Alt-Caffarelli-Type Functionals Overview

• Alt-Caffarelli-type functionals are variational free-boundary models that blend an elliptic bulk term with a measure penalty to define an unknown positivity set.
• They extend the classical Bernoulli free-boundary framework to include two-phase, higher-order, anisotropic, and nonlocal problems via techniques like blow-up analysis and Weiss monotonicity.
• Applications range from combustion-type singular perturbation limits to biharmonic and anisotropic models, advancing regularity and uniqueness results in free-boundary theory.

Alt–Caffarelli-type functionals are variational free-boundary energies that combine an elliptic bulk term with a measure penalty for an active phase, classically

$$\mathcal{E}(u):=\int_D |\nabla u|^2\,dx+|\{u>0\}\cap D|,\qquad u\in H^1(D;\mathbb R^+),$$

and, in later developments, through two-phase, higher-order, anisotropic, nonlocal, and almost-minimizing analogues. Their common structure is that the positivity set is not prescribed a priori: the Euler–Lagrange theory therefore couples an interior PDE in the phase region with a Bernoulli-type condition on an unknown free boundary. In the classical one-phase case minimizers satisfy $u\ge0$, $\Delta u=0$ in $\{u>0\}$, and $|\nabla u|=1$ on $\partial\{u>0\}$, while in the two-phase Alt–Caffarelli–Friedman setting a scale-invariant product of weighted Dirichlet energies governs interfaces between disjoint phases (Engelstein et al., 2018, Allen et al., 2022).

1. Classical variational model and singular perturbation origins

The foundational object is the one-phase Bernoulli functional

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

posed on a connected Lipschitz domain $D\subset\mathbb R^d$. Minimizers are defined by comparison with all competitors sharing the same boundary values, and the classical Alt–Caffarelli theory identifies the weak free-boundary problem

$$u\ge0,\qquad \Delta u=0\ \text{in }\{u>0\}\cap D,\qquad |\nabla u|=1\ \text{on }\partial\{u>0\}\cap D$$

as its Euler–Lagrange system (Engelstein et al., 2018). In this formulation the free boundary is the interface where harmonicity in the positive phase meets a geometric jump condition.

A distinct but closely related entry point is the singular perturbation problem

$$\Delta u_\varepsilon=\beta_\varepsilon(u_\varepsilon)\quad\text{in }B_1,\qquad |u_\varepsilon|\le1,$$

with $u\ge0$0, $u\ge0$1, $u\ge0$2, and $u\ge0$3. In the limit $u\ge0$4, solutions converge to stationary points of the Alt–Caffarelli functional

$u\ge0$5

so the one-phase Bernoulli condition appears as the singular perturbation limit of combustion-type equations. The same framework also makes clear that stationary points are more general than minimizers: they remain harmonic off the free boundary but need not inherit the full minimizing structure of the original Alt–Caffarelli problem (Karakhanyan, 2017).

This variational core persists through most later generalizations. Whether the bulk term is $u\ge0$6, $u\ge0$7, $u\ge0$8, or an Orlicz integral $u\ge0$9, the characteristic feature is the competition between PDE energy and a phase-measure term. That competition is what produces a free boundary and makes blow-up analysis, monotonicity formulas, and rigidity the central tools of the subject.

2. Blow-ups, Weiss monotonicity, and singular free boundaries

For the classical one-phase functional, a minimizer $\Delta u=0$0 at a free-boundary point $\Delta u=0$1 is rescaled by

$\Delta u=0$2

The Weiss functional

$\Delta u=0$3

is monotone nondecreasing in $\Delta u=0$4, its limit $\Delta u=0$5 exists, and every blow-up is a global 1-homogeneous minimizer of the Alt–Caffarelli functional (Engelstein et al., 2018). This converts local free-boundary geometry into a classification problem for global cones.

