Why monotone Allen–Cahn transitions must be planar through dimension eight

← All digests

Family 375 · Partial differential equations
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-26

1. The question

Imagine a material with two preferred phases, labeled $-1$ and $1$. A transition between them occupies a thin layer. The simplest equilibrium has parallel, flat layers: the phase changes only with distance in one direction.

Could an equilibrium transition be curved everywhere, even if the phase always increases as we move upward? This is the question behind De Giorgi’s conjecture for the Allen–Cahn equation. It connects a nonlinear differential equation to the geometry of minimal surfaces.

The manuscript in this family claims that every bounded equilibrium in eight-dimensional Euclidean space, strictly increasing in one fixed direction, is a flat one-dimensional transition. It assumes neither a bound on the amount of transition energy in large balls nor prescribed phases at the two ends of that direction. Those properties are conclusions of its argument.

Eight is significant. Nonplanar monotone solutions are already known in dimensions nine and higher. A second threshold is equally important: the manuscript claims a classification of stable solutions in dimension seven, while stable nonplanar solutions are known in dimension eight. Monotonicity and stability are related conditions, but they have different endpoints.

2. Background and essential definitions

A diffuse interface equation

The scalar Allen–Cahn equation considered here is

$$\Delta u=u^3-u.$$

Here $u:\mathbb R^n\to[-1,1]$ describes the phase and $\Delta$ is the sum of the second derivatives. The potential

$$W(s)=\frac14(1-s^2)^2$$

has minima at the two preferred values. On a region $\Omega$, the energy is

$$E(u;\Omega)=\int_\Omega\left(\frac12|\nabla u|^2+W(u)\right)\,dx.$$

The first term discourages abrupt spatial changes; the second discourages intermediate phase values. Equilibria satisfy the displayed differential equation.

An explicit transition is $g(s)=\tanh(s/\sqrt2)$. If $e$ is a unit vector and $c$ a real number, then

$$u(x)=g(e\cdot x-c)$$

is a solution. Its level sets are parallel hyperplanes. Such a solution is called a planar heteroclinic: it connects the two constant equilibria as its one relevant coordinate runs from one end of the real line to the other. “Entire” means defined on all of $\mathbb R^n$, with no finite boundary.

Stability versus minimization

A solution $v$ is stable if small, compactly supported perturbations cannot lower its energy to second order. Precisely, every continuously differentiable test function $\varphi$ of compact support must satisfy

$$Q_v(\varphi)=\int_{\mathbb R^n}\left(|\nabla\varphi|^2+(3v^2-1)\varphi^2\right)\,dx\geq0.$$

A global minimizer satisfies a stronger requirement: in every ball it has least energy among all Sobolev competitors with the same boundary values. Stability tests the quadratic response near a solution; minimization compares it with potentially large changes.

Strict monotonicity means $\partial_nu>0$ everywhere, after choosing coordinates. This positive derivative solves the linearized equation and supplies stability. But a general stable solution need not have a preferred direction of increase.

Energy density

A planar transition has infinite energy in the whole space, but its energy in a ball of radius $R$ grows like surface area, $R^{n-1}$. The normalized density is

$$M_R(y)=\frac{E(v;B_R(y))}{s_0\omega_{n-1}R^{n-1}}, \qquad s_0=\frac{2\sqrt2}{3}.$$

Here $B_R(y)$ is the ball centered at $y$, $\omega_{n-1}$ is the volume of the unit ball in $\mathbb R^{n-1}$, and $s_0$ is the energy per unit area of the one-dimensional transition. Its limiting density is one. A finite-density assumption bounds $\sup_{R>0}M_R(0)$; it does not require finite total energy.

3. Development history and previous partial results

De Giorgi proposed the monotonicity question in the late 1970s, in the setting of phase transitions and the Bernstein problem for minimal graphs. The connection comes from sharpening the interface: appropriately rescaled phase energies converge to surface tension times perimeter.

The geometric analogy contains a crucial distinction. Nonflat minimizing cones occur in ambient dimension eight. Nonaffine entire minimal graphs first occur in ambient dimension nine. A graph has directional structure that an arbitrary minimizing surface lacks; likewise, a monotone phase transition carries more information than stability alone.

Ghoussoub and Gui established the Allen–Cahn conjecture in dimension two in 1998. Ambrosio and Cabré proved the three-dimensional result in 2000. These were classifications without an assumption that the solution already approaches the two wells along the monotone direction.

