Convergence to global equilibrium for the semiconductor Boltzmann equation

Technical Summary: Convergence to Global Equilibrium for the Semiconductor Boltzmann Equation

Problem Statement
This paper addresses the long-time behavior of the semi-classical Boltzmann equation describing electron transport in semiconductor devices under an external electrostatic potential. The system is defined on a spatial torus $T^d$ and a wave vector domain $B$ (the Brillouin zone), which may be a bounded set or $\mathbb{R}^d$. The governing equation is:
$$\partial_t f + \nabla_k E(k) \cdot \nabla_x f - \nabla_x V(x) \cdot \nabla_k f = Q(f)$$
where $f(t,x,k)$ is the distribution function, $E(k)$ is the electron energy function, $V(x)$ is a given external electrostatic potential, and $Q(f)$ is the collision operator incorporating the Pauli exclusion principle. The study focuses on proving the exponential decay of solutions towards the global equilibrium state, the Fermi-Dirac distribution $f_\infty$, without relying on restrictive assumptions common in prior literature.

Methodology
The authors employ a hypocoercivity framework, specifically the construction of a modified Lyapunov functional, to overcome the lack of dissipation in the spatial direction inherent to kinetic equations. The methodology proceeds through the following steps:

1. Existence and Uniqueness: The paper first establishes the existence of a unique distributional solution $f \in C([0, \infty), L^1(T^d \times B))$ that satisfies specific bounds relative to the equilibrium distribution. This is achieved via a fixed-point argument on a linearized transport equation, ensuring the solution remains within physically relevant bounds ($0 < f < 1$).
2. Relative Entropy and Dissipation: The standard relative entropy $H(f|f_\infty)$ is analyzed. It is shown that its time derivative equals the negative of a dissipation functional $D(f)$, which is non-positive. However, $D(f)$ only controls the deviation of $f$ from a local equilibrium in the velocity variable, not the full spatial-velocity deviation.
3. Non-linear Projection: A non-linear projection operator $\Pi f$ is introduced, mapping the distribution $f$ to a local Fermi-Dirac distribution with a spatially dependent chemical potential $\mu(t,x)$ that preserves the local mass density. Estimates are derived relating the $L^2$ distance between $f$ and $\Pi f$ to the dissipation functional $D(f)$.
4. Construction of the Lyapunov Functional: The core technical innovation is the construction of a modified Lyapunov functional $\mathcal{E}(f)$:
   $$\mathcal{E}(f) = H(f|f_\infty) + \delta \int_{T^d} \nabla_x \phi \cdot J \, dx$$
   Here, $\phi$ is the solution to an elliptic equation driven by the mass density difference $(\rho - \rho_\infty)$, and $J$ is a current-like vector function involving the logarithmic ratio of the distribution. The parameter $\delta > 0$ is chosen sufficiently small.
5. Hypocoercive Estimates: By differentiating $\mathcal{E}(f)$ in time, the authors derive a differential inequality. The cross-term involving $\phi$ and $J$ couples the spatial and velocity dynamics, allowing the dissipation from the collision operator to control the transport term. This leads to a differential inequality of the form $\frac{d}{dt}\mathcal{E}(f) \leq -2\lambda \mathcal{E}(f)$.

Key Contributions

• Removal of Parabolic Band Approximation: Unlike previous works that often assumed a quadratic energy dispersion $E(k) = |k|^2$ (parabolic band), this result holds for general smooth energy functions $E(k)$, provided the matrix $\int_B \partial^2_{k} E(k) e^{-E(k)} dk$ is positive definite. This includes the physically relevant case where the wave vector $k$ is confined to a bounded Brillouin zone.
• Removal of Electrostatic Potential Restriction: The analysis accommodates non-zero, smooth external electrostatic potentials $V(x)$, whereas prior exponential decay results often required $V \equiv 0$.
• Relaxation of Initial Data Assumptions: The paper removes the "close-to-equilibrium" assumption on the initial datum. The initial data is only required to be bounded by the equilibrium distribution (i.e., $f_0$ lies between two Fermi-Dirac distributions with different chemical potentials), rather than being a small perturbation in a Sobolev norm.
• Explicit and Constructive Rates: The convergence rate $\lambda$ and the constant $c$ in the decay estimate are explicitly computable based on the physical parameters and the bounds of the initial data.

Main Results
The primary result (Theorem 2.1) states that under the stated conditions on $E$, $V$, and the collision kernel $\sigma$, the unique solution $f(t)$ converges exponentially to the global equilibrium $f_\infty$ in a weighted $L^2$ norm:
$$\int_{T^d} \int_B (f(t, x, k) - f_\infty(x, k))^2 e^{E(k)} \, dk \, dx \leq c e^{-2\lambda t} \int_{T^d} \int_B (f_0(x, k) - f_\infty(x, k))^2 e^{E(k)} \, dk \, dx$$
This result implies that the relative entropy also decays exponentially. The proof relies on the equivalence of the modified Lyapunov functional $\mathcal{E}(f)$ to the squared weighted $L^2$ norm and the derived differential inequality.

Significance
The paper claims to improve upon existing results in the field of semiconductor kinetic theory (specifically citing works by Neumann & Schmeiser, Chen et al., and Pirner & the author). By eliminating the need for the parabolic band approximation, the zero-potential assumption, and the smallness of initial data, the authors provide a more robust and physically realistic framework for analyzing the long-time behavior of semiconductor devices. The technique of modifying the relative entropy with a specific non-linear functional involving an elliptic potential is presented as a novel approach that may be applicable to other semi-classical kinetic equations. The results are constructive, offering explicit constants for the convergence rate, which is valuable for quantitative analysis.