The regular–singular dichotomy is encoded by the possible blow-ups. A point is regular if some blow-up is a half-space solution; then the free boundary is $\Delta u=0$6 nearby. A point is singular if no blow-up is of half-space type. The known dimension restrictions are sharp in the current literature: there are no non-flat 1-homogeneous minimizers for $\Delta u=0$7, whereas in $\Delta u=0$8 De Silva–Jerison constructed minimizing cones with isolated singularity at the origin. Those cones have positivity set

$\Delta u=0$9

a symmetric double cone whose free boundary is smooth away from the vertex (Engelstein et al., 2018).

The main singular regularity theorem available in the provided literature states that if one blow-up at a singular point is a cone with isolated singularity, then the blow-up is unique and the free boundary is a $\{u>0\}$0 graph over that cone near the singular point. When the cone is “integrable through rotations,” the regularity improves to $\{u>0\}$1; the paper further proves that De Silva–Jerison cones satisfy this integrability condition, hence enjoy uniqueness of blow-up and $\{u>0\}$2 graphical convergence (Engelstein et al., 2018). The analytic mechanism is a local epiperimetric or log-epiperimetric inequality for the Weiss energy, obtained through a reduction to an analytic functional on the sphere and a Łojasiewicz inequality for its finite-dimensional kernel.

A complementary rigidity principle comes from the strong maximum principle for one-phase minimizers. If $\{u>0\}$3 are minimizers and their regular free boundaries do not intersect, then their full free boundaries cannot touch. This yields a Hardt–Simon-type foliation associated to any 1-homogeneous global minimizer: there exist global minimizers lying on either side of the cone whose dilates foliate the complement of the conical free boundary (Edelen et al., 2022). This places 1-homogeneous global minimizers in a role analogous to area-minimizing cones in geometric measure theory.

3. Alt–Caffarelli–Friedman functionals and two-phase interface geometry

The two-phase branch of the theory is organized around the Alt–Caffarelli–Friedman functional. For nonnegative continuous functions $\{u>0\}$4 satisfying

$\{u>0\}$5

the ACF functional is

$\{u>0\}$6

It is nondecreasing in $\{u>0\}$7, and constancy characterizes the model pair of complementary truncated linear functions $\{u>0\}$8 and $\{u>0\}$9 (Allen et al., 2022).

The distinguished set is

$|\nabla u|=1$0

the portion of the interface where the ACF limit is strictly positive. On $|\nabla u|=1$1, the interface admits a sharp geometric description. It is $|\nabla u|=1$2-rectifiable, and for $|\nabla u|=1$3-almost every $|\nabla u|=1$4 the Lipschitz rescalings converge strongly in $|\nabla u|=1$5 to a unique pair of nondegenerate truncated linear functions

$|\nabla u|=1$6

with $|\nabla u|=1$7 and $|\nabla u|=1$8 (Allen et al., 2022). The proof combines a sharp quantitative remainder term in the ACF monotonicity formula with the Naber–Valtorta rectifiable Reifenberg framework.

The spectral side of the ACF formula is encoded by characteristic constants of spherical phases. If $|\nabla u|=1$9 is the spherical positivity set of phase $\partial\{u>0\}$0 on $\partial\{u>0\}$1, and $\partial\{u>0\}$2, then

$\partial\{u>0\}$3

This relation is the basis for a higher-dimensional analogue of Carleson’s $\partial\{u>0\}$4-conjecture: at tangent points of pairs of Wiener-regular domains, finiteness of

$\partial\{u>0\}$5

is forced by the ACF monotonicity formula (Fleschler et al., 2023).

The two-phase theory also exhibits sharp limitations. Positivity of $\partial\{u>0\}$6 at a single point does not imply tangent planes or unique blow-ups; the literature provided includes a spiraling planar example with $\partial\{u>0\}$7 but no approximate tangent and non-unique blow-ups. Likewise, rectifiability of an interface does not imply positivity of the ACF limit almost everywhere: the counterexample built from two planar Wiener-regular domains with common rectifiable boundary portion $\partial\{u>0\}$8 satisfies $\partial\{u>0\}$9 for $$\mathcal{E}(u)=\int_D |\nabla u|^2\,dx+|\{u>0\}\cap D|,$$0-almost every $$\mathcal{E}(u)=\int_D |\nabla u|^2\,dx+|\{u>0\}\cap D|,$$1 (Allen et al., 2022, Fleschler et al., 2023). These examples delimit the scope of purely pointwise ACF criteria.

