Vectorial Bernoulli Free Boundary Problem

• The vectorial Bernoulli free boundary problem examines minimizers of vector-valued energies, balancing gradient energy with a measure penalty to form free boundaries.
• It employs variational methods, viscosity solutions, and blow-up analysis to achieve regularity results and decompose the free boundary into regular and singular parts.
• Extensions to nonlocal, thin, and metric-space settings reveal broader applications and open problems in singularity classification and shape optimization.

The vectorial Bernoulli free boundary problem concerns the structure and regularity of free boundaries arising for minimizers of vector-valued variational energies that generalize the classical scalar Bernoulli or one-phase Alt–Caffarelli functional to systems. The central object of study is a vector-valued function $U: D \to \mathbb{R}^k$ (with $D\subset\mathbb{R}^d$ a bounded domain) that minimizes an energy of the form

$$\mathcal{J}(U,D) = \int_D |\nabla U|^2 \, dx + \Lambda\, |\{|U| > 0\} \cap D|$$

among Sobolev maps with prescribed boundary data, where $|U|$ is the Euclidean norm. The problem extends naturally to non-Euclidean settings, nonlocal or thin-variant energies, and is linked both to PDE regularity theory and to shape optimization. The free boundary refers to $\partial\Omega_U$, where $\Omega_U = \{|U|>0\}$; its analytic, geometric, and measure-theoretic properties exhibit rich behaviors extending and complicating those of the scalar case.

1. Mathematical Formulation and Variational Structure

Given a bounded Lipschitz domain $D\subset \mathbb{R}^d$, integer $k\geq 1$, and parameter $\Lambda>0$, one considers the minimization problem: $$\min_{ U \in H^1(D;\mathbb{R}^k),\, U=G \text{ on } \partial D} \mathcal{J}(U,D)$$ with

$$\mathcal{J}(U,D) = \int_D |\nabla U|^2\,dx + \Lambda\, |\{|U| > 0\} \cap D|$$

where $$G = (g_1,\ldots,g_k) \in H^{1/2}(\partial D;\mathbb{R}^k)$$. The positivity set is $\Omega_U := \{|U|>0\}$.

The Euler–Lagrange system associated to this variational problem, in the classical regions, reads: $$\begin{cases} \Delta u_i = 0 & \text{in } \Omega_U, \;\; i=1,\ldots,k \ U = G & \text{on } \partial D \ |\nabla |U|| = \sqrt{\Lambda} & \text{on } D\cap\partial\Omega_U \end{cases}$$ The free boundary condition is interpreted either in a weak (viscosity) sense or via inner variations, making this fundamentally a non-linear and non-uniform elliptic free boundary system (Tortone et al., 10 Oct 2025).

2. Weak and Viscosity Solution Frameworks

The natural function space is $U \in H^1(D;\mathbb{R}^k)$; minimizers are locally Lipschitz. The free boundary condition for $|U|$ holds in a scalar viscosity sense: if $\varphi\in C^2$ touches $|U|$ from above (resp. below) at a free boundary point, then $|\nabla\varphi| \geq \sqrt{\Lambda}$ (resp. $\leq \sqrt{\Lambda}$). This generalizes the Caffarelli viscosity solution framework from the scalar one-phase problem to genuinely vector-valued systems, and is further refined for multi-component maps by requiring the applicable test directions to span and probe the sphere $S^{k-1}$ in component space (Silva et al., 2019, Tortone et al., 10 Oct 2025).

This approach accommodates the absence of a sign-preserving property and the possible local vanishing of components, sharply distinguishing the vectorial theory from scalar theory.

3. Regularity, Free Boundary Decomposition, and Flatness Improvement

Any minimizer $U$ exhibits the following properties:

• Local Lipschitz regularity of all components $u_i$.
• The positivity set $\Omega_U$ has locally finite perimeter.
• The free boundary $\partial\Omega_U\cap D$ decomposes into three disjoint sets, according to Lebesgue density $$\theta(x) := \lim_{r\to0} |\Omega_U \cap B_r(x)|/|B_r|$$:
  • Regular set $\Reg(\partial\Omega_U):$ $\theta(x)=1/2$, relatively open and locally $C^{1,\alpha}$.
  • One-phase singular set $\Sing_1(\partial\Omega_U):$ $1/2 < \theta(x) < 1$, closed with Hausdorff dimension at most $d-d^*$, with $d^*\in\{5,6,7\}$.
  • Two-phase (branching) set $\Sing_2(\partial\Omega_U):$ $\theta(x)=1$, relatively closed, rectifiable of dimension at most $d-1$, and admits a finer stratification by the rank of linear blow-ups (Mazzoleni et al., 2018, Tortone et al., 10 Oct 2025, Siclari et al., 31 Jan 2026).

On the regular part, the improvement of flatness method applies: if the free boundary in $B_1$ is sufficiently flat in the geometric sense,

$$|U(x) - f(x\cdot e)_+| \leq \epsilon \text{ in } B_1,\;\; |U| \equiv 0 \text{ in } \{x\cdot e < -\epsilon\}$$

for some unit vectors $e\in\mathbb{R}^d$, $f\in\mathbb{R}^k$, then inside smaller balls the approximation improves, leading by iteration to $C^{1,\alpha}$ regularity (Silva et al., 2019). This is established using a vectorial Harnack inequality, blow-up compactness arguments, and regularity for harmonic/Neumann–Dirichlet systems in half-spaces.

4. Blow-up Analysis, Monotonicity, and Free Boundary Stratification

A central analytical tool is the Weiss monotonicity formula, which for all $x_0\in\partial\Omega_U$ and $r>0$ is

$$W(U,x_0,r) = \frac{1}{r^d}\int_{B_r(x_0)}|\nabla U|^2\,dx + \Lambda\,|\Omega_U\cap B_r(x_0)| - \frac{1}{r^{d+1}} \int_{\partial B_r(x_0)}|U|^2\,d\mathcal{H}^{d-1}$$

with $W(U,x_0,\cdot)$ nondecreasing in $r$ (Mazzoleni et al., 2018, Tortone et al., 10 Oct 2025). Blow-up limits at free boundary points are homogeneous minimizers of the same energy in $\mathbb{R}^d$ and yield the trichotomy in the free-boundary decomposition:

• Regular points: blow-up is a one-phase planar profile $U_0(y) = g (y\cdot\nu)_+$.
• One-phase singular points: blow-up is $U_0(y) = g \, u(y)$, $u$ a scalar one-phase cone.
• Two-phase (branching) points: blow-up is linear, $U_0(y) = A y$, with stratification according to rank$(A)$; each stratum is rectifiable and of codimension dictated by the rank (Mazzoleni et al., 2018, Philippis et al., 2021, Siclari et al., 31 Jan 2026).

Uniqueness and rectifiability of blow-ups in the two-phase regime are established using Alt–Caffarelli–Friedman monotonicity functionals and quantitative stratification methods (Philippis et al., 2021).

5. Extensions: Nonlocal/Thin Problems and Metric Measure Spaces

The vectorial Bernoulli framework encompasses several generalizations:

• The “thin” Bernoulli problem, minimizing

$$\mathcal{J}(G,\Omega) = \int_\Omega |\nabla G|^2 dX + \mathcal{L}_n(\{x\in\Omega\cap\{x_{n+1}=0\}: |G(x,0)|>0\})$$

exhibits existence, optimal regularity, and a similar dichotomy of regular/singular free boundary parts, though without branching points—the phases separate and singular cones only arise in high dimension (Silva et al., 2020).

• On metric measure spaces with Riemannian curvature-dimension bounds (RCD$(K,N)$), the variational structure is retained, with analogous regularity, nondegeneracy, and dimension reduction for the singular set—mirroring the Euclidean theory (Chan et al., 2021).

A summary table of free boundary regularity in representative settings:

Context	Regularity of $\Reg$	Nature of Singular Set
Local, smooth	$C^{1,\alpha}$, $C^\infty$	Disjoint, stratified by blow-up type
Thin/local	$C^{1,\alpha}$	No two-phase (branching) points
Metric-space (RCD)	Reifenberg, $C^{0,\alpha}$	Hausdorff codim $\ge 3$ for singularities

6. Analytical Techniques and Key Formulas

The analysis employs:

• Variational methods: outer and inner variations to extract stationarity and stability.
• Monotonicity formulas: Weiss energy and ACF functionals to control blow-up limits and stratification by symmetry and rank.
• Boundary Harnack principle and NTA domain theory: used in scalar reduction approaches for the regular set (Mazzoleni et al., 2018, Silva et al., 2019).
• Viscosity improvement-of-flatness: direct for vectorial maps (without reduction to scalar case) (Silva et al., 2019).
• Quantitative stratification (Naber–Valtorta): for the rectifiability and structure of singular sets (Philippis et al., 2021).

Key formulas central to the theory include the Weiss energy, the ACF monotonicity functional,

$$\Phi(r; U, \sigma) = r^{-4} \int_{B_r} \frac{|\nabla (U\cdot \sigma)^+|^2}{|x-x_0|^{d-2}} dx \cdot \int_{B_r} \frac{|\nabla (U\cdot \sigma)^-|^2}{|x-x_0|^{d-2}} dx,$$

and capacitary characterizations of critical Bernoulli constants for global minimizers.

7. Open Problems and Recent Developments

Major open directions include:

• Sharp estimates on the Hausdorff dimensions of singular sets in degenerate or non-sign-definite settings.
• Classification and regularity of high-rank (branching/two-phase) blow-ups, including $C^1$ rectifiability of strata of $\Sing_2$.
• Regularity up to the fixed boundary for eigenfunction-driven Bernoulli problems.
• Analysis for fractional/nonlocal operators and the corresponding changes in regularity and singularity classes.
• Stability and asymptotics of minimizers with measure constraints, particularly in the singular, high-density regime (Siclari et al., 31 Jan 2026, Tortone et al., 10 Oct 2025).

Recent work has completed the classification of blow-up limits at two-phase points, providing a precise threshold for the Bernoulli constant $\Lambda^*(A)$ associated to linear profiles, revealing a spectrum of behaviors tied to the rank and structure of the matrix $A$ (Siclari et al., 31 Jan 2026). In metric measure space extensions, dimension reduction and stratification results parallel the Euclidean regime (Chan et al., 2021). Viscosity-based approaches have established $C^{1,\alpha}$ regularity for regular free boundary parts without sign conditions, using intrinsic vectorial arguments (Silva et al., 2019).

These results position the vectorial Bernoulli free boundary problem as a rich, deeply structured class within the wider free boundary and shape optimization landscape, central both for mathematical theory and modeling multicomponent phenomena in applied contexts (Tortone et al., 10 Oct 2025).