The transition problem between time-independent motions of a body in a viscous liquid

Technical Summary: The Transition Problem between Time-Independent Motions of a Body in a Viscous Liquid

Problem Formulation
This paper addresses the "transition problem" for a rigid body $B$ moving in an unbounded, three-dimensional viscous Navier-Stokes liquid filling the exterior domain $\Omega = \mathbb{R}^3 \setminus B$. The study investigates the asymptotic behavior of the fluid when the body's motion changes from an initial steady translatory state to a final, arbitrary steady rigid motion.

Specifically, let the body initially translate with a constant velocity $\gamma_1$ (with zero angular velocity), generating a steady-state flow $(v_1, p_1)$. At time $t=0$, the body undergoes a smooth transition over the interval $(0, 1)$ to a final state characterized by a constant translational velocity $\gamma_2$ and a constant angular velocity $\omega_0$. This final state generates a different steady flow $(v_2, p_2)$. The central question is whether the unsteady flow generated during the transition converges to the final steady state $v_2$ as $t \to \infty$.

This formulation generalizes the classical "starting problem" (where $\gamma_1 = 0$ and $\omega_0 = 0$) and extends previous work by Sazonov, which was limited to cases where the initial and final velocities were parallel and the domain was the whole space (no body).

Methodology
The authors employ a functional analytic approach based on the evolution operator theory for non-autonomous linearized Navier-Stokes equations.

1. Reformulation: The problem is transformed by decomposing the total velocity field $v(x,t)$ into the final steady state $v_2$, a lifting function $b(x,t)$ handling the time-dependent boundary conditions, and a perturbation $w(x,t)$. The goal becomes proving that $w(t)$ decays to zero as $t \to \infty$.
2. Integral Equation: The evolution of the perturbation $w$ is recast as a nonlinear integral equation in a suitable function space:
   $$w(t) = T(t, 0)v_0 + \int_0^t T(t, s)Pg(s) \, ds + \int_0^t T(t, s)P\text{div}(Fw)(s) \, ds$$
   Here, $T(t, s)$ is the evolution operator associated with the linearized operator around the time-dependent motion of the body, constructed in prior work by Hansel and Rhandi. $P$ denotes the Helmholtz projection.
3. Function Spaces: The analysis is conducted in Lorentz spaces $L^{q,r}(\Omega)$, specifically focusing on the weak-$L^3$ space $L^{3,\infty}_\sigma(\Omega)$. This choice is critical because the steady-state solutions in exterior domains possess a "wake structure" that prevents them from belonging to standard $L^q$ spaces for $q \le 3$, but they do belong to $L^{3,\infty}$.
4. Fixed Point Argument: The existence and uniqueness of the solution are established using the contraction mapping theorem. This requires deriving precise decay estimates for the evolution operator $T(t, s)$ and its adjoint, as well as handling the nonlinear term $Fw$ (representing convective terms) and the forcing term $g$ (arising from the time-dependence of the transition).

Key Contributions and Results

• General Transition: The paper proves that for sufficiently small initial and final velocities ($\gamma_1, \gamma_2$) and angular velocity ($\omega_0$), there exists a unique global solution connecting the two steady states. Unlike previous results, the final state allows for arbitrary rotation ($\omega_0 \neq 0$) and non-parallel velocities.
• Decay Rates: The authors establish algebraic decay rates for the perturbation $w(t)$ as $t \to \infty$. Specifically, for any $\epsilon \in (0, 1/4)$, the solution satisfies:
  • $\|w(t)\|_{(3,\infty)} \le C(1 + |\psi'|_0)D$
  • $\|w(t)\|_r = O(t^{-1/2 + \epsilon})$ for $r > q_0$ (where $q_0 = 3/2\epsilon$)
  • $\|\nabla w(t)\|_{(3,\infty)} = O(t^{-1/2 + \epsilon})$
    Here, $D = |\gamma_1|^{1/2} + |\gamma_2| + |\omega_0|$ represents the magnitude of the data.
• Handling Rotation: A significant technical achievement is the treatment of the term $(\omega_0 \times x) \cdot \nabla v_2$. The authors show that if the initial state is purely translatory ($\omega_1 = 0$), the forcing term $g$ remains in $L^{3,\infty}$ even when $\omega_0 \neq 0$. They note that if the initial state also had rotation ($\omega_1 \neq 0$), the transition problem remains open due to difficulties in estimating the forcing term.
• Uniqueness: The paper proves that the solution is unique within a class of global solutions, even if a competing solution is large, provided the data satisfies a specific smallness condition.

Significance and Claims
The authors position this work as a rigorous solution to the transition problem in a general setting, overcoming the limitation of previous results that required parallel velocities or the absence of a body.

• Motivational Context: The paper acknowledges that the "starting problem" (accelerating from rest) was solved by Shibata and others for small velocities. This work extends that to the "transition" between two arbitrary steady states.
• Limitations and Open Questions: The authors are modest about the optimality of their decay rates. They conjecture that the $L^\infty$ decay rate might be $O(t^{-1/2})$, but their current proof yields $O(t^{-1/2+\epsilon})$. They explicitly state that if the initial state involves rotation ($\omega_1 \neq 0$), the problem remains unsolved because the necessary estimates for the forcing term are not available.
• Methodological Contribution: The work relies heavily on the $L^q-L^r$ decay estimates of the evolution operator $T(t, s)$ and its adjoint, refining the application of these estimates to the nonlinear transition problem in exterior domains with moving boundaries.

In summary, the paper provides a positive answer to the transition problem for a body moving from a translatory state to an arbitrary rigid motion, provided the velocities are sufficiently small, establishing the existence, uniqueness, and asymptotic decay of the connecting flow.