Savin’s 2009 theorem reached dimensions through eight with an additional hypothesis: for each fixed transverse point $x'$, the solution tends to $-1$ as the last coordinate tends to minus infinity and to $1$ as it tends to plus infinity. These are pointwise directional limits, not a requirement of uniform convergence in $x'$. They supply substantial information about the global interface.

Del Pino, Kowalczyk and Wei subsequently constructed nonplanar strictly monotone solutions in every ambient dimension at least nine, using large dilations of nonaffine minimal graphs. Thus the proposed dimension-eight conclusion is at the expected sharp threshold.

For stability, Pacard and Wei constructed nonplanar stable solutions in dimension eight. Recent work cited by the manuscript includes stable rigidity in dimension three and monotone rigidity in dimension four, obtained independently by Liu, Luo, Wang, Juncheng Wei, Yong Wei and Wu, and by Chan, Fernández-Real, Figalli, Florit-Simon and Serra. Florit-Simon and Serra also proved stable rigidity in dimension four under bounded energy density.

The present manuscript therefore addresses two gaps together: extending the stable classification to dimension seven and dispensing with the density assumption there. That stable theorem then feeds a geometric reduction to the monotone theorem in dimension eight. The earlier reduction through minimizing end profiles is credited to Alberti–Ambrosio–Cabré and Jerison–Monneau.

4. What the new manuscript claims

Monotone rigidity in dimension eight

Theorem 1.1 states that if $u\in C^2(\mathbb R^8,(-1,1))$ satisfies

$$\Delta u=u^3-u,\qquad \partial_8u>0,$$

then there are a unit vector $e\in\mathbb S^7$, with $e_8>0$, and a real number $c$ such that

$$u(x)=\tanh\left(\frac{e\cdot x-c}{\sqrt2}\right).$$

Here $\mathbb S^7$ denotes the unit sphere in $\mathbb R^8$. The theorem does not assume minimization, surface-order energy growth, or directional limits at the wells.

Stable rigidity in dimension seven

Theorem 1.2 states that every stable $C^2$ solution $v:\mathbb R^7\to[-1,1]$ is either the constant $1$, the constant $-1$, or a planar heteroclinic. There is no energy-growth assumption.

A subsidiary theorem first gives the same classification in dimensions four through seven when $\sup_R M_R(0)<\infty$. Removing this extra assumption in dimension seven is a separate part of the manuscript, not a consequence of merely restating the finite-density result.

Together with the known counterexamples, these claims give sharp dimension endpoints for the stated scalar quartic equation: seven for stable rigidity and eight for strictly monotone rigidity.

5. Why the advance matters

Importance: 5/5. If correct, the claims settle a central symmetry problem linking nonlinear elliptic equations, variational phase transitions and minimal-surface geometry. The significance includes both the endpoint conjecture and a stable classification requiring no global energy bound.

A density bound can quietly encode much of the desired simplicity. It limits how many transition layers can accumulate as one looks farther away. Proving rigidity without that bound means controlling configurations whose large-scale complexity has not been restricted in advance.

The two dimension thresholds also clarify a physical and geometric distinction. Resistance to small perturbations does not force the same structure as a globally consistent direction of increase. The manuscript’s conclusion uses this extra directional information at exactly the dimension where stable nonplanar patterns are already possible.

6. The main idea

Measure departure from the one-dimensional profile

For a nonconstant solution, introduce the inverse-profile coordinate $U=g^{-1}(v)$. For a planar transition, $U$ is affine with unit gradient. Set

$$q=1-v^2,\qquad D=1-|\nabla U|^2.$$

Modica’s gradient bound gives $D\geq0$. The quantity $D$ measures failure to follow the ideal one-dimensional profile. Testing stability with $q\eta$, where $\eta$ is a compactly supported cutoff, yields the identity

$$0\leq Q_v(q\eta) =\int q^2|\nabla\eta|^2\,dx-\int q^3D\eta^2\,dx.$$

This converts stability into control of a nonnegative defect. It remains useful near irregular portions of the zero set, where describing the transition as a smooth surface is harder.

Under finite density, the manuscript combines diffuse estimates with descriptions of separated sheets. When sheets lie close together, their interaction constrains the gaps between them. Comparing those constraints with what stability allows gives the finite-density classification through dimension seven. Estimates for transition layers developed in earlier work of Wang and Wei enter after the required local regularity has been obtained.

Obtain the missing density bound

Without finite density, one cannot begin by assuming a uniformly manageable collection of sheets. The manuscript introduces a penalized version of geometric reach: a local scale sensitive both to bending of a sheet and to the proximity of other sheets.

Near that scale, local density control permits use of the preceding regularity estimates. The argument combines curvature information on good interface regions with ambient stability tests that also cover bad regions. Fourth-moment estimates control the accumulation of curvature and profile defect.

A delicate issue is joining test functions on neighboring layers. Joining arbitrary amplitudes could consume the positive margin needed in dimension seven. The construction arranges that the cost depends on squared differences of neighboring amplitudes; close layers receive nearly equal amplitudes. The same radial weight is then used in both the geometric reach estimate and the ambient stability estimate.

According to the manuscript, the remaining positive margin contradicts the amount of interface area that unbounded density would force. This establishes the missing density bound and closes the stable classification. The argument is not simply an appeal to a corresponding minimal-surface theorem.

Recover the monotone theorem

For a monotone solution in eight variables, boundedness supplies end profiles

$$u_\pm(x')=\lim_{t\to\pm\infty}u(x',t), \qquad x'\in\mathbb R^7.$$

They are stable seven-dimensional solutions. At this stage the classification allows planar transitions as well as constant wells; the argument must not assume that the end profiles are already $-1$ and $1$.

Each allowed end profile is a global minimizer. By clipping competitors between the profiles and sliding translates of the original solution, the manuscript applies the classical reduction to show that $u$ is itself a global minimizer. Comparison with a constant phase then supplies $E(u;B_R)\leq CR^7$.

Zooming out gives a minimizing boundary preserved by dilations, called a minimizing cone. The argument identifies the full limiting energy as surface tension times its perimeter, excluding additional hidden energy. Directional monotonicity gives a consistent sign to a component of the cone’s normal.

In ambient dimension eight, the minimizing cone can have at most a singular vertex. The directed-cone argument forces a hyperplane, including the case where the distinguished normal component vanishes identically. Finally, a near-unit-density theorem of Wang returns from this flat limit to the original solution. The one-dimensional equation fixes its profile and orientation.

7. Scope, limitations, and open questions

The claims concern the scalar equation with the particular quartic potential above. They are not stated here for vector-valued systems, arbitrary double-well potentials, fractional operators, or curved ambient spaces.

The monotonicity hypothesis is an everywhere strictly positive derivative in a fixed direction. The result does not classify all bounded eight-dimensional solutions. The stable theorem concerns dimension seven; known stable nonplanar examples prevent extending that conclusion unchanged to dimension eight.

Likewise, known monotone nonplanar examples in dimension nine rule out an unrestricted extension of the monotone theorem. Further questions concern what other hypotheses recover rigidity beyond these endpoints, and which parts of the density-removal method extend to other phase-transition energies.

The manuscript’s mathematical conclusions and analytical mechanisms are reported as stated. This digest explains the role of its assumptions and reductions without independently establishing the claimed classifications.

8. Sources and further reading

• Ghoussoub and Gui, On a conjecture of De Giorgi and some related problems: the two-dimensional theorem.
• Ambrosio and Cabré, Entire solutions of semilinear elliptic equations in three space dimensions and a conjecture of De Giorgi: the three-dimensional theorem.
• Savin, Regularity of flat level sets in phase transitions: the result through dimension eight with directional well limits.
• Del Pino, Kowalczyk and Wei, On De Giorgi’s conjecture in dimension N ≥ 9: nonplanar monotone solutions above the threshold.
• Pacard and Wei, Stable solutions of the Allen–Cahn equation in dimension 8 and minimal cones: the different endpoint for stability.
• Liu and collaborators and Chan and collaborators: recent stable three-dimensional and monotone four-dimensional classifications cited by the manuscript.
• Florit-Simon and Serra, their bounded-density stable rigidity work: a preceding finite-density result.

Source manuscripts

• M1. OpenAI, A positive resolution of De Giorgi’s conjecture in dimension eight (September 26, 2026).

Sources consulted on 26 October 2026 at repository revision 1581d4c. Download the manuscript citations.