Dice Question Streamline Icon: https://streamlinehq.com

Controllability without surface diffusion in the dynamic boundary condition

Determine the null and approximate controllability at time T for the forward stochastic parabolic system on Q = (0, T) × G with dynamic boundary conditions when the boundary equation has no surface diffusion term, namely: dy − ∇·(𝒜∇y) dt = (a1 y + B1·∇y + 1_{G0} u) dt + v1 dW(t) in Q, dy_Γ + ∂_ν^𝒜 y dt = (a2 y_Γ + B2·∇_Γ y_Γ) dt + v2 dW(t) on Σ = (0, T) × Γ, and y_Γ = y|_Γ on Σ, where 𝒜 and 𝒜_Γ are uniformly elliptic diffusion matrices and a1, a2, B1, B2 are bounded adapted coefficients. Establish whether controls (u, v1, v2) can drive the state to zero (null controllability) or arbitrarily close to a target (approximate controllability) under this boundary configuration.

Information Square Streamline Icon: https://streamlinehq.com

Background

In the main results, the authors rely on a surface diffusion term on the boundary, ∇Γ·(𝒜Γ∇Γ yΓ), to absorb a critical boundary integral in their Carleman estimate. This term is essential in their proof of Lemma 2.2 (Step 2 in Appendix A).

The open problem concerns replacing the boundary diffusion operator with a purely dynamic boundary coupling dy_Γ + ∂ν𝒜 y dt = (a2 yΓ + B2·∇Γ yΓ) dt + v2 dW(t), i.e., removing the surface Laplace–Beltrami diffusion term. The authors note that known techniques fail to absorb problematic boundary contributions in this setting.

They point out that in one spatial dimension (N = 1), controllability under such boundary conditions is known, but for higher dimensions it remains unresolved.

References

However, the study of the controllability of 1.4 with the following dynamic boundary condition: \begin{equation*} \begin{cases} dy_\Gamma + \partial_\nu\mathcal{A} y \, dt = (a_2 y_\Gamma + B_2 \cdot \nabla_\Gamma y_\Gamma) \, dt + v_2 \, dW(t) & \textnormal{on} \; \Sigma, \ y_\Gamma = y|_\Gamma & \textnormal{on} \; \Sigma, \end{cases} \end{equation*} remains an open problem.

1.4:

$\begin{cases} \begin{array}{ll} dy - \nabla\cdot(\mathcal{A}\nabla y) \,dt = (a_1 y + B_1 \cdot \nabla y + \mathbbm{1}_{G_0} u) \,dt + v_1 \,dW(t) &\text{in}\,\, Q,\\ dy_\Gamma - \nabla_\Gamma \cdot (\mathcal{A}_\Gamma \nabla_\Gamma y_\Gamma) \,dt + \partial_\nu^\mathcal{A} y \,dt = (a_2 y_\Gamma + B_2 \cdot \nabla_\Gamma y_\Gamma) \,dt + v_2 \,dW(t) &\text{on}\,\, \Sigma,\\ y_\Gamma = y|_\Gamma &\text{on}\,\, \Sigma,\\ (y, y_\Gamma)|_{t=0} = (y_0, y_{\Gamma,0}) &\text{in}\,\, G \times \Gamma, \end{array} \end{cases} $

Carleman Estimates and Controllability of Forward Stochastic Parabolic Equations with General Dynamic Boundary Conditions (2510.12345 - Boulite et al., 14 Oct 2025) in Remark, Section 1 (Introduction and Main Results)