4. Higher-order and anisotropic generalizations

A substantial extension replaces the Dirichlet term by fourth-order bending energies. In two dimensions, one biharmonic Alt–Caffarelli problem uses

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

with Navier boundary conditions. Minimizers are biharmonic in $$\mathcal{E}(u)=\int_D |\nabla u|^2\,dx+|\{u>0\}\cap D|,$$3, the free boundary is the nodal set $$\mathcal{E}(u)=\int_D |\nabla u|^2\,dx+|\{u>0\}\cap D|,$$4, interior regularity reaches $$\mathcal{E}(u)=\int_D |\nabla u|^2\,dx+|\{u>0\}\cap D|,$$5, and the free boundary consists of finitely many $$\mathcal{E}(u)=\int_D |\nabla u|^2\,dx+|\{u>0\}\cap D|,$$6-hypersurfaces. The same work proves that minimizers are in general not unique and computes radial minimizers explicitly (Müller, 2020).

A later biharmonic analogue adopts

$$\mathcal{E}(u)=\int_D |\nabla u|^2\,dx+|\{u>0\}\cap D|,$$7

under either Dirichlet or Navier boundary conditions. In this model the flat region

$$\mathcal{E}(u)=\int_D |\nabla u|^2\,dx+|\{u>0\}\cap D|,$$8

is the free region, minimizers are biharmonic on $$\mathcal{E}(u)=\int_D |\nabla u|^2\,dx+|\{u>0\}\cap D|,$$9, and the optimal regularity is $D\subset\mathbb R^d$0 rather than $D\subset\mathbb R^d$1. In the Navier case, radial boundary data produce radial minimizers, and explicit radial solutions show that this $D\subset\mathbb R^d$2 regularity is optimal (Grunau et al., 2023).

The two-sided biharmonic functional

$D\subset\mathbb R^d$3

has a markedly different global theory. Half-space solutions of the form $D\subset\mathbb R^d$4 are global minimizers for the two-sided problem, but not for the one-sided functional. A second class of global minimizers is given by functions with constant Laplacian $D\subset\mathbb R^d$5 for $D\subset\mathbb R^d$6. At the same time, minimizers of the two-sided biharmonic problem need not satisfy any PDE, not even with a signed Radon measure on the right-hand side, which the paper emphasizes is in sharp contrast with the one-sided problem (Grunau et al., 15 Jan 2026).

The anisotropic higher-order model replaces $D\subset\mathbb R^d$7 by $D\subset\mathbb R^d$8 and studies

$D\subset\mathbb R^d$9

In dimension two, with $$u\ge0,\qquad \Delta u=0\ \text{in }\{u>0\}\cap D,\qquad |\nabla u|=1\ \text{on }\partial\{u>0\}\cap D$$0 smooth, symmetric, and uniformly elliptic, minimizers are $$u\ge0,\qquad \Delta u=0\ \text{in }\{u>0\}\cap D,\qquad |\nabla u|=1\ \text{on }\partial\{u>0\}\cap D$$1, smooth away from $$u\ge0,\qquad \Delta u=0\ \text{in }\{u>0\}\cap D,\qquad |\nabla u|=1\ \text{on }\partial\{u>0\}\cap D$$2, and the zero set is a $$u\ge0,\qquad \Delta u=0\ \text{in }\{u>0\}\cap D,\qquad |\nabla u|=1\ \text{on }\partial\{u>0\}\cap D$$3 curve; the negative phase is a finite union of $$u\ge0,\qquad \Delta u=0\ \text{in }\{u>0\}\cap D,\qquad |\nabla u|=1\ \text{on }\partial\{u>0\}\cap D$$4-domains compactly contained in $$u\ge0,\qquad \Delta u=0\ \text{in }\{u>0\}\cap D,\qquad |\nabla u|=1\ \text{on }\partial\{u>0\}\cap D$$5. The structural tool is an anisotropic Frehse-type decomposition of $$u\ge0,\qquad \Delta u=0\ \text{in }\{u>0\}\cap D,\qquad |\nabla u|=1\ \text{on }\partial\{u>0\}\cap D$$6 into a singular scalar part proportional to $$u\ge0,\qquad \Delta u=0\ \text{in }\{u>0\}\cap D,\qquad |\nabla u|=1\ \text{on }\partial\{u>0\}\cap D$$7 plus bounded-kernel potentials of the free-boundary measure (Müller, 27 May 2025).

Setting	Functional	Representative conclusion
Biharmonic one-phase	$$u\ge0,\qquad \Delta u=0\ \text{in }\{u>0\}\cap D,\qquad |\nabla u|=1\ \text{on }\partial\{u>0\}\cap D$$8	$$u\ge0,\qquad \Delta u=0\ \text{in }\{u>0\}\cap D,\qquad |\nabla u|=1\ \text{on }\partial\{u>0\}\cap D$$9 minimizers in 2D; finitely many $$\Delta u_\varepsilon=\beta_\varepsilon(u_\varepsilon)\quad\text{in }B_1,\qquad |u_\varepsilon|\le1,$$0 free-boundary components (Müller, 2020)
Biharmonic two-sided	$$\Delta u_\varepsilon=\beta_\varepsilon(u_\varepsilon)\quad\text{in }B_1,\qquad |u_\varepsilon|\le1,$$1	half-space and constant-Laplacian global minimizers; no general measure-valued PDE (Grunau et al., 15 Jan 2026)
Anisotropic higher-order	$$\Delta u_\varepsilon=\beta_\varepsilon(u_\varepsilon)\quad\text{in }B_1,\qquad |u_\varepsilon|\le1,$$2	smooth anisotropy preserves optimal $$\Delta u_\varepsilon=\beta_\varepsilon(u_\varepsilon)\quad\text{in }B_1,\qquad |u_\varepsilon|\le1,$$3-regularity in 2D (Müller, 27 May 2025)

These higher-order problems show that the Alt–Caffarelli mechanism survives beyond second order, but with qualitatively new features: sign changes, flat cores, measure terms on $$\Delta u_\varepsilon=\beta_\varepsilon(u_\varepsilon)\quad\text{in }B_1,\qquad |u_\varepsilon|\le1,$$4, and the possible failure of any simple PDE characterization for global minimizers.

5. Almost-minimizers, weak solutions, and boundary regularity

A second major direction replaces exact minimizers by almost-minimizers. For variable-coefficient one-phase and two-phase Bernoulli energies

$$\Delta u_\varepsilon=\beta_\varepsilon(u_\varepsilon)\quad\text{in }B_1,\qquad |u_\varepsilon|\le1,$$5

almost-minimality is imposed with an additive error $$\Delta u_\varepsilon=\beta_\varepsilon(u_\varepsilon)\quad\text{in }B_1,\qquad |u_\varepsilon|\le1,$$6 on balls or coefficient-adapted ellipsoids. In this setting, almost minimizers are continuous, $$\Delta u_\varepsilon=\beta_\varepsilon(u_\varepsilon)\quad\text{in }B_1,\qquad |u_\varepsilon|\le1,$$7 inside each phase, and locally Lipschitz across the free boundary in both one- and two-phase problems. The two-phase argument requires an almost-monotone Alt–Caffarelli–Friedman formula adapted to variable coefficients (David et al., 2019).

The Orlicz and vectorial theory generalizes the gradient term further. For

$$\Delta u_\varepsilon=\beta_\varepsilon(u_\varepsilon)\quad\text{in }B_1,\qquad |u_\varepsilon|\le1,$$8

where $$\Delta u_\varepsilon=\beta_\varepsilon(u_\varepsilon)\quad\text{in }B_1,\qquad |u_\varepsilon|\le1,$$9 is a Young function in the Lieberman class $u\ge0$00, weakly coupled vectorial almost-minimizers are shown to be Hölder continuous, $u\ge0$01 in the non-coincidence set under additional structural assumptions, and locally Lipschitz in the interior. The same framework yields universal gradient estimates for nonnegative almost-minimizers in non-coincidence regions (Pontes et al., 21 Jun 2025). A subsequent paper proves optimal Lipschitz continuity up to the boundary for $u\ge0$02-domains, extending the interior theory to boundary points and free-boundary contact points (Pontes et al., 7 Dec 2025).

The weak-solution theory for generalized one-phase Bernoulli problems also advances the classical picture. For

$u\ge0$03

the functional

$u\ge0$04

provides the Alt–Caffarelli-type variational structure. Weak solutions are shown to be stationary solutions of this functional; under a density condition on the zero phase, blow-ups are weak solutions of the homogeneous problem and are 1-homogeneous. Most significantly, if the free boundary is Lipschitz and $u\ge0$05 are smooth, then the free boundary is actually $u\ge0$06. This extends to weak solutions a regularity theorem previously known only for viscosity solutions, and it yields an alternative proof of Serrin’s problem for Lipschitz domains as well as a Poisson-kernel characterization of boundary smoothness (Domingo-Pasarin et al., 28 Jan 2026).

These almost-minimizer and weak-solution results show that the regularity theory of Alt–Caffarelli-type functionals is not confined to exact minimization. It remains effective under coefficient oscillation, non-standard growth, vectorial weak coupling, and distributional formulations, provided one retains sufficiently strong comparison principles and blow-up compactness.

6. Nonlocal, non-Euclidean, and geometric reinterpretations

The nonlocal extension replaces $u\ge0$07 by intrinsically nonlocal energies associated with the fractional Laplacian. One paper introduces one-phase fractional ACF-type functionals

$u\ge0$08

where $u\ge0$09 is the fractional Poisson kernel and $u\ge0$10 is either a Gagliardo-type quadratic density $u\ge0$11 or the fractional-gradient energy $u\ge0$12. Under $u\ge0$13-subharmonicity of these nonlocal gradient squares, the functionals are monotone increasing, and as $u\ge0$14 they converge to the classical one-phase Alt–Caffarelli functional. The same work derives nonlocal gradient estimates and a nonlocal Bochner identity, all without using the Caffarelli–Silvestre extension (Ferrari et al., 30 Sep 2025).

By contrast, in the Heisenberg group the Euclidean ACF paradigm can fail. The natural Heisenberg analogue of the FeFo companion functional is not monotone increasing in general: for the intrinsic harmonic polynomial

$u\ge0$15

the associated functional is strictly decreasing near the origin. Consequently, the Heisenberg ACF-type functional with the natural exponent $u\ge0$16 is not universally monotone, showing that Euclidean ACF theory does not transplant directly to the noncommutative setting (Ferrari et al., 2022).

A different geometric reinterpretation comes from singular perturbation limits in three dimensions. Blow-ups of stationary limits of

$u\ge0$17

are 1-homogeneous, and their spherical trace $u\ge0$18 satisfies $u\ge0$19. The associated map

$u\ge0$20

parametrizes a minimal immersion inside the sphere of radius $u\ge0$21, meeting the sphere orthogonally. Disk-type positivity sets yield half-plane solutions, while ring-type sets produce the Alt–Caffarelli catenoid cone (Karakhanyan, 2017). This establishes a direct bridge between stationary Alt–Caffarelli-type free boundaries and capillary minimal surfaces.

Several open directions recur across the literature. The classical singular theory leaves open the classification of minimizing cones with isolated singularities beyond the De Silva–Jerison family, the structure of higher-dimensional singular strata, and extensions to almost minimizers or two-phase and vector-valued Bernoulli functionals (Engelstein et al., 2018). The ACF program raises corresponding questions for more general elliptic operators, dynamic free boundaries, and converse rectifiability criteria (Allen et al., 2022). In higher-order settings, boundary regularity in dimensions $u\ge0$22, lower regularity assumptions on anisotropies, and complete classification of global minimizers remain unresolved (Müller, 27 May 2025, Grunau et al., 15 Jan 2026). Taken together, these problems indicate that Alt–Caffarelli-type functionals form a unified but still rapidly developing framework for variational free-boundary analysis across local, nonlocal, geometric, and higher-order regimes.