka formula, the main results are extended to the case with locally Lipschitz continuous diffusion coefficients.
In conclusion, through introducing new comparison methods and techniques, this paper successfully provides explicit local upper bounds for the density of Itô processes under weaker assumption conditions, which is of great significance for further understanding the behavior of solutions to stochastic differential